new-matter-server

[fringe fjord]

would be interesting to wait tbh ... I would hope they resume from themself with time

[deep drift]

ok, then I wait

[fringe fjord]

I would report that to the eve guys, so knowing" did not fixed itself in 2h" is a better point 🙂

[deep drift]

right. 😉

[fringe fjord]

and as we know especially thread and stuff can take time here and there

[deep drift]

I already had a timmerflotte dropping out. but the thermo immediately recovered. lets see.

[deep drift]

Just to confirm my understanding of Matter binding is correct: via Matter.js's UI we configure the target by defining the node, the endpoint and optional the cluster. In my case Eve Thermo knows the node and the endpoint it is bound to. In order to connect to/poll the target, doesn't it have to communicate to Matter.js first e.g. to get the actual connection details (IP, etc.) or anything else of the specific node? Or is it directly connecting to the node via Thread without involving Matter.js at all?

[fringe fjord]

No, if matter node need to communicate with an other matter node (and here device <--> controlelr or device <--> device is basically the same). if you know an IP then you use it, else MDNS is used to discover ... So matter-server is just setting up the binding and ACLs ... rest happens between the devices themself, so the controller can be offline and the binding will still work

The binding defines the ndoe.id pf the target and so the device can get any IP or anything needed itself.
Try it (when it is back working). turn off the server and force temp change and verify that the thermostat gets new temp ... should work without server being online

[deep drift]

Thanks for the explanation. Now it's clear to me! 🙏

[fringe fjord]

But that also brings problems because right now you do not have any real clue why it oes not work if a binding does not work because you can not look into the device errors or such

[deep drift]

The Matter binding was recovered nearly 2 hours later.

Interestingly the sensor was "turned off" approximately 15 minutes before it became unavailable. So yes, no clue. 😉

[spare wraith]

Just had all mater device on HA become unavailable. Matter work no problem on my other controllers. Not sure why it happened. No clear error message. Looks like it can see the devices on mDNS happen with out there being any changes. Is it a known issue?

[manic lichen]

Are you using the new matter server for testing?

[spare wraith]

Yes did all the updates

So I assume so

[manic lichen]

Erm... just to be sure. This channel is just for testing of the beta versions of the Matter server. If you are having issues with the non-beta Matter server, #1284966617670881350 with the "Matter" tag would be the right place.

[spare wraith]

Probably worng place then...

[sturdy ivy]

I haven't been able to reproduce this blinds problem for the last couple of days, so you should probably ignore me. I actually suspect that this was a side-effect of the problematic Nanoleaf bulbs that we discussed in another thread. I have since replaced some of those with the new Ikea Kajplats bulbs. They are sooo much better (and also more affordable). I think the Thread network clutter is now much reduced which has probably helped with the other Invoke clashes. Anyway, all good for now.

Is it by design that when adding devices to the Matter integration, all its entities are initially disabled? This seems odd. I would expect that for a light for example, at least it's primary light control entity should be enabled when added. Or is it just some setting that I've inadvertently pushed?

[fringe fjord]

What exactly you mean?

[haughty stump]

How do I find or trace these unknown devices? When I click it, it shows it as an unknown router, but I have no matter/thread devices that are not in HA. I do have 4 apple TVs and 4 HomePods, but there are like 20 unknown devices.

[fringe fjord]

There should be a setting per integration how that should behave. Check this maybe

[fringe fjord]

Good question but I basically have no real answer. In fact these are all entries the devices „see“ or have in their neighbor lists and routing table.

[sturdy ivy]

Ok, so it's definitely not a server issue then?

[fringe fjord]

If you like compare the shown data of the nodes that connect to such unknown devices check the data in the dashboard in the thread network diagnostic cluster attribute neightbors.

But honestly I am unsure if fighting the unknowns is really the point of the chart. Honestly maybe better check out the devices or „many connection unknowns“ that have a red color 😉

[silk plover]

I am curious how many devices are listed as potential TBR under HA/Settings/thread?

[haughty stump]

6 TBRs 5 apple 1 HA OTBR

[silk plover]

Ok that only accounts for 12 of those unknown devices

[heady token]

batt powered devices can take up to 2 hours

[hollow cloak]

@signal drift I converted your message into a file since it's above 15 lines :+1:

[signal drift]

I am adding discory schema for Heiman clustom attribute, but it can not be discoveried by HA core, anyone konws that? thanks

the matter server currently is matterjs-server

even if I replaced the attributes with existing required_attributes=(clusters.SmokeCoAlarm.Attributes.DeviceMuted,),, it still can not shows there

[fringe fjord]

@signal drift As already told ... when chaning this we also need to update the python client but thats currently not added for the new matterjs server stuff, so right now thats a blocker

For all with Ikea "devices going offline till we resubscribe and then work again short time" please check https://github.com/matter-js/matterjs-server/issues/187#issuecomment-3852760704 and provide information if any devices beside MYGGBETT, MYGGSPRAY, Bilresa. thanks. We are currently supporting Ikea to find the cause.

[signal drift]

@fringe fjord Thanks for th reply, even though to use the standard attributes for testing? like clusters.SmokeCoAlarm.Attributes.DeviceMuted

[deep drift]

I just added yesterday’s Matter binding issue - originally caused by two TIMMERFLOTTEs going unresponsive and/or offline - as a comment to the GitHub issue.

[fringe fjord]

Yeahh but going fully offline right? Then other issue

[deep drift]

True. I removed it.

[silk plover]

Any plans to change the Matter server dashboard icon to the matter symbol in future Matter-server app updates?

[raven harbor]

Is it possible to combine info from OTBR REST API to fill part of them

IKEA Dirigera even offers full TLV Dataset in the app, importing should be easier

[junior sedge]

That's something that's probably better implemented in Home Assistant itself rather than the matter server, since Home Assistant knows about OTBR, it knows about Matter over Thread, it knows about Homekit over Thread, and in the future might even know about ESPHome over Thread.

so combining multiple sources of info seems like it would fit better there

[fringe fjord]

I think I said before … please create an issue for the addon.

[silk plover]

Will do did not see a prior ask/request.

[fringe fjord]

In future we plan to have an enhanced variant of the graph in HA and then we can enrich such infos. But. Only when HA orbr is used. Others are still unknown

[fringe fjord]

Or directly do a Pr 😉

[raven harbor]

That would be good 👍

[junior sedge]

It might be possible to include info from other OTBR based TBRs too, but until recently they haven't really considered the diagnostic related apis to be stable so there's a few different apis on different implementations that might need to be supported :/ (and of course a lot of commercial devices disable the apis)

[heady token]

app lol

[silk plover]

App not add-on issue created 🤪 , I think. https://github.com/home-assistant/addons/issues/4391

[heady token]

noice

[deep drift]

I already had created an issue 2 weeks ago.

https://github.com/home-assistant/addons/issues/4308

[silk plover]

I’ll reference yours and try to get mine closed.

[fringe fjord]

Soo, then here also my reminder for the upcoming Matter Community Sync as open meeting using Google Meet coming/next Monday at 19:00-20:00 CET. 🎉

Google Meet Link: https://meet.google.com/uch-rsxw-txh

If you want to hear the latest news for the OHF Matter Server, matter.js, the HA Matter integration and Matter in general, join, bring in your questions or information and lets sync!

[raw veldt]

I've been having some weird OTA behavior with one matter-over-thread device with the new server. I'm not sure what to make of it. I think it might be a bug but perhaps it's a local problem. I have one battery operated device (an Eve Motion) that is really fixated on using a particular pair of powered router devices. Thread packets go to it via a nearby powered device, but its replies seem to go via an iffy connection with low link quality via an entirely different device. The logs imply there are packets being lost when talking with this device. Amyway, that's fine, packet loss with weird routing is hardly surprising.

What is causing trouble is that it seems that the OTA process goes into a non-recoverable state if there's 4 retransmits in a row. At some point in an OTA update it'll abort. Anywhere from 5% to 90% progress so far. The Matter Server UI shows the device in "Waiting (Querying)" state and it never recovers.

I'm not sure what I'm looking at in the code. It seems that the OTA server gets shut down on the server side after the 4th retransmit in a row, but the device seems to be left in an "OTA in progress" state forever. It logs periodically that it's waiting for the device to complete even though the server had shut it down previously.

I must be misunderstanding. I'm not familiar with JS so I'm really not sure what I'm looking at. (C and assembler is more my speed). Anyway, is this plausible?

[deep drift]

Ikea devices don’t have serial numbers?

I just checked all of my Ikea devices - no serial numbers.

[humble citrus]

I am running into issues commissioning a new Aqara Door and Window P2. It comissions to Apple Home on the same Thread network fine. Looking at the Matter Server logs shows communication is fine but commisioning fails here:
Error post-commit of server-2-134b.peer10.commissioning.state: [validation-out-of-bounds] Error in reactor<server-2-134b.@1:28.commissioning.#peerAddressChanged>: (ValidationOutOfBoundsError/135) Invalid type suffix for MEI: 65535

[heavy dirge]

Manufacturer-specific clusters are identified by a Manufacturer Extension Identifier (MEI).

You should open an issue in GitHub to get this fixed.
https://github.com/matter-js/matterjs-server/issues/

[humble citrus]

https://github.com/matter-js/matterjs-server/issues/214

[heavy dirge]

Can you specify the firmware version of the device in the issue?

[left rose]

Is it this issue behaviour-wise?
https://github.com/matter-js/matterjs-server/issues/210

[silk plover]

Had not noticed previously, but neither do mine.

[deep drift]

Interesting approach to narrow down potential production issues without the possibility to identify serial or batch numbers. There are some ID codes on the packaging box, but who keeps those? The Matter codes are - to my knowledge - not unique, are they?

[silk plover]

Not sure on the codes. Working today, but while looking up the IKEA devices, noticed Nearly 30 of my 91 Matter devices are offline in the Matter-js server this morning. All of the devices are working and online in Apple home as far as I can tell from my workplace. Will have to try and check that out this evening.

[fringe fjord]

grab a debug log ... but when device states "delay on query then this is the device state ... it shoulkd recover 2h-1d later or such ... reasons are hard to predict because it is the devcie that controls the whole OTA process. The controller just tells "hey i have an update". rest is device side, so in such situations that extremely hard to say. Try to bring device more near to a better connection? But if you like grab debug log and I have a look

[fringe fjord]

might be. It is an optional field when it comnes to matter spec

[fringe fjord]

please post the full stack error trace., and ideally a full debug log of the process. There is an invalid ID coming from the devcie ... now need to find out where and what. ... ok I see the isseue ... I have a look what the device does there

[fringe fjord]

Yes but the MEI 0xffff is forbidden 🙂 ... sooooooo Lets see

[fringe fjord]

you know the answerr ... any log? even Info should give some picture for the start

[silk plover]

Pretty sure logs are at info, but will see if I can get them to you. Work at a hospital so finding time can be an issue during the short-staffed weekend workday. Should I start an issue or post them here?

[fringe fjord]

better issue .. here chance is too hight to leave from radar

[deep drift]

For being a standard Matter leaves a lot of options... 😉

[silk plover]

Numerous devices offline overnight. #217.

Hopefully logs uploaded correctly, first time I try to upload to GitHub from my iPad.

[humble citrus]

The log I added was the debug log (level set to debug) from the time I started the commissioning until I received an error in the iOS commissioning interface

[humble citrus]

I also added the firmware version that shows in Apple Home: 1.0.0.0

[grizzled burrow]

I think I saw a similar issue. I was updating 31 Inovelli White switches. All but about 5 updated with one click without issue. However, several got stuck in the above loop. When I looked at the Matter server page and tried an update from there, it appeared to be stuck in the Query phase but the OTA never started. To fix the issue I did a complete HAOS VM reboot, and then magically the remaining switches completed their OTA on the first try from the Matter server 'update' button.

[raw veldt]

There's definitely some fragility in the OTA system. I'm still trying to get a handle on what's going on but I did find a way to make things angry. I had a spare Matter/thread device that had a lot of test-dcl updates that had to be applied one after another - the Eve Motion. It's a perfect test device for trying to break things. I did find a (so far) 100% repeatable failure mode that I think is server-side. If you look at the test-net browser for vendor 0x130a (eve), product 0x59 (eve motion) you can see that, It can do 8+ updates in a row from the current prod firmware.

However, every single step breaks the OTA server. Suppose I updated the device from build 6711 to 6713. The file /data/ota/bin.130a.59.test is stored. When updating from 6713 to 6715, I think I see this happening:

1. OTA server starts and notes that /data/ota/bin.130a.59.test exists.
2. device OTA is activated
3. server receives the connection and opens bin.130a.59.test.
4. server discovered it's the wrong file. Logs a checksum mismatch and schedules a re-download
5. server shuts down.
6. client device is left in limbo
7. Any further interactions with the device cause the Waiting (Querying) state and logs that it's waiting for the client to get out of OTA state.

This is 100% repeatable so far. One thing I tried was deleting /data/ota/bin.* with the assumption that it would just download a fresh copy. Nope, invalid assumption. It starts the OTA process, discovers the missing file, and the server shuts down just like with the checksum mismatch.

The convoluted workaround I found was: delete the OTA files, restart the entire matter server. For each step. This worked reliably for 5 steps in a row.

Protip: make sure the batteries are fully charged. You get less retransmits.

[raw veldt]

Sort-of. The outcome is the same, but in the PR the reporter interrupted the OTA update on the client side. In my case I believe the server initiated the interruption after a few retransmits in a row. So "sort-of". The didn't-recover state is the same but mine was the result of a server-side footgun. I think.

BTW; I'm really liking the new server in general. I've even survived setting up device->device bindings with it.

[humble citrus]

It looks like there is a version 1.0.1.0 on TestNet. I can try installing that or I can hold off if you would need me to reproduce as is.

[humble citrus]

Oh, I actually see a version 1.0.2.0 on MainNet that I’m not getting offered

[silk plover]

Aqara made 1.0.2.0 for the door and window sensor public for a short while, then for reasons unknown to me it was no longer available. Not sure if they ever made it publicly available again.

[humble citrus]

Ah, in the details it’s listed as “software version valid: false”. Was still able to download the OTA file.

[fringe fjord]

Debug log for that please. But would be bad is the checksum is broken tbh because then there is not way to apply it at all. the query state swould be cool to see in log. So a debug log with a but time infos what happened when from user/your perspective would be cool.
But please also consider: test dcl updates are experimental and might not what you want and at own risk ....

[fringe fjord]

in theory a newer version on prod should win against test-net. also here a log from an "check update" bis dashboard would be cool

[fringe fjord]

yes a valid=false also blocks it

[silk plover]

We should be getting a MainNet inovelli switch firmware update soon that will test what most users will experience with OTA, as several of us beta testers also have 20 or more of those devices at home.

I’ve only tested OTA thru the private upload firmware option made available by Matter-JS server so far.

[grizzled burrow]

Yup. I manually downloaded the 1.0.2.0 from the main DCL, dropped in locally into HA Matter, and did the update. No issues.

[raw veldt]

Would I be right in assuming that this is an impatience (in my part) issue? 3 of 4 recently commissioned Matter-over-thread devices (Onvis S4) showed as having no neighbors but were definitely online and fully operational. I reset and recommissioned the 3 and that increased it to 2 of 4. We were out for a few hours and 2 became 1. Presumably I just be more patient and wait for it to converge/settle?

[silk plover]

Onvis S4 are missing the optional thread diagnostic network cluster so they can’t inform matter server of their routing/neighbors. If they show up, it will be with a dotted line because another node has included the S4 in its routing/neighbor table.

Most of my S4 remain disconnected from the network graph as well. It’s expected.

[deep drift]

My ZBT-1‘s connections (mostly with Eve Energys) are also shown as dotted lines. As the Eves have the diagnostics clusters, does this mean the ZBT-1 doesn’t have it?

Further, the ZBT-1 is shown as „Unknown“.

[raw veldt]

Aha! That makes sense.

Hmm. Trying to add an Aqara P2 errors out with:

2026-02-07 23:34:40.197 INFO Endpoint server-2-134b.peer39 ready endpoint#: 0 type: RootNode (0x16) behaviors: 💤parts ✓index ✓commissioning ✓network
2026-02-07 23:34:40.211 ERROR Transaction Error post-commit of server-2-134b.peer39.commissioning.state: [validation-out-of-bounds] Error in reactor<server-2-134b.@1:44.commissioning.#peerAddressChanged>: (ValidationOutOfBoundsError/135) Invalid type suffix for MEI: 65535
This happens both for direct commissioning via the app, and when commissioning first in apple home and trying to share to HA.

Apple home says the firmware is 1.0.0.0 - which is old according to the DCL. I'm going to try and nudge it to update and see if that helps with getting it onto HA. It used to work with HA on the old matter server.

Hah, nope, never mind. Apple home won't be nudged on such things.

[humble citrus]

I opened an issue on this yesterday

[humble citrus]

Also see: #new-matter-server message

[raw veldt]

Oh, bleah. "Software version valid: false" in the DCL for the updates. wow.....

Actually, no. This device is clearly cursed. Back in its box it goes. Again.

[fringe fjord]

Yes. These infos build up over time and also roles could change (like router or not). If you feel info is outdated selectivity node and click the reload button to force reload of the data

[fringe fjord]

The HA OTBR is no matter node so we have no data. Hence it us an unknown node seen by the devices. That’s the limit of just looking at the devices data.

[fringe fjord]

Was reported above already. I need to check. There should be be issue. Add you log there please.

[raw veldt]

I scrolled back and looked. It is identical. The firmware situation for these devices is nuts. It was an unreliable device when I tried to use it last year - I'm going to box it up until they get their act together. I can do without a flakey device making things worse.

[left rose]

I put that under "steps to reproduce", because the server side interruption isn't reproducible reliably... but yeah, I also think it is the server just stopping retransmissions after a while. After I placed my bulb somewhere else with a better uplink chain, I was able to update it. None the less, I shuldn't have to wait for two days or have to restart the matter server in order to be able to give it a new try.

[junior sedge]

Does the new matter server support batching matter commands? (for example, if i turn on a light setting both brightness and color temperature, will it combine the moveToColorTemperature and moveToLevelWithOnOff into a single invoke on matter 1.3 compatible devices?)

(it's also possible i'm misunderstanding what batch commands are for)

[raw veldt]

That is what I saw. I was looking at the logic around retransmit limits in MessageExchange.ts etc but wasn't paying enough attention at the time to capture decent logs. It appeared that the OTA session was sending with a retransmit limit of 4, which was pretty easy to hit at the time with transient events. The device responded to normal commands the whole time - aside from the transient packet loss.

IMO, given the nature of the OTA process taking so long to recover from and being client-side driven, perhaps the OTA sender should be more persistent? A retransmit limit of 4 seems reasonable for something that takes a few seconds to recover from but not so much for something that takes a day or more. I think it would be more appropriate to make sure it's really, really, definitely, absolutely dead before dropping the OTA server session.

The cache of the OTA file is definitely incomplete though. It doesn't consider the version. It stores a file called bin.{vendor}.{product}-{test|prod}. If I've just updated to version 6711 then that's what is cached. If I try to update from 6711 to 6713 then the cache doesn't get cleared and the OTA server starts with the wrong data in place. It does catch it eventually and won't serve the v6711 data to the client looking for v6713, but it kills the server in the process and initiates the day+ timeout.

[hollow cloak]

@modern olive I converted your message into a file since it's above 15 lines :+1:

[modern olive]

Also I have a device that received a different IP address after a power cycle. The Matter Server doesn't seem to be picking it up. What should I look for? It has popped up on mDNS with the new IP.

The Matter server logs took quite a bit to show anything for that device and it's trying to connect to the wrong IP.

[reef crane]

Since my update from matter.js 0.3.4 to 0.3.5 some Thread devices have gone on and offline. I have 9 offline now, out of 45. Not sure if it has anything to do with matter.js, but they were stable under 0.3.3.

[modern olive]

What could cause this device to suddenly gave most entities unavailable, and new ones created on the side? The new/current ones are called the same as the previous ones, with a _2 suffix. Interestingly, the switch entity hasn't changed.

[fringe fjord]

Ok, the reason is, as also added to the github issue ... Aqara is announcing itself with "Device Type ID" 65535 which is no valid Device Type id ... I will add a workaround to ignore that and also informed aqara

[fringe fjord]

The answer today, same as the python server, is no. Answer in a week (or whenever I manage to finish https://github.com/matter-js/matter.js/pull/3065) is "it depends" 🙂
Depends because:

• not all devices support multiple paths in one invoke. Hue Bridge supports it but I doubt that that many use this. If "normal" devices support it depends on the Matter version AND the settings. If you like verify in Basic Information cluster on the Root endpoint. If you have a maxPathsPerInvoke existing and has a value >1 then yes.
• The Idea of the linked PR is to just collect the commands that happen "inside one Macrotask", means the commands would need to come in really "in parallel" Websocket wise. We also could define a "collection time", but then we come again in a question "how long is too long" where likely asking 10 people leads to 12 answers 🙂 ... soo how many people we have here? 😉

Lets maybe do it that way ... I post some ideas and you give "thumbs up" or "down" 🙂

• 0ms (aka Instant aka "Same macrotask")

• 10ms

• 25ms

• 50ms

• 100ms

[fringe fjord]

Ok, the 4 retries is the Matter standard for transmissions and also incorporates the device Session definitoons and such. by default BDX messages have a longer "waiting timeout" of 5mins instead the typincal some seconds. So I need to analyze it in depth ... but the question is if the devce really answered to it or not . I also saw devices just stopping stufff and then because of the hight BDX timeouts the process is blocked 5-10 mins at minimum ... this is currently the way it is.

For your comment on OTA storage ... tbh ... which reason you would have to use different versions on your devices? For now it is intended to npt use the version. The newest version wins, thats the idea. Else we need way for you to select the version whcih is not the idea currently. So why you want older versions? YOu can not downgrade anyway because also devices will not accept that. We alwaydownload the latest version relevant and this should be nothing about "wrong verison in place". Yes this approach can lead to "you need several downloads" when you need to do a "version a->b->c->d marathon but that should be not that often.
ALso why should we clear the latest version? The idea is that if you have a device multiple times then you usually want to update them all - so in this case the update is already there and does not need to be re-downloaded.
So i would love to discuss the real use cases where exactly issues arise and not theoretical things that could happen. If thats in the issues on GitHub, great then I will come to it ... else ... open to discuss but all these things are intended to be that way for now because I decided it to be that way because of reasons ... I am totally fine to adjust if this is not working in reality, but for now I did not saw anything why i should change 🙂

[fringe fjord]

basically this seems to be a bit strange device rechability issues where devices send very old cks for stuff already solved. Try updating the matterjs server ot the latest verison. If they still happen please use the Dashboard to temporariely enable debug log and grab a debug log and open issue and I take a look.

[fringe fjord]

Without a log hard to say. Try a restart of the server. If the devices keep offline try to re-power them and see if this solves it. A debig log would help to maybe see whats up

[fringe fjord]

if you still have the log (even if info) then we could look into, else no clue tbh, also as discussed on GitHub

[junior sedge]

0ms would fit what I'm thinking of, which is the fact that a single light.turn_on call can get expanded into multiple matter commands to set both color and level, and also into multiple matter commands to different lights (that only matters for bridged devices).

[fringe fjord]

The question is how fast and how exactly the Python HA layer sends these things over Websocket to the matetr server and how exactly the timing here is. we likely need to try that out. So yes also i would strat with 0s tbh and then see if/how it works ... guibven the fact having a device that suppportes it

[junior sedge]

i suppose finding a device that supports it is a challenge. I'll take a look at what i have, but i guess i shouldn't get my hopes up. I wonder if it might make sense to add something to the websocket protocol to allow the integration to explicitly frame a series of commands as a group.

[modern olive]

This happened again about an hour ago (I think after a restart of the Matter Server). I'm trying to grab logs but even the maximum lines the GUI allows me to download only covers about 5 min.

[junior sedge]

Oh, hey, Aqara T2 bulb has MaxPathsPerInvoke=5

[deep drift]

Any news?

[fringe fjord]

I ask

[modern olive]

I've attached logs and a screenshot to the GitHub MR.

[fringe fjord]

Matter Community Sync starts now ... come and Join ... https://meet.google.com/uch-rsxw-txh

[flint compass]

I'm noticing this exact same thing from 0.33 to 0.35

[fringe fjord]

In 0.33 we sometimes stored wrong former IPs when they changed, which leads to the fact that a restart always had wrong IPs and soe was more relying on MDNS ... this should be fixed since 0.3.5, but basically that means you need to get the devices back online and announcing via MDNS once. As said several times: if you still have devices offline and also re-power them does not solve it (and ideally you are sure they are discoverable via MDNS) then create an issue an d add debug log pelase

[tranquil agate]

Last two nights I found several restarts of the Matter Server (0.35 i assume) during the nights. I cannot really see the logic begind this, because the server and all matter devices are stationary at that time. It seems to be happening close to the time my daily automatic backup is created. Matter Server Log is at info level. I also include the Supervisor Log, to show the coincidence with the backups. Is there something I can do to shine some more light on this?

[fringe fjord]

how is your RAM looking? I do not know how exacly the backup works but if that needs more RAM then maybe you come over a lomit and the OS kills the container?

[tranquil agate]

Is there a way to view the RAM history. I do have 8GB installed, and right now still 5,6GB available. But I don't know how to tell how much was in use at the time of the back-up

[boreal coyote]

https://www.home-assistant.io/integrations/systemmonitor

[fringe fjord]

I saw "service exited with code 134 (by signal 6)." in your logs which means SIGABRT ... so the questin is whay ... so usually I would look into the /var/log/syslog on a normal linux system, but no clue how/where to gain such details in HAOS 🙂

[tranquil agate]

I just installed System Monitor, so I can tell you tomorrow what it looked like, this night.

I'll invoke a backup manualy, to see if I can reproduce the behaviour...

[fringe fjord]

Fir kill was 05:50:31.245, then in the middle of startup it was shot down again 7.2. 05:55:39.277 and 05:57:39.125 and 07:29:14.254 and 05:57:39.125 and 07:29:14.254 and then 02-09 05:00:34.813 and 05:12:52.781 .... kw I stop ... so yeah something is strange and something is killing the container

[tranquil agate]

Hmm, there's also something in the Home Assistant Core logs, I wasn't aware of:

[hollow cloak]

@tranquil agate I converted your message into a file since it's above 15 lines :+1:

[tranquil agate]

Logger: homeassistant.components.matter
Bron: components/hassio/addon_manager.py:408
integratie: Matter (documentatie, problemen)
Eerst voorgekomen: 9 februari 2026 om 05:18:16 (1 gebeurtenis)
Laatst gelogd: 9 februari 2026 om 05:18:16

Failed to start the Matter Server app: Another job is running for job group addon_core_matter_server

A "manual automatic" back-up does not lead to significant memory use

And no restart of the Matter Server

[subtle flint]

As someone who is not running the new matter server (I have a WAF budget and I need to rebuild it before I can chance running beta Matter after dealing with a complete "delete everything and start over" event a couple of weeks ago) is there any point to updating to any OTBR version that only mentions the beta in the release notes? This is for the current 2.16.2 but has been true for a couple/few lately.

[tepid spindle]

Hey everyone!
Please help me. Where should I dig?
After restarting my OS matter server freezing on startup on INFO [matter_server.server.vendor_info] Fetching the latest vendor info from DCL. Nothing happens after and matter server doesnt work at all untill I restart it

[flint compass]

In my case all devices come online but then several of them go offline and back on much more frequently than in 0.33. They do tend to be the ones more on the edge of the mesh. With 0.33 I rarely had a device go offline (except my every-6-hour storms), but with 0.35 devices blink off and on more frequently.

[silk plover]

Better place to ask this question: https://discord.com/channels/330944238910963714/1458060881970856051

[junior sedge]

the most recent OTBR app update only affects people having the OTBR beta enabled; if you're not using beta there is no reason to update.

[compact root]

That's in the development category and shouldn't be used for regular user questions.

@subtle flint Your regular question would normally be asked in #1284966582375813201, but kepstin is correct with their answer. I often ignore those updates myself.

-# A general note: We prefer the community use the term Home Approval Factor (as seen [in this newsletter](https://newsletter.openhomefoundation.org/open-home-approval-factor/) ) - even if it's only you and your wife, she's still part of the Home and so are you, so your whole home needs to feel happy about the changes.

[fringe fjord]

Why you would need to delete everything and start over? The new server automatically migrates everything and on issues you can switch back to the python server. So what do you mean?
For the OTBR beta please go to https://discord.com/channels/330944238910963714/1467964294837833859

[fringe fjord]

That logline feels like belonging to the pyythin matter server? Then please ask in the Matter channel. Likely check internet access. If you use the new Matter Server Beta then I would wonder isf that would happen and woul dlike to see an issue and a debug log

[fringe fjord]

Hm, nothing was changed in that regard ... maybe more traffic in your thread network and better to start improvig with some more routers 😉 The matter.server is just the receiving and sending part ... hard to blame the server if devices have bad thread connection ? 😉

[subtle flint]

That delete-and-restart was unrelated to anything new-matter/OTBR. It was a last-ditch attempt to fix a different issue. I only brought it up to mention where I'd spent my HAF budget (and as @compact root rightly pointed out, as both a part of the household and the person who was re-binding everything, I was probably more affected / less pleased than my beloved...)

[fringe fjord]

hey guys, here is the link to the transcript and recording of the latest Matter Community Sync from monday. https://docs.google.com/document/d/133vRNDa6QqJm-3yTJf0m8DBzwhjCyKA-OFCj_rcBUr8/edit?usp=sharing
I currently prepare 0.3.6 of the Matter server ... so stay tuned ....

[fringe fjord]

Hey everyone, the new version of the Matter Server Beta 0.3.6 was just relöeased and contains several fixes, optimizations and adjustments. Changelog: https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#036-2026-02-11

[deep drift]

The upgrade was again flawless. However, the Eve Energy child lock switch is not available.

[fringe fjord]

Yes. because this change was only about the Matterserver side ... noone said that this will become avaioable as entity in HA with this update 🙂

[silk plover]

New icons look great . Only small thing I see is that after restart I lost the manually input 'Node Label ' of one device an Ikea bilresa dual button. all other 92 Node Labels came over correctly, including second Bilresa button. Pulled the logs, will submit issue in a few.

[fringe fjord]

strange ... the NodeLabel ist stored on the devcie ... what does the Debivice show on Root endpoint > Basic Information?

if missing there maybe an other fabric (is it paired anywhere else?) overwritten it?

[silk plover]

Was missing at EP0/Basicinfo/nodelabel. The words ‘Node Label’ were in the text input box after restart. Also paired to Apple Home.

[fringe fjord]

@flint compass I also hear that 6 hour issues with Apple networks again from someone else... maybe a good Idea to open a apple ticket with all infos? https://github.com/project-chip/connectedhomeip/blob/master/docs/guides/darwin.md#providing-feedback-to-apple

[fringe fjord]

Then set it again and see what happens next time ...

[silk plover]

I did exactly that earlier today. Thanks.

[silk plover]

Suspect it is something specific to the Ikea Bilresa and/or in combination with apple home with the Node label. Just updated to tvos/ios 26.3 final and everything came back online fine, but now both bilresa dual buttons I own have lost the manually input node Label names. Other Ikea water leak, door/window sensors all came back up fine with the previously input node label names.

[fringe fjord]

Please try to Repro. I can try to report to IKEA if behavior is consistent

[silk plover]

Heading out to work, but will try to do some restarts later after family goes to sleep.

[reef crane]

Upgrade to 0.3.6 had a couple problems for me:

2 Eve Dimmer Switch lost NodeLabel
1 Eve Light Switch lost NodeLabel
NodeLabel for all 3 switches worked correctly before. All 3 are multi admin to Apple Home

HA menu item Settings/DevicesAndServices/:
Devices and Entity displays are lost (they initially flash on, then disappears). The failure only occurs on web portal. iPadOS HA app shows both displays correctly. Under same menu item, Integrations and Helpers display correctly.

[silk plover]

Interesting. Did you also upgrade or restart your Apple home network?

[reef crane]

I upgraded my appleTV that was the Apple home hub. Other Apple border routers have not been upgraded.

I think I upgraded the Apple home hub before moving to 0.3.6, but not sure.

After more thinking, I believe I upgraded to 0.3.6 before tvOS 26.3

[grizzled burrow]

0.3.6 and tvOS 26.3 here. Upgraded to tvOS 26.3 RC last week, GA version today. ALL US Eve energy in-wall outlets lost their labels, along with both Ikea buttons.

[silk plover]

Had created an issue earlier for the Bilresa Node Label deletion, added some additional context from your replies here. In case either of you would like to follow along. https://github.com/matter-js/matterjs-server/issues/243

[reef crane]

I had to reduce my Eve Energy Outlets (in-wall) to one because the aggressive polling to keep up with the energy messages was bringing down my HA OTBR fabric.

They are matter 1.1 devices, so polling is required. Lots of failed communications.

Eve energy wall plug is matter 1.3, so
polling is not required. But even with it, the very high number of energy messages is a serious issue for my thread fabric.

They do not have to be polled, so it is better than the in-wall version, but still a problem. I have 5 on my fabric now, looking to move some to an Apple Home fabric, which does does not try to keep up with them so it works better.

I am minimizing eve energy devices on my network. The reporting toll is a big problem for a thread network.

[grizzled burrow]

Yup I reached out to Miranda at OHF to complain about the terrible Eve energy firmware.

[reef crane]

It’s probably more accurate to say it was bringing down my matter fabric. I think the problem is bigger than thread. I doubt I could multi admin them from Apple home to HA without causing the problem to happen again.

I use Apple home for voice communication and button pushes. HA does all the automations.

All HA scenes a user needs are duplicated on Apple home and HA.

If matter.js could do something to reduce the impact of Eve energy excessive reporting, that would be great.

My impression is that Apple handles them better. Likely because of the close Eve Apple relationship.

[raw veldt]

Hmm. I've got some Bogons on my node list now that don't seem to exist. They weren't there yesterday.

[raw veldt]

I had a theory what might have caused it, but it doesn't quite fit. A couple of odd things happened over the last day or two.

1. I was goofing around with an Aqara P2. I'm pretty sure I had multiple failed attempts at pairing it, but I do not recall the details.
2. I deleted/re-commissioned some Onvis S4s. I thought those would have fit the gap in the node numbers but it's not quite right... so that's not a viable theory.

[deep drift]

I can relate and confirm your observation. I have seven Eve Energy plugs with Matter 1.3.0 in my mesh. Looking at my logs it is obvious that their „energy messages“ are quite frequent. They cause no issue (yet). Maybe @fringe fjord could comment if this chatty behavior is normal or should be attended to. I am happy to provide logs.

[raw veldt]

I'm wondering about that Eve Energy reporting thing. In the ZIgbee universe there are controls for reporting rates. eg: I have battery temperature sensors all over the place. I can set minimum/maximum reporting interval, what the temperature delta needs to be etc. Does something similar exist in Matter?

[deep drift]

Good question. I have no idea.

It would be interesting to understand how much bandwidth the energy messages are using. To see them in the logs is one thing their real impact is another.

[reef crane]

I can also provide logs.

[raw veldt]

For a zigbee battery device with reporting interval controls. I didn't see anything even remotely like these when browsing around an eve energy, nor in the python matter server source either. (It has custom clusters defined for things like Eve but none related to this)

Hah, this came up on https://www.reddit.com/r/MatterProtocol/comments/1mfn84m/reporting_rate_in_matter_smart_plugs_with_power/ not too long ago.

Eve Energy reports every 60 seconds 
Meross MSS315 reports every 10 seconds
Tapo P110M reports every 10 seconds

[reef crane]

I think Eve energy reports a lot of attributes. I have not analyzed in detail. The big problem was Eve energy outlets. Just 2 on my network caused it to become totally unstable. My network does ok with 1 eve outlet and 5 plugs. But I am going to drop to 3 plugs.

[deep drift]

This post is specific about Eve with an Eve guy commenting: https://www.reddit.com/r/EveHome/s/bib5Dp4uL6

[deep drift]

The congestion is probably depending on the mesh quality. As I stated, I run 7 Eve plugs without any issues and the responsiveness of my devices is really good.
And my mesh quality is not perfect.

[reef crane]

Also, I found the Eve Energy Plug has a much higher rate of reporting if it is plugged into a varying load. With no load, it is not bad. I plugged it into an entertainment system with ethernet switch, MOCA adapter, Xfinity Modem Extender, TV, Sound system, AppleTV, and the problems increased dramatically. Had to remove entertainment system from Eve Energy Plug, left Eve Energy plugged in with no load, because I wanted it as a REED.

[deep drift]

That’s a good point. I would estimate in average only 1/3 of my plugs are with constant load. Maybe even less.

[raw veldt]

I'm in the process of moving my leviton gen2 homekit/matter devices over the inovelli thread switches. I like that the inovelli stuff can actually do bindings. and it actually works. (!)

I'm trying to get solid coverage without relying on things like having eve energys plugged into odd locations.

[reef crane]

I don't have any device, except the Inovelli that does binding. Looking forward to some binding targets to see how it works. Matter.js has made using binding possible.

[raw veldt]

The python matter server had binding support (ish). You could enable it by setting the beta flag. Except that now does something entirely different (heh).

[deep drift]

I am using Matter bindings between Eve thermostates and Ikea temp sensors and they work great.

[raw veldt]

The HACS "Matter Bindings Helper" has been a problem for me with the new server. It always seems to fail/blow up on ACL configuration.

My particular use case for bindings was to use the inovelli ceiling fan canopy remote control thing. It works although not as correctly as I'd like. It has stopped the spouse complaints about the "stupid fan!!" etc.

Matter has some really strange design choices. eg: there's an on/off control. And an entirely separate fan speed control system that looks entirely different to what a dimmable light control is. In other words, you can bind an on/off light switch to the fan controller but it'll be 0 or 100% only. I kind of expected that dim up/down would have mapped into faster/slower, but nope.

I started skimming through the Matter 1.4 core spec looking for the subscribe/reporting interval stuff. It's "only" about 1200 pages. But then I got an instant headache. And realized that if I really wanted to hurt my brain then it's a lot more fun to do it with Factorio than reading specs.

[fringe fjord]

As said again "interesting and no clue why NodeLabels shluld get lost ... unless someone else writes them ... We do not write them in any case on a connection or such. strange

[fringe fjord]

We only poll whennneeded, but sure if stuff changes then eve devices report changes to energy measurements sometimes very often ... bit thats nothing where can really do anything on 🙁

[fringe fjord]

Maybe we are also again in the topic: "Do not use thread devices in nmultiple fabrics!"... because you multiply the data reports. So in case of eve energies reporting energy values every whatever seconds ... if you have the device in Appple and HA then the data report is sent twice over the thread network ... now multiply that woth number of devices and such. So the "best practice" for thread devices, especially in such cases, is clear: try to limit to ONE fabric/ecosystem and prevent multiple!

[fringe fjord]

puhh.... sorry but how they could? That impression is wrong, sorry to state that.
The matter.js server polls the same "every 30s" as the python server was doing ... I can reduce that to 60s or 2min or 3min ?

[fringe fjord]

There are discussions goinfg on. effectively they update the data in ways that are compliant to the spec ... if thats "too mucH" is on another piece of paper and mainly the part of the discussion ... but no clue if there are changes. But yes rest - see my latest posts

[fringe fjord]

Yes, but I can not say "give me all changes in realtime beside energy values, these I want to have more seldom" ... so we can say "report data only every 30s" but then we would also get "turn on/off" changes just that late ... i guess this is for 90% of the users not acceptable.
So the easiest way to optimize - reduce fabrics"

[fringe fjord]

hionestly ... logs do not help here because too many other factors play a role and also about your thread network and such
But I get the feeling we have some discussions here that do not really belong to "it is an issue of the new server", right? We do not act differently than the old in all this regard

[fringe fjord]

Shure it did not worked in the new server? I would more say that the dashboard in the old server only checked the very first status returned and if that was ok then all was ok ... reality was that the first is basically always ok, so the dashboard logic of the poython server simply never showed any error if if not all ACLs were written as expected. Same for bindings. So you ended up thinking all was working but it actually wasn't. This is fixed in the dashboard orf the matter,js server now and it correctly shows the status of the ACL change - and yes this leads to showing errors when not all data could be written ...

[deep drift]

Keeping the number of fabrics low is for sure the way to go. Might still be a challenge as many vendors are pushing their Matter controllers/TBRs into our home networks.

[deep drift]

For example Eve’s iOS app is (still) not able to communicate directly with other Matter controllers than Apple’s. So, to change the „Transmit Power“ setting of Eve‘s battery powered devices, you need Apple Home. I am not suggesting, that Matter.js should automatically support all vendor-specifics, but all vendor specific apps should be able to access their own Matter devices through whatever Matter controller is present in the fabric. Of course, this is nothing we can solve here with Matter.js. This is a strategic topic of the CSA and their Matter specs that would also need to include vendor specific apps.

[fringe fjord]

We already have custom clusters and details ... so if someone knows details make a PR and add it

[deep drift]

I know and I’m grateful they are available via Matter.js. But this wasn’t my point. Anyways.

[fringe fjord]

ahh I see ... yeah but thats hard ... usually the apps use the OS Matter frameworks and they only work if the devcie is in the fabric there

[fringe fjord]

Hey everyone, 0.3.7 available (https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#037-2026-02-12) ... just some optimizations on startup and websocket optimizations.

It would still be interesting if you have more that 50 devices on matter to get a log from such a start with this version (info loglevel is enough). I want to look into some devcie startup behavior things and so would be cool to see. But really mainly relevant with many devices

[silk plover]

@fringe fjord DM'ed log of restart and devices coming back online after 0.3.7 install.

Ikea released a new Bilresa dual button firmware 1.9.11 to Testnet DCL, will try it out on one of my 2.

[deep drift]

Please check if it now correctly reports the battery level in %.

[silk plover]

install quits/fails on every attempt so far. Giving the thread network and Dual button a few minutes and will retry after pulling batteries.

[solid flare]

I have a 53 device network (6 Wi-Fi, the rest Thread). How can I get you logs? I can update shortly.

[fringe fjord]

just send me in discord as PN ... i guess most easy

[solid flare]

sent!

[fringe fjord]

lol ... you were faster than me posting it 🙂
But be aware ... yes it passed internal tests at IKEA and also cert wise is ok, but still it is a beta and there is no way to restore to the former version! Do on your own risk!

[silk plover]

still failing to install, gets to 90% on the update then just stalls out. EP0|Cluster 42|UpdateState changes to 3 and gets stuck there seemingly until I restart Matter server.

[fringe fjord]

yeahh for the ota state when an update stalls or devcue restarts or such there is still one ossue I am currently looking into in the beta server

[silk plover]

I give up for now. Seems somewhat similar to eve weather Matter OTA update issues for anyone that dealt with those previously where the device itself is failing to complete the install (matter server/thread gets blamed), but now added on Matter-JS server is getting stuck in a blocking state to reattempt the update again later.

[grizzled burrow]

IKEA button update stuck at 2%.

[deep drift]

Somehow, it gives me a flashback to the last firmware update of Eve Weather to 3.5.0. I must have tried at least a hundred times.

[fringe fjord]

Lets see if I can finish the OTA fixes today, latest tomorrow

[rigid bone]

I'll just add that my IKEA Bilresa gets part way loading the new update and then things stall. The Device is reporting update state =3, update progress = NULL.

[silk plover]

had the Bilresa node details page open in the back ground on another monitor while working on something else. As stated before noted the bliresa EP0|Cluster 42|UpdateState changes to 3 when update/install manually and then it does revert to the default 1 at some point after the Node fails or stops responding/reporting OTA info and all seems appropriate. Then 5-10 mintutes later the bliresa EP0|Cluster 42|UpdateState changes to 3 again and the update button as well as the HA firmware update notice go to an endless 'Querying/waiting" state.

[rigid bone]

I should also add, that the logs also reveal a BDX retransmission limit reached time out

[fringe fjord]

Maybe wait for the next update of the server. But thread updates for batter stuff is sometimes always a bit tricky

[rigid jackal]

Is anyone else having issues with inovelli dimmers on 0.3.7? I'm not home rn so cant get logs, but after moving to 0.3.6 and 0.3.7, I am having issues keeping my inovelli dimmers connected

[reef crane]

Upgrade to 0.3.6 (and 0.3.7) had a problem for me:

HA menu item Settings/DevicesAndServices/:
Device and Entity displays are lost (they initially flash on, then disappear, leaving message: "no data"). The failure only occurs on web portal. iPadOS and IOS HA app show both displays correctly. Under same menu item, Integrations and Helpers display correctly.

[reef crane]

Here is my INFO log after turning on 0.3.7

[left rose]

Please create an issue here: https://github.com/matter-js/matterjs-server/issues/

[hollow raven]

It’s possible to update the bilresa but when the update stall you need to restart the matter server. I had to do it two times then the update went though.

[fringe fjord]

What does "web portal "mean? What means "flash on" means? if you mean the dashboard of the server, did you tried to reload?

[fringe fjord]

see above, working on fixes ... new version next days

on first view some nodes are strange and do not connect or mix ack's but thats all "potentially common stuff". but in fact "start_listen " is returning all 9x nodes ... need way more details to maybe tell you whats up ... but I can only check the Matter server side ...

@reef crane Your log is very interesting and you should check your wifi. The thread nodes connect as expected and thatc ool ... but your wifi nodes are struggle a lot to connect and have maany subscription timeouts and reconnects and such ... Maybe check that the thread and wifi use different channels and how good the wifi nodes are connected and such ,... there is definitively something strange - but pure behavior wise from the server I can not see anything which is not as it should

[reef crane]

I mean the browser interface versus the HA IOS/iPadOS app

I did reload, I upgraded to 0.3.7, I tried on multiple computers (the browser is Safari). By flash on, I mean the display appears for a fracton of a secomnd, then is replaced by a screen that is all black except the header and the "no data" display in the area the device list should bel.

Interesting about my wifi. It is xfinity xb8 modem with several xfinity extenders (all extenders are connected with ethernet cable). It is a rigorous wifi infrastructure. There is full bar wifi everywhere. But we have a growing neighborhood and sometimes have slow internet. But i don't think that was the case when i took the log. One issue, is the wifi is both 2.4Ghz and 5Ghz. Xfinity does not allow me to create a separate 2.4Ghz wifi.

I just tried the Devices and Entity display on my only MacOS 26.2 version (others are 26.3), and it displays correctly. Upgrading that one to 26.3 to see what happens.

I will check the 2.4Ghz channels. I do use a lot. I have 1 Zigbee, 5 WiFi extenders, 1 cable modem, 1 Apple Home thread fabric and 1 HA thread fabric (independent fabrics, each with a different Pan ID). That is 9 channels used..

[rigid bone]

Success updating a Bilresa using 0.3.7 🙂

[reef crane]

I checked my WiFi. All the subscription churn is THIRDREALITY Smart Color Night Light, which is a mains powered motion sensor with nightlight. All other WiFi nodes are maintaining their subscripton.

[grizzled burrow]

On 0.3.7, did a full HAOS reboot, and Bilresa downloading gets stuck at 0%. Many hours later, the updates are still stuck.

[reef crane]

I updated the firmware on all my ThirdReality motion sensors and the subscription churn has stopped. Thanks for the Heads Up.

Turns out the xFinity extenders all use the same channel as the modem, so my actual 2.4Ghz channel count is about 4.

[deep drift]

Is BILRESA’s battery level in % now correctly reported?

[fringe fjord]

pan ids are irrelevant .. channels are relevant ... especially also where "much used wifi" channels might interfere with thread channels

[left rose]

Exactly. Also, placing the button near a BR might also help in subsequent tries.

[left rose]

Ultimately see here: https://support.metageek.com/hc/en-us/articles/203845040-ZigBee-and-Wi-Fi-Coexistence
In that context you can read ZigBee and Thread interchangeably wrt channel consumption because they both use IEEE 802.15.4.
Turn off auto-channel wherever you can (esp. WiFi). Try to join your Thread PANs, because then all BRs create a single mesh using only one channel. This is different/independent from Matter fabrics, you still can have multiple fabrics on a single Thread mesh.

disclaimer: not every mix of BR brands/makes does actually play well, though, which would be considerable when joining Thread networks.

[tranquil agate]

Well, just to close the loop on this one. No more restarts for the last three nights, so probably it was just a glitch in my system. Thanks anyway for the help and all the effort on the Matter Server

[rigid bone]

I tried updating a second Bilresa today (using 0.37), and it failed many many times before finally succeeding. A few times were BDX timeouts, even when moved close to the OTBR. But others were seemingly weird, like receiving a pairing request from the Bilresa which ended the session. Another was receiving a message from unexpected protocol from Bilresa. Anyway once it got going well past 15 to 20% it made it the rest of the way (but I did see the downloading pause while it waited to do an mDNS query)

[heavy dirge]

Matter Fan oscillate feature doesn't work with matter.js-based Matter server #162930

[raw veldt]

Waitasecond. Aqara P2 door & window sensor appears on my device list. That was broken during pairing before. Is its protocol violation "fixed" now?

[raw veldt]

Hah, nope, its somehow still wonky. After a server restart the ghosts are back:

How I got that: it appeared at node 69. I removed, reset it. Recommissioned. It came up at 73. Restart server. Now 73 is a ghost and the P2 is on 74 (!).

[raw veldt]

Eww. With an OTA update tranfer in process (about 60% complete) the DCL checker tried to re-download the file.

2026-02-13 19:28:02.210 INFO DclOtaUpdateService Update available in Test-DCL new: 1010 vid: 4447 pid: 8194 prod: false current: 1000
2026-02-13 19:28:02.232 INFO DclOtaUpdateService Existing OTA image validation failed, Re-downloading ... [ota-image] OTA full file checksum mismatch: expected "Zm7Hua+iaiRl53NC01Z2NeV3h1sfKGpCKALU1+aAR2w=", got "rFZ6WdH0DuuCf7HVoRmNjCF73mYZ98DGYpHoDKmf0Bw="
at Function.file (/opt/matterjs/node_modules/@matter/protocol/src/ota/OtaImageReader.ts:77:19)
It failed to download because Aquara used an internal URL (sigh). It tried to replace a local prod file with a test-dcl file because two different software versions had the same name.

[raw veldt]

The device is quite thoroughly bricked at this point. I don't know whether the re-download attempt while halfway through the flash was the culprit or whether it was Somebody(TM) tempting fate too much.

[humble citrus]

That was fixed in 0.3.6

[fringe fjord]

see above, fixes in progress, wait till 0.3.8 of the server

[fringe fjord]

please open issue with a full log yu have. usually if thisis logged the file got not stored, but will check. interesting case

[rigid bone]

Both of my Bilresa's are now reporting around 60% (using the LADDA rechargeables).

[fringe fjord]

Ps: normally the device validated the downloaded file after download again before applying. So in theory should not be an issue. But yes log and issue please

[left rose]

The Aqara P2 contact sensors are known to get bricked by regular Matter OTA updates, at least if you update from <1.0.2.0, and for sure when you come from 1.0.0.0. This is also why it says "Software Version Valid: false" in the production DCL. The firmware will work if you manage to flash it, but how you get there (outside the Aqara ecosystem) seems to be a mystery.

[raw veldt]

That than explains the situation. It comes under "Somebody(TM) tempting fate." - I copied the ota file to /addon_configs/core_matter_server/updates (which it assimilated). Presumably the only way to have done it without bricking it would have been to let their hub or cloud do it?

[fringe fjord]

I think I will make at least the dialog in the matter server for updates sources from
Test DCL or local way more „warning-ish“

[raw veldt]

Even if it had issued a scary warning along the lines of "The local file is marked as invalid by the vendor in the DCL" it wouldn't have saved this device. A certain somebody would have tried it anyway.

[fringe fjord]

I could only warn for „this is a local update. Know what you do“. When you add it yourself I can not check DCL because origin is unknown.

And yeah again somebodies(TM) decisions I can not do anything anyway

[raw veldt]

What does concern me though is that the new server can only seems to handle one version for a vendor:product:[main|test]. Take the eve motion sensor, they have a test-dcl chain of about 20-30 sequential updates that must be applied in order. The P2 had two in-order updates on prod (although marked invalid). Based oin what I saw earlier the OTA process choked on each step.

It appeared that the OTA process was:

• I have a cached OTA file for this vendor:product
• start OTA server, tell device to initiate OTA
• OTA server opens file and discovers its the wrong checksum, marks it for re-download.
• client left hanging for its OTA timeout
  I could recover by powering off the client, deleting the cache in /data/ota/*, restarting the server, waiting for the server put the client into MIA mode and invalidating its sessions, then power up the client. Then reinitate OTA which will work. For one step only. Repeat.
  I know you want logs for this but my test-case device doesn't have a path back to prod firmware yet and I didn't save the logs at the time.

[raw veldt]

It most definitely did already give a warning to that effect. I vaguely recall that it was the same warning text as a test-dcl update.

[fringe fjord]

That’s sounds all wrong tbh. So yes when he finds a local update which is not applicable but a better one then it is downloaded. End. Invalid checksum means invalid file. End. It does validate the file already initially so finding invalid things just in the update is strange.
Most updates I know are not sequential but have higher version ranges. So again why I should save multiple version files? I still miss the usecase and real world examples.

So if you think there are issues because of that collect it. Repro it and create issue with details and a debug log. Thanks.

I know with it’s hard and pot Saldo Info log is ok. But I need examples else I am blind

[raw veldt]

The scenario was Eve Motion. They have:
v6703, min 1, max 9999
v6705, min 0, max 6703
v6707, min 6705, max 6706
v6709, min 6707, max 6708
.... repeat for another 10 versions.
The OTA update process fails for every single step.

I've just noticed that they have a downgrade/reset if I can get the OTA process to put v6703 on it.

But again, the error flow for every single step was:
6705 installs, Applying, offline, online.
Click check for updates. Offers 6707. Click confirm.
According to the logs, the process is started, the client-side OTA is initiated, the server starts and notices the checksum mismatch (because it's got 6705 in the cache but this is an OTA for 6707). The server marks it as invalid and aborts. The client is left hanging with an OTA trying to start and the server has already table flipped.

Anyway, I'm going to try forcing it to downgrade to 6703 since that appears to be allowed. Then capture debug logs of the OTA server failing at each sequential update.

Logistics question. Firmware is v6721. v6703 is within the min/max rules (1..9999). Before Somebody(tm) breaks something again, what are my best chances of forcing this? Best guess: turn off test-dcl for this device, provide v6703 locally, will the server allow this with a version number downgrade?

[fringe fjord]

I have an eve motion too. I will see what’s the firmware. Please create issue with above details.

[fringe fjord]

There is no downgrade. 6703 should never be offered as update when current version is higher. So easy that is. Else we always check the highest available valid version

[raw veldt]

Obviously it shouldn't go backwards. Out of personal curiosity, what prevents it? Is it the server's responsibility or will the device refuse it if its an older version?

[fringe fjord]

Server ideally never offers these. But device also verifies

[raw veldt]

I'm curious whether the device does a simple version(new) > version(current), or whether it looks at the header to see of version(current) is in version(min..max). Or maybe my assumption about min..max is wrong and that its just a DCL attribute (rather than an ota header) to allow vendors some flexibility in what is offered to devices in the field.

This is how Somebody(tm) gets into trouble.

[mellow escarp]

I have RODRET and STYRBAR remotes from Ikea could you help me troubleshoot them in home assistant I’ve downloaded one blue print from forum I’ve tried to configure them but I don’t think I’m doing right and it doesn’t work properly, would someone be kind and guide me to make them work ?

[silk plover]

Anecdotally Eve does multiple small sequential updates via test-net, and every so often does a larger main-net all inclusive firmware update.

[heavy dirge]

This is the latest main version available. We are still waiting for an upgrade.

[heavy dirge]

Add endpoint-level icon selection to the dashboard by deriving icons from each endpoint’s DeviceTypeList https://github.com/matter-js/matterjs-server/pull/257

[lament vessel]

Can anyone explain why there are so many unknown nodes in the thread network. ZBT-2 and Tado Wireless x as my TBRs

[raven grotto]

In my case my HomePods and Nest Hubs show up that way

[lament vessel]

See I don’t use multi admin. My matter over thread devices are only in HA. So I don’t understand what the unknown devices are.

[raven grotto]

Well does the number of devices add up

Should be process of elimination, I think it’s down to the Matter version supported? Could be talking out my rear lol

[raw veldt]

As I understand it, this is the nature of Thread. Thread is self organizing. The mesh tries to have 16-23 routers. The limit is 32. In general (and oversimplified):

• If a router eligible device can see a neighbor node that can't each the mesh, it will become a router itself.
• if a router eligible node sees the mesh has less than 16 routers it'll become a router.
• If a router sees that there are more than 23, one will stop being a router (so long as it won't cause a partition).
  Border routers are a separate function to a mesh router.
  Complication: border routers can also do TREL - they can become a thread router but tunnel packets over wifi/ethernet instead of thead mesh.
  Anyway, as I understand it, the router funtion is a virtual node. They aren't a commissioned device that Matter has any involvement with so they show as unknown. They're ephemeral and come and go over time. They have a separate address so nodes can use a radio address filter and stop receiving packets if they shut down the router function.
  However, people like to know what's going on. All these vague "unknown router" things drive people batty. There is an optional matter cluster "ThreadNetworkDiagnostics". It has thread routing tables, neighbors lists, etc etc. IIRC this is how graphs attribute the "unknown" routers with a physical device. But it only works if the devices have the optional ThreadNetworkDiagnostics cluster in endpoint 0.

The openthread.org docs have a lot of detail on the mechanics and the intro/primer docs are approachable to read.

Its been a long time since I've looked, I'm probably misremembering some of this.

Caveat: things like the Eve app Thread network page doesn't know about things like TREL. It's quite difficult to generate an accurate map and the eve app often gets things wrong these days. Or breaks.

[soft idol]

Is it possible to disable nodes from being used somehow? Like if there's a smart bulb with a regular switch that you don't want to be used?

[raw veldt]

To my knowledge: no.

I can see why you might want that though. For what it's worth, I have two smart bulbs (nanoleaf). I used to have the switched outlet for it bypassed (with a wago connector), removed the switch, and put a blank plate in its place. Later on when we had electrical work done in the house, the electrician "fixed" that properly.

For the second bulb, it's just plugged into a non-switched outlet.

[fringe fjord]

Are they matter? If not then that’s something for a different channel I guess.

[fringe fjord]

Yeah as said testnet -> be careful

[left rose]

Plus, being a (border) router is essentially a function of Thread (transportation layer), not Matter (orchestration layer), so the matter server only uses the information available from devices on the matter layer. The information about thread topology is more accurate/current/complete within the thread layer itself, centrally available (as opposed to having to ask every node) and can be asked from an API-enabled OTBR or using e.g.:

ot-ctl meshdiag topology children

which yields a graph data structure as a whole.

[grizzled burrow]

I've done everything I can think of to reset the 'stuck' status of the Ikea button update. This has included a full HAOS VM reboot. Matter server 0.3.7.

[silk plover]

Assume you pulled the battery on the device itself at some point?

Definitely waiting for Matter Server 0.3.8 release before I try updating those again.

[cerulean saddle]

Heyyy I don’t know why but can’t add my Bilresa remotes to the matter integration while everything already registered works fine. I reboot ha, reset the remotes, try other batteries nothing work any idea did something change recently

[silk plover]

Are you using the Matter-JS server beta or the default Python Matter server?

[cerulean saddle]

Défault one. I am using zbt 2 adapter and followed défaut instructions

https://www.home-assistant.io/integrations/matter

I am using that

I am honestly not really expert on the matter. Any idea of the isssue ? Thread ? Matter ? Hardware ?

[compact root]

This channel is specific to the beta for matter.js. You'll want to post in #1284966617670881350 with the Matter tag, or #1284966353798697001 with the ZBT-2 tag if you think it could be hardware related.

[grizzled burrow]

Yup...indefinitely stuck

[silk plover]

Silver lining is that it’s an 8$ button and not a 40-65$ device.

[deep drift]

Maybe we should add „beta“ to the channel name?

[raw veldt]

Hrm. Something very strange happened a few hours ago and I'm not sure what to make of it. I'd love to hear some thoughts. I have a hybrid apple/HA(otbr) thread network.
Yesterday, I replaced a number of leviton matter devices with inovelli. This went well.
This evening, I replaced two more and added another. All hell broke loose.
The first incident was the inability to commission (to HA) of two of the replacement devices. The phone app would connect, handshake, go into Setting Up mode. I'd see some log entries with its UDP address and then .. nothing. The phone HA app would report that it failed to set up the node.
I powered off all the apple thread gateways and was able to commission all 3 devices.
After setting things up in HA, I turned the apple thread devices back on (5x homepod mini, one appletv). Things seemed to be working. Until I tried to share the devices from HA to apple. In the HA device info UI a click on the "Share Device" would activate pairing mode and show a QR code. In apple home, trying to add that device from the HA QR code lead to an instant pairing failure.
Turning off the apple thread gateways again - leaving just one non-thread apple home controller - and I could share the devices fine.
After a bit of head-scratching, I turned the apple thread gateway devices back on again. A little later I tried to pair an ikea (temp/humidity sensor) device. Absolute failure to pair again.
Turned the apple thread gateways off again, and pairing succeeded with HA.

I don't know what to make of this. I cant leave the homepods turned off or the spouse will have Things to Say. I don't see any kind of messages in either the otbr logs nor the new-matter-server logs that give me any useful hints to go on. I did note that the thread fabric ipv6 prefix changed from fd0e:... to fd1e:.... earlier today but that's happened before.

With the apple thread devices powered off, the timmerflotte ikea devices are doing their firmware updates (and succeeded. 1.0.18 -> 1.0.21). With the apple thread devices powered off, sharing the matter node from HA -> apple home (with its one remaining non-thread appletv) worked just fine.

Ideas for what to do or look at from this point? If I were to speculate, I'd guess that perhaps network size reached a threshold (31 nodes now) and apple maybe doing TREL more aggressively? Or maybe there's an disagreement about security keys? It's a bit frustrating because I can't see anything obvious for a clue.

[raw veldt]

Hrm. Maybe just a coincidence but I powered the appletv with thread on first and let it sync up for about 5 minutes before turning the rest of the homepods on. The problem didn't occur. Maybe the mistake was having all the homepods and appletv in a HA group and powering them on/off at exactly the same time? Nah, surely not .... (Why do I have them on remote power?? It used to be that they interfered with OTA updates. But this doesn't seem to be a problem with OS version 26.x. Maybe they fixed the mdns forwarding issue?)

[fringe fjord]

Interesting but in fact this shopws how fragile thinsg sometimes are with multiple BRs

[silk plover]

Hrm. Maybe just a coincidence but I

[fringe fjord]

Hey all, I just release version 0.3.8 of the matterjs based OHF Matter server (Changelog: https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#038-2026-02-16) We have again multiple optimizations. Especially OTA updates should not be more stable when crashing or such. There are still potentially timeouts on device side according to the error case of 5-105 mins but all should be more stable now.
Also the Thread graph has a good chance to show less "unknown" nodes or mybe label them more clear.
Thanks for your feedback.

@raw veldt I also re-thought about the version topic and how many to store... so in a later versiuon (not 0.3.8!) I will adjust to storing multiple versions.

[junior sedge]

oh, neat, the batch invoke change made it in? I saw that my aqara t2 bulbs claim to support it (MaxPathsPerInvoke=5), i'll have to try that out.

is there anything in the logs to indicate whether it's being used?

… either it's not working, or home assistant doesn't send commands sufficiently "immediately" for it to do it.

[silk plover]

The Bilresa still timing out, almost certainly a device issue, but no longer getting stuck in an endless Waiting(querying) loop in matter server, so additional update attempts no longer blocked.

Had 31 Inovelli test-net firmwares to update, so far no issues, other than its been 30-60 minutes post update for a couple and the Matter device page still showing the pre-update firmware version. Apple Home which was slower to show the new version for me in the past has already changed to show the latest version is installed. Did a re-interview of the node from Matterjs-server and no change on HA device page firmware version so far.

[fringe fjord]

show me a debiug log ... but it can easiely nbe that different "sequenceal" websocket commands are not coming fast enough for "at the same timepoint". We can later tweak the processing time ... but thats a balance as discussed (nd my voting above was not really attracking anyone) 🙂

[junior sedge]

opened https://github.com/matter-js/matterjs-server/issues/270 with the debug log

[fringe fjord]

thats still an open issue where I need more logs and time (and device updates) to test myself... normally the updates shpuld come in when device reconnects ...but I need to see. so any devic logs of update processes please add top the relevant issue

[junior sedge]

I wouldn't be surprised if some changes are needed on the HA integration side to get the batching to work (or maybe even a new websocket api that allows sending multiple invokes in a single websocket message)

[fringe fjord]

sure

[compact root]

Maybe in the future. Right now the intent of this channel is for the new server itself. There is a plan of it being locked down rather than changing the name, but that doesn't mean it's the only plan.

I'll consider this.

[silk plover]

So the bilresa I mentioned timed out a couple times did ultimately update to test-net 1.9.11 apparently but it did so without indication of success after the HA update window closed or timed out.

Trying a second bilresa update now from the matter server node detail update to see if that looks any different.

Second bilresa updated on first try. Both came back up with the pre-existing manually input node label missing as reported previously, which is likely a device manufacturer issue.

[raw veldt]

FWIW, the ikea Timmerflotte battery temperature sensor update (which worked, yesterday's 0.3.7 build) was a little odd. The UI showed the Downloading progress get up to 98% then it just disappeared off the network for a few minutes. Came back and the firmware was the new one after having impatiently hit the "re-interview" button.

[raw veldt]

I think I am inclined to call this a Footgun incident for now and assume that I provoked a problem on the apple side. I don't think it's too much of a leap of logic to assume that starting ALL 6 BRs of an ecosystem at the exact same time could lead to things getting upset.

[raw veldt]

Hah, this is a little more noticable 🙂

[grizzled burrow]

Upgraded to 0.38 but for multiple days the Bilresa updates are shown as stuck. Did a full HAOS VM reboot and power cycled the buttons. Help!

[raw veldt]

What is going on with those devices, anyway? It looks like they crash, drop off the network, and eventually restart during a transfer attempt.

[silk plover]

Does the updating bilresa still function as a button? You may have to factory reset it as a last resort.

If that doesn’t work, you may have , unfortunately, sacrificed that bilresa to the test-net dcl gods.

[grizzled burrow]

Yes they function just fine.

[silk plover]

If you can get the update to complete the bilresa now at Matter spec 1.4.1. Still no Binding cluster though.

[raw veldt]

When I was tinkering an hour ago, it looked like it was doing its Thread "I'm a sleepy device, I'll go to sleep now" while the OTA was running. Pushing a button every second made it run much longer - got to 88% vs the usual 8-12% - but it eventually fell over again.

[raw veldt]

Fingers crossed

Unrelated, but a wishlist request. Could we get names or something on here?

[silk plover]

You can manually input Names (node labels) now with the Beta. Node details page then select into Endpoint 0/Basicinformation cluster and at the top click Node Label to reveal the input field. Then hit save after.

[raw veldt]

Oh, how did I miss that?!

[silk plover]

Be aware the node label is saved on the device and the ikea bilresa is among a couple devices that has been reported to lose the Label on matter server restart. Eve smart plugs are the other by memory.

[raw veldt]

Oh my, it made it!

[reef crane]

I've powered down all my HA OTBR except one. I leave all my apple Homepods (2) and appleTVs (3) running. I have not and will not join the apple thread network with HA thread network. I had a lot of problems with multiple border routers early on. I'm planning on moving all thread devices off the apple network. My plan is to use HA as thread network and multi admin all user controllable devices to Apple. The HA documents say HA does not support firmware updates if Apple thread network is enabled, at least that's how I understand it. I think one HA OTBR only is optimum for now. When all the software stabilizes, maybe multi-vendor thread networks or multiple OTBR will be ok. Keeping it simple for now.

[raw veldt]

The docs do say that (re: OTA vs apple) - I encountered it constantly and that';s why I have remote power to kill the apple thread border routers. However, it was a specific bug on the apple side where something like they didn't forward an mdns query so the client could find the OTA server. That seems to be fixed now. I have not had that kind of problem since *OS 26.

[silk plover]

OTA works with a joined Apple thread and ha otbr since a little before iOS 26. OTA Updated 33 devices today and have done similar over the last year or so.

[raw veldt]

However, I would very much prefer to have a HA-only thread network and export selectively. Unfortunately, as I understand it, it wouldn't help with multi-admin-induced traffic at all though. Whether there was one HA-OTBR or a hybrid of 10 BRs, the thread load is still the same with multi-admin.

I would just like to leave the homepods alone while messing with stuff in order to avoid Spouse-aggro.

Multi-admin certainly causes some surprises. eg: Apple sees the composite Inovelli wall switches and their RGB ledbar as something that would be a great match for Apple's adaptive lighting.

[grizzled burrow]

I ONLY do multi-admin for a handful of devices that have an iOS app that I deem 'required'. Eve Motion blinds, Aqara U400/W200, and I think that's it. Everything else I use "Matter Hub" to expose 80 devices to Apple Home. Rock solid and avoids bogged down Thread networks.

[raw veldt]

That's what I was planning to do at some point.

[raw veldt]

Pity that's archived/abandoned though 🙁

[grizzled burrow]

no it's not...there's a VERY active fork

[raw veldt]

oh, nice! pointer?

[grizzled burrow]

https://github.com/riddix/home-assistant-matter-hub/

EXTREMELY active development. Seems like nearly daily updates.

[raven grotto]

Hell yeah thank you, I was mourning the loss (although not enough to find this fork)

[raw veldt]

I had been using the homekit bridge to expose some zigbee stuff. Thanks for the pointer!

[raven grotto]

I used it to get my vacuums into HomeKit, even though for now it just does start/stop lol

[raw veldt]

For what its worth, zigbee stuff has been driving me crazy. eg: the temp sensor in the garage that stubbornly uses the router on the far side of the house that it can just barely maintain contact with and chews up its battery to do so. vs using the repeater in the same room. I'm kind of hoping that matter/thread is a bit smarter. Zigbee has a lot of devices that are stubbornly resistant to any sort of self-organizing.

[buoyant nova]

Off topic, but with both ZHA + Z2M, you can allow joins via a specific router only and reset the device again. If you're lucky, it sticks to that router and doesn't jump to another one.

[lament vessel]

0.38 - still showing lots of unknown devices

[raven grotto]

Amazing, thanks again for the heads up

[fringe fjord]

Not really. You see external devices vs routers. So is really all your thread stuff only commissioned into HA?

[lament vessel]

Yeah, they’re only in HA using the ZBT-2. So I’m not sure why it’s showing up like this. The second router is the tado wireless bridge but when it was only the ZBT-2 it was like this as well

[raw veldt]

I only have 7 unknowns now. Also 7 apple TBRs. And no islands.

[lofty moth]

Why must switching from Google thread border router to HA be a headache... no need to answer, just venting frustration.

[junior sedge]

it really shouldn't be, if you can get the google thread credentials into ha from your phone, then HA can use those to join the network, and then you can just turn off the google hardware if you no longer want to use it.

[raw veldt]

I've thought about this many times but with Apple for many reasons. Some good, some silly. If I could have HA create a thread network of its own, I would have, and then selectively proxied things I wanted to share with apple home via homekit bridge (or now matter hub). Sure, I could share a thread network with apple and use its homepod border routers but when weird things happen (eg: the OTA update problem that plagued us for years) I'd rather have a clear idea of where the problem happened.

I actually thought about resetting the HA OTBR and put the the dongle in a foil enclosure so it can't hear other networks and thus trick HA into creating its own thread network/credentials. But that seems a little excessive.

Oh my, I hard crashed the server. It bounced.

2026-02-16 20:37:08.561 FATAL Unhandled Unhandled error detected: [unknown-node] This node is decommissioned and cannot be connected to.

I'm not entirely sure how I got there. Approximately:

• Spent about 10 minutes trying to factory reset an old battery device to make sure it actually wiped its credentials rather than just rebooting for 10-20 seconds. I went through 2 reboots.
• pressed its button to see if it appeared in the logs.
• did a re-interview a few times
• at some point it was genuinely off the network. Thread/Matter marked it as offline.
• I think a re-interview was still pending, retransmitting.
• I deleted the node in the new-matter gui
• server crashed and restarted

[reef crane]

I don’t know a way to migrate HA from a multi-vendor Pan ID once it has joined an HA Pan ID.

But I made a couple of HA OTBR with independent Pan ID. I can’t remember the details, but it wasn’t hard. It’s a lot of work to reset all the thread devices and move them to a new fabric. I was able to leave the old system running in parallel with the new, so I had two HA fabrics running in parallel. Old system continues to run automations until i removed the devices.

I first multi admin all the wifi devices to the New HA Hub from Old HA. Then I removed the thread devices, one at a time, from the Hub hosting its fabric, both the Home and HA devices and factory reset it. Then pair it with New HA Hub and multi admin to Home. It took a month. Finally, I deleted the Old HA matter over WiFi pairings and shut it down forever. Only had to factory reset the Thread devices, not the matter over WiFi.

I also avoided mixing zwave and Zigbee devices into it. A clean Matter system, no third party software, just Matter over WiFi and Thread devices and Matter specified multi admin.

I used a new HA system (Pi 5 with ZBT-2) to create the new fabric. Probably turned off all the other HA OTBR to create the new system using the Pi 5 HAOS download. It just came up as a new fabric. I rebuilt the old system and copied the TLV dataset from the New HA Hub to make it a second OTBR, but ultimately decided to stick with one OTBR for now.

[raw veldt]

Did you have HA or the OTBR addon create the thread network? The native openthread gui is a bit vague. There's also warnings in the HA docs to not do it in the OT gui.

One thing I noticed: apple seem to like to change the on-mesh prefix periodically.

[reef crane]

I installed HAOS, Matter Server was automatically installed, I installed the OTBR with the ZBT-2 and I think that created the new fabric. You can inspect the Thread specs from the OTBR integration to make sure you have a new fabric. And you need to turn on the matter.js beta switch to get rid of the Python Matter.

[raw veldt]

The docs have some real gems in there. 1) it will offer to create a thread network only if there are no existing thread networks. 2) if there is a thread network, you can only join one of them. 3) fun: "Note: The preferred network function isn’t completely implemented yet. When adding Matter devices through the companion apps, the preferred network of the mobile device is being used." That implies that commissioning a device with your phone puts the device on the phone's thread network regardless of what HA wants.

Its a bit ambiguous on https://www.home-assistant.io/integrations/thread/ - at least the way I read it. On one hand it says its the phone that picks the thread network to use, but there's also bits about syncing credentials to the phone. There's also

There is also a web interface provided by the OTBR. However, the web interface has caveats (e.g. forming a network does not generate an off-mesh routable IPv6 prefix which causes changing IPv6 addressing on first app restart).

[reef crane]

I never shut down my apple network. If you click the HA fabric button (in Thread integration) that says something like “share thread credentials with phone”, the add device utility will use the HA fabric. When you move the first thread device, be sure and check the device info and make sure it is on the right fabric. You have to make sure delete the thread device from both Home and Old HA before factory reset and pair to your new fabric. Mistakes are not allowed on deletion from Home (ok if left on HA, but causes confusion).

You can copy the YAML for all your automations for documentation purposes so you can rebuild them on New HA Hub.

[raw veldt]

Mistakes cause spouse aggro.

[reef crane]

It is a treacherous process. I backed up to my hard drive every few devices. I had to rebuild from backup a few times, which worked great.

[raw veldt]

Spouse: what did you do to the lights now? it used to be simple. Why can't we just have the normal on/off switch like normal people?
also Spouse (on a different day): via text: can you turn off the lights for me? I dont want to call Siri and wake up the cat.
also Spouse: Don't make me go over there to turn the lights off. Its so long since I've used the switches that I dont remember which is which.

[reef crane]

Yes, been there, done that. Spouse only uses Apple Home. All scenes are duplicated on Home. Spouse never sees HA. Siri does all voice control. Only multi admin devices to Home that need to be controlled by people. All automations are HA.

[raw veldt]

The same rules apply to home internet. A great number of demerit points can be assigned if you break the internet or TVs. We have boring internet at home so I'm not tempted to tweak anything.

For what it's worth, that is my motivation for Matter/Thread. I want the essentials to work via local control and not break if there's a wifi or internet hiccup. I spent a few years with Insteon which was all local control. I have no taste for cloud stuff.

Zigbee or Zwave would probably be a more logical choice but where's the fun with those?

[reef crane]

I left Zigbee and Zwave running on SmartThings. Replaced almost all of them now. I was running 2 HA, 1 Apple Home, and SmartThings paired with Alexa to control zigbee and zwave for a few months. Amazing it could all function in the same house, but it did. A lot of work, but wanted a pure system moving forward. Almost there..

[raw veldt]

I am keenly awaiting the Ikea GRILLPLATS plugs. Rumor has it April-ish. I used to be an Eve fan but they seem to have lost their way since being bought out.

eg: Eve Light switch. Firmware 3.5.1 was certified in june last year. Not available to users. Everyone is stuck on 3.4.1 from August 2024, which crashes periodically.

[fringe fjord]

Issue with the log please. Here it is gone too fast

[fringe fjord]

Just to be sure … pan id ? Or different channels (and pan ids)? Using two thread networks on same channel will not bring any benefit

[fringe fjord]

Yeah. Somehow it seems they restart their border routers sometimes and this then leads to that.

[fringe fjord]

Yes this is because emit all ecosystems really allow you to choose to use an other network. (Apple might work. Google not when intensiver right)

@raw veldt @reef crane please limit conversation to matter server topics. I get the feeling we got „a bit“ offtopic above 😉 thanks.

[left rose]

Is the matterjs server App (Add-on) somehow affected by the current HA/DockerHub issue?
https://github.com/home-assistant/core/issues/163198

Hesitating to restart the matter server fearing it doesn't come up again because it couldn't download stuff...

[fringe fjord]

I can only guess but it shoukd not try to update anything on docker. updating the matterjs server is from npmjs.org

[fringe fjord]

tracked and fixed by next verison https://github.com/matter-js/matter.js/issues/3237

[quasi arch]

Hi all, on the new matter server, are the following suppoered ? Engery Management things

I can see the logs coming in on the matter server and decode the TLV, but HA doesnt present these

[fringe fjord]

if thats the case then also HA would not present these with the python server 🙂 I thought that electrical measurements are supported and also discovered. But thats then a more HA question and not that much server related

[fringe fjord]

FYI: It seems that a change introduced in 0.3.7 that fixed processing of command fields introduced a different handling leading to issues with locks ... fix comes in next version of the server tomorrow - issue to follow if interested https://github.com/matter-js/matterjs-server/issues/271

[quasi arch]

thanks, I didnt know if it was the server ignoring or HA not knowing what to do.

[fringe fjord]

As said in theory should all work ... details need to che checked in detail

[flint compass]

Question about the "Thread Network Mesh" ... should there be an icon for every Thread Border Router in the mesh? I have two Apple TVs and 4 HomePods, but I only have three Unknown Devices listed as routers in the mesh. Looking at the Thread app, I see all 6 devices as TBRs for the thread network. Also, is there a way to figure out which "unknown device" is which Apple TBR?

[haughty stump]

dumb question.. how do we update matterjs-server to current release 0.3.8? I have the beta flag on python matter srever and it's running the new version but I can't see or identify which version matterjs-server is running? Does it auto-update on restart or something?

[junior sedge]

The Thread stuff shown in the Matter server comes from Matter diagnostics, so it's only able to get information from Matter devices. The thread border routers themselves are not Matter devices.

(although technically there is a matter device type for thread border routers, but i don't think anything has implemented it yet?)

[deep drift]

Simply enable the Beta switch and restart the Matter App. Matter.js will be automatically downloaded, installed or updated with every restart. Additionally you can enable the Matter App in the sidebar.

[reef crane]

It would be a great feature if we could use the “Customize” feature of the Matter “Device” display to show Node ID in addition to attritibutes like device name, model, battery, floor, display, etc. It is hard to determine the correct Node Label with only the device type and Node ID shown in matter.js.

I had 5 devices that had their Node Label erased when I updated from 0.3.7 to 0.3.8. The character count was set back to 0/32. They were all Eve devices. I had 4 Node labels erased when updating to a prior release. Some of those were not Eve.

[grizzled burrow]

That's why I maintain an Excel spreadsheet. Shouldn't be needed, but......keeps my devices straight though.

[heavy dirge]

Will it be possible to release version 0.3.9 (including the patch) today?

[deep drift]

General question: Is there any update on the discussion regarding how the future integration of the three components - HA, Matter.js, and OTBR - is envisioned?
I’m particularly interested in whether there are plans for consolidated, comprehensive data across all three layers. For example: Unified friendly names; Consistent areas/rooms/categories/labels across all components; End-to-end diagnostic data that can be visualized in a consolidated way.
Please don’t get me wrong - there has been great progress on all three sides, which is highly appreciated. I’m simply curious about the longer-term integration vision and how tightly these components are expected to work together in the future.
Perhaps this could also be a topic for the next community sync?

[heavy dirge]

This is a next step, but this thread is dedicated to testing the new Matter server.

[fringe fjord]

The data come from the devices, so it is in what the devices "see" ... I can not say more or different than that 🙂 So ... maybe they are passible or not really used or or or ... no clue.

[fringe fjord]

thats my plan still

[fringe fjord]

In general I would say "idea phase" 🙂 but not yet any concrete details. But sure bring it up in the next community sync

[grand kite]

Trying out the matter beta for the first time. Using the old python server, it would generally take 15-20for my network (with 71 matter devices) to come fully online. Looks like the beta will take significantly longer. Maybe this is a one time thing though on first use? I've got all kinds of network logs I don't understand yet scrolling through the logs.

[fringe fjord]

The very first start of the matterjs server takes longer because we need to migrate data and then discover and fully reconnect all devices including loading all data. Give it it's time. When all nodes are online again then you can try a restart and this should go faster.

In general it also is relevant what exactly you measure. Whenyou measure from the "klick restart button in the UI" then it includes downloading the newest version from npmjs and some more things which will be better when we have it as real version.

Else I'm always happy to check logs and I can tell you whats up

[grand kite]

It's no big thing, I was mostly just curious. I was measuring just from when building the netwrok started (so not the NPM download). It seems to be moving now though. I'll let the first run complete and stabilize then maybe do another one for a fair comparison. Thx!

[fringe fjord]

So or so love to check the log ... both ...so from this initial start and also from a later restart if you like. Just send me here in discord as PN ...

[grand kite]

OK, ya a 2nd run was much faster. Not bad. Took about 20 minutes (not counting the npm download & build). 15 minutes (so about the same) building the network, and about 5 minutes before that in some new setup steps. mostly this:

[fringe fjord]

if you have info level logs from initil and the restart just drop over ... I am analyzing them curremtly

So would be interesting ... e.g. some logs start from these "create paired node" till network is on is 4:30 for 150 devices ... so would be interesting how yours look

[grand kite]

let me see what I can grab

[flint compass]

How often does the thread mesh map update? As a test, I unplugged all 4 HomePods and left just 2 Apple TVs managing my Thread network. However, about 8 minutes later, I still see the same three "External" thread BRs in my mesh. Not sure why there are three now instead of just two.

[fringe fjord]

dependson the devcie. The mesh view when you klick on a node has a refreh button at the right side ... klick it and you can force reload

[fringe fjord]

So guys, 0.4.0 is out ... (Changelog as usual: https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#040-2026-02-19). A Breaking change is only relevant for the Docker users, HA Addon is not affected!
Other than that should be again more stable, reliable and the two issues last days reported (and more) are fixed.

If you have many thread nodes (>50 or such) I would love (again) to get logs from your restarts with this version if you have time and can do it (info loglevel is sufficient) for analysis. Just send as PN.
Thanks for all infos.

[grand kite]

Giving this a shot. FWIW, it doesn't look like the server version is recorded in the log anywhere on startup? Perhaps it should be?

[fringe fjord]

it is

[grand kite]

I'm not seeing it? Perhaps I'm just missing it though. Lotta logs messages.

[fringe fjord]

some lines below the "Commnd line:" which should be the first real server log

[grand kite]

ah, I see it now.

It was below all the LegacyDataInjector logs, and I didn't look down that far

Startup seemed a bit faster (will send you the log in just a sec) but my Aqara U200 lock appears to not be connecting (it was fine on the previous version).

[fringe fjord]

give them a bit ...and yeah send logs

[grand kite]

Ah, ok. It picked up the lock too (yay!). Just took a bit of extra time.

[fringe fjord]

I can tell you why tomorrow ... but I guess thread ...

[shadow horizon]

Dumb question as someone who has been behind for ages now but has been invested in thread since the first handful of devices and has 30+ at this point - how do you see the mesh map with only Apple TBRs? I thought you had to have an OTBR for this 🤔

am idiot this is not in HA accessed on default port 5580 and works fine with Apple only TBRs - neat.

Also to be relevant I recently migrated to the new matter server just before latest release and had no issues - no real delay on first startup either despite all the devices - though my HA instance is on arguably beefy hardware.

[heavy dirge]

Thank you, I don't have any exceptions. It works with the new version.

[reef crane]

Log for upgrade to 0.4.0. Upgrade at 14:53

[silk plover]

Had previously updated 2 Ikea bilresa to test-dcl 1.9.11 and for the first time in a long neither lost/deleted the manually input Node label with a Matterjs server Update/restart (updated to 0.4.0). Hopefully Ikea fixed whatever was off in the firmware causing this previously.

[silk plover]

BTW , the map only uses routing table and neighbors list reported back to matterjs-server by Matter devices that have an optional threadnetworkdiagnostics cluster. So apple TBR are pulled into the Map when a thread devices connects to them and reports it back to Matter server. Same for a HA OTBR. Both would be listed as external device. no TBR/OTBR info used directly.

[grizzled burrow]

0.4.0 seems pretty broken? In the node server view nearly everything is listed as offline 5 minutes after a full HAOS VM reboot. Normally with the beta most everything is online in <5m

Lots of: [discovery] @1:78 is not reachable right now.

[reef crane]

All my devices came back quickly, thread and wifi.

[grizzled burrow]

9 out of 73 in 10 minutes

[reef crane]

Bummer.

[grizzled burrow]

20-25 minutes later things are back online. A far cry from the typical 5m

[reef crane]

I did not have any Node Label losses. My devices were back in 5 minutes.

[silk plover]

Collect and send those logs to @fringe fjord with a restart

[silk plover]

0.4.0 restarted and sent logs to apollon, 90 of 92 nodes up in about 5 min, 2 Ikea window and door held out for about 10 min total. All manually input Node Labels retained (only ever been an issue with 2 Bilresa when they were on the original firmware)

[fringe fjord]

Yeahh collect the full log from
The start and send me as pn. I have a look

[fringe fjord]

But also here measures from „I completely restarted HA“ can be completely misleading because on system start a lot stuff goes on and all depend also on CPU and io. But log will show what happened

[deep drift]

Yesterday, 0.4.0 installed and started flawlessly. Today, I restarted Matter.js and I saw something like an exception in the log:

2026-02-20 07:31:36.904 INFO PairedNode @1:10 Simple re-establishing session did not work. Reconnect ... [received-status-response] Received error status: 128 (SubscribeResponse)
at InteractionClientMessenger.throwIfErrorStatusMessage (/opt/matterjs/node_modules/@matter/protocol/src/interaction/InteractionMessenger.ts:199:19)
at InteractionClientMessenger.#nextMessage (/opt/matterjs/node_modules/@matter/protocol/src/interaction/InteractionMessenger.ts:177:14)
at async subscribe (/opt/matterjs/node_modules/@matter/protocol/src/action/client/ClientInteraction.ts:713:37)
at async QueuedClientInteraction.subscribe (/opt/matterjs/node_modules/@matter/protocol/src/action/client/ClientInteraction.ts:754:28)
at async PairedNode.subscribeAllAttributesAndEvents (/opt/matterjs/node_modules/@project-chip/matter.js/src/device/PairedNode.ts:1093:30)
at async PairedNode.#initialize (/opt/matterjs/node_modules/@project-chip/matter.js/src/device/PairedNode.ts:882:37)
at async PairedNode.reconnect (/opt/matterjs/node_modules/@project-chip/matter.js/src/device/PairedNode.ts:663:17)
at async StandardTimer.callback (/opt/matterjs/node_modules/@project-chip/matter.js/src/device/PairedNode.ts:1205:89)

Unfortunately, I only caught it logging on Info level, but I retained the full log. I restarted again with Debug log level activated, but all seems ok. So I could not reproduce it. In all cases all devices came online and worked without issues.

[fringe fjord]

yes we have some places where we log full stack traces also for "can happen" errors just in case to ensure we see where it comes from for optimization.
if you still have a log from aroundthat time send me, but usually it is such a "can happen ... will be retried automatically" case

[deep drift]

Just sent the log via PM.

[fringe fjord]

Then also for the reound here: In some cases logs on some error cases a a but too verbose still ... but in that case subscribe errored once because device responded with an error, worked 2s later

[shadow horizon]

Yeah noticed that thanks with those and a lot of my non-matter thread only gear. (Early adopter problems 😂)

It would be nice to be able to label them accordingly myself, happy to take a look at PR’ing such functionality in the next couple of weeks if it would be desired by the devs.

[fringe fjord]

You can set the device "node Label" (Basic Information cluster). This is then already used. Else we would need to maintain extra names, this makes no sense. One idea is to have such a map also in HA then names from HA can be mapped, but currently the server does not have these information (and also do not need them!), because the server can also be used standalone from HA

[meager zenith]

Is anyone else getting this error or is it just me? I can't find any relevat errors in the supervisor log.

[meager zenith]

@fringe fjord Any chance this gets implemented in the near-term? Would it help if I moved it to the matter.js git repo? I've been questioning if I put it in the wrong place. Thanks for everything you do my friend!
https://github.com/orgs/home-assistant/discussions/2518

[fringe fjord]

Maybe create additionally an issue in the matterjs server. Basically for this we need to enhance the websocket protocol and this can only be done once the python server is gone. so „near term“ I can not promise.

[meager zenith]

Gotcha, thank you and happy Friday!

[flint compass]

Just a quick shoutout to the devs who've made Matter.js possible. I know there are issues here and there, but, man, in my experience, the difference between the Python Matter Server and this beta is night and day. Previously, my mesh was never reliable, with periods almost every day that devices would go offline for 30-60 minutes making living in our house annoying. Oddly, even watching a 4k movie on our Apple TV would bring down the mesh. Now, I hardly have any disturbances at all, and if I do, they resolve in seconds or minutes. Also, we could never say "Hey, Siri, turn off the 2nd floor lights" without bringing the mesh down. We had to do it room by room. Now I can say "Siri, turn off all of the lights in the house" and it works. Also, I like my lights to transition from daytime to evening scenes over 20 minutes every night, but that always brought the Python matter server to its knees. Now it just works. On top of that, being able to visualize the thread mesh is so helpful in beefing up problem spots and making changes. And, finally, to have access to the devs in this channel and to get such quick responses is incredible. Thanks to @fringe fjord and the dev team! I'm very excited to see where this is going!

[fringe fjord]

@flint compass great to hear that it makes a difference. That basically was the plan.
If you like -maybe after release of the next version. Grab an info log after the update and after all nodes are back - or even longer timeframe including such mass actions and watching a stream. And send me the log. I love to look into such problematic setups to see if we can improve things.

[fringe fjord]

Soo, good morning ... new weekend time new version 🙂 0.4.1 comes with some improvements and a fixed custom cluster definition. https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#041-2026-02-21

Again- because I optimized the connection logic in some places would be cool if the "big thread network users" could send me a startup log. thanks a lot

[mint thorn]

How many devices would class somebody as big? I'm willing to turn on beta to get you logs if I qualify

[fringe fjord]

😉 so far the most interesting users had 50+ thread devices. WiFi is usually simple 🙂 and if you decide to then the first start after switching on beta takes longer naturally. Then see that all devices are back and then do another restart. I am happy to take both logs … but the initial one is special anyway

[mint thorn]

A few more needed. I will reluctantly purchase more devices so I can help 😝

[fringe fjord]

LOL. Any feedback is interesting. So give the beta a try and report back your results even without logs. Ideally all goes well and all is stable.

The more feedback we get the better. Also smaller systems count. The „bigger“ ones are just the more challenging ones to get them rocking stable.

[grizzled burrow]

I have about 74 Thread devices

[fringe fjord]

Great. If you like send me a info level log of startup or longer. As you like. But when longer ensure startup is included or extra. Thanks.

First logs I saw from 0.4.1 are as expected. The optimizations work as expected

[grizzled burrow]

ok with 0.4.1 all 73 were back online in 5 minutes

[pearl crown]

Hey guys, sorry if this should be obvious but I'm new here and not sure how to upgrade to the latest version. I enabled the beta and I appear to be running 0.3.7 but I'm not sure how to go from here to 0.4 or 0.4.1.

[silk plover]

Just restart the matter app.

Restart the matter server app and it will auto update to the latest beta automatically.

[pearl crown]

Awesome, I should have at least tried that. 😆

And for anyone curious, two things brought me here - I occasionally have some IKEA MYGGSPRAY motion sensors going offline until I pop the battery in and out. And the new matter server is also showing some "timeSynchronization.timeFailure" notifications for my IKEA ALPSTUGA environment monitor. Still learning!

[raw veldt]

I'm having some issues when updating firmware on Inovelli White matter/thread switches. For background, going from 1.0.0 to 1.1.5 moves endpoints and attributes around. The aftermath of this is that HA has a bunch of dead controls on the device page. eg: before the 1.0.0->1.1.5 update, there will be a dimming speed setting. After the update, there will be two. One causes an error popup if you click it, the other may or may not work. Things I've tried:
Matter App restart (Settings -> Apps -> Matter Server -> restart): No effect
Device reinterview in Matter Server UI: No effect
HA Matter integration reload (Settings -> Matter -> Reload): No effect
HA core restart (Settings -> Restart): No effect
Full, complete, HA system reboot (a hard reset, Settings -> Restart -> Advanced -> Reboot): fixes it.
After that system reboot (Home Assistant Yellow, external SMLIGHT SLZB-07 otbr thread radio) the troublesome controls are grayed out on the device info page in HA, and an entity search for "Not provided" quickly generates a list for deleting them quickly.

What is supposed to happen here? Is there a less drastic option that I might be able to use to get these controls to sync up? Where might this be going wrong?

Is this a case of trying to hurry something up that will eventually get around to fixing itself with patience?

I've since seen docs on the Inovelli site where they recommend a HA restart to resolve this. Hmm.

[west kraken]

Hello i need some help i have all of my light switches in the house that are connected to homeassistent with matter and one day they all went ofline and how when i try to reconnect then in getting this error. iv tryed almost everything i could think of and still cant get it to work. in not even sure if in posting to the right place. Hopefully someone can help

[silk plover]

This is expected. Inovelli firmware moved many endpoints around from 1.00 to 1.1.5. Most are now in their own mode select clusters whereas previously many were in custom clusters. You will need to delete the orphaned endpoints. Nothing matter server can do about that.

[raw veldt]

Yep, the problem was HA wasn't showing that they'd been deleted. When I had a big list in the device info page it wasn't obvious which was which.

A hard HA reset caused it to resync against the matter server. I had hoped there might be a less disruptive way to force a resync.

After the hard reset, the config entries showed as "Not provided" and were greyed out.

[raw veldt]

I'm sorry, that's the old matter server. You're probably going to have to ask here https://discord.com/channels/330944238910963714/1284966617670881350

To try and put it another way: After the update, the new config entries appeared immediately in HA. But the old dead config entries were still active. Some had the same name as the new ones. It took a HA reboot to get HA to notice that they were no longer provided and were orphaned. I was looking for a better way to make HA notice this without needing to resort to the big 10 minute reboot hammer. My poor HA yellow needs a CM module with more ram and a reboot is painfully slow at the moment.

[cedar pawn]

Restarting is also part of inovelli's official guidance for that update https://help.inovelli.com/en/articles/13539955-vtm31-sn-firmware-1-1-0-update-advisory

[fringe fjord]

hey guys, 0.4.2 release but this time only minor stuff and syncing with matter,js release version, so no need for any new logs. should behave exactly as the 0.4.1 ... https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#042-2026-02-22

[fringe fjord]

time sync failure is expected. issue exists. we need to add support to sync the current time to the devices. for MYGGSPPRAY the main issues known there is that it becsomes unresponsive here and there bit then recoinnects. also here an issue exists where we try to get to the bottom of it.

[fringe fjord]

Better lets discuss that in normal matter channel because unrelated to the beta server!

[fringe fjord]

This log shows you use the olf Matter server, please use the other channel. But as hint: Error says that the device do not respond to any data. try power the device off and on again and see if it works, else factory reset and repair might be an option,

[fringe fjord]

Formally thats a huge issue because with such a behavior they break all controllers (not just Home assistant) ... thats just meehh 🙁

[cedar pawn]

Just made the transition to the new matter server, and the logs look (to my non-expert eyes) like the system handled the transition very well. Love the UI visualizations of the Thread and Wifi networks.

[silk plover]

They decided to bite the bullet and move towards standards now that matter allows AND when it only really affects HA and Homey users (none of the other major ecosystems display any of the changed options) instead of later when it could effect more people.

[silk plover]

Stumbled across some evidence that aqara has had issues with matter node label implementation for a while.
https://forum.aqara.com/t/matter-attribute-nodelabel-is-not-taken-into-account-for-bridges-endpoints/70738/7
https://forum.aqara.com/t/issue-with-with-nodelabel-attribute-of-the-bridgeddevicebasicinformation-cluster/2541/6

[west kraken]

thats the thing is that i tryed factory resetting the lightswitch and repairing it and thats the error i get

[fringe fjord]

again: wrong channel!
Factory resetting a device means you need to manually delete it from the controller/Matter server ...

[manic lichen]

Please use the #1284966617670881350 channel with the "Matter" tag as you have been told.

[west kraken]

i did yesterday and nothing

[manic lichen]

No reason to post here off-topic.

[west kraken]

i was just replying to a response that i got sorry

[fringe fjord]

I answered in the oter thread now ...

[pearl crown]

Thanks for the info. Glad to know that these are known issues and my network doesn't have some unique problems.

[bright rune]

Hi all, does anyone know a way to tell the Matter server to ignore an interface. I have a disabled wifi interface in HA but the Matter server keeps grabbing it and trying to use it. I have a multi lan network so its causing lots of err fun

[fringe fjord]

just set the interface to use in the addon settings

[bright rune]

The only option I have found to do that is on the matter server adding extra parameter of --primary-interface eno2. The addon / app in home assistant uses host networking so it seems to grab the interfaces from the host

[grizzled burrow]

@fringe fjord 0.4.2 is having terrible startup times to get all 73 devices online. I set logs to debug and will send them your way.

[fringe fjord]

Sure. First report of such issues. Let’s see

[fringe fjord]

So guys, here we go ... 0.4.3 released (https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#043-2026-02-25) ... nothing big jkust some smaller UI things for the dashboard and hopefully fixes the version update in HA after an OTA update.

[deep drift]

The descending sort order (LQI) doesn't seem to apply for external routers. The other devices' neighbors are sorted correctly.

[fringe fjord]

yeah ,.. will prepare for next version

[deep drift]

@fringe fjord Would it require much effort to add a search function to the Matter.js visualization?
My use case: my Thread mesh isn’t very strong, and a few extended addresses show up repeatedly with NoAck failures in my OTBR log. It would be very helpful if I could paste e.g. an extended address into the Matter.js visualization and have it jump directly to the corresponding node.

[fringe fjord]

create an issue ... I see what I can do

[bright rune]

@fringe fjord I'm sorry to bother you I may be asking in the wrong place should I be on open-home-protocols for the current matter server not the js one ? Still can't get it to start without grabing the disabled (in home assistant) interface and can't see anywhere to tell it not too. Have spent 5 days on this really struggling to find any ways to fix it

[shadow kelp]

Hello everyone, I have a quick question.

I’m using IKEA bulbs that support Matter over Thread. I first added these bulbs via Thread/Matter to my Apple HomePods. After that, I used pairing mode / the Matter setup code to additionally add the same devices to Home Assistant, so I can create automations there.

However, I’ve now noticed that I can also add a device directly to Home Assistant via the Home Assistant app using the Matter code. In that case, the bulb becomes part of the Thread network created by the Apple HomePods. To see it in Apple HomeKit, I then have to share the Matter device from Home Assistant and scan the generated QR code with the Apple Home app.

Is there any significant difference between:
• Adding the device to Apple HomeKit first and then sharing it to Home Assistant
vs.
• Adding the device to Home Assistant first and then sharing it to Apple HomeKit?

All devices are in the same Thread network created by the Apple HomePods. I am not using a Thread stick with Home Assistant — I am only using the existing Thread network from the Apple HomePods.

Thanks in advance for any advice.

[reef smelt]

Are you using the new matter server? If not, you should ask in #1284966617670881350

[fringe fjord]

I can not tell too much about the python server. So the other channel is better I guess

[fringe fjord]

Then both is equivalent.

[bright rune]

Thanks for replying, I have up and moved over to the js one which solved the problem. At least I can report any issues for that here if I find any and so far it's working well. Pity I can't get rid of the ghost thread network on my android phone.

[fringe fjord]

Hey guys, does anyone here that uses the new matter server has issues with ny bridges or such? In GitHub a user was telling that he has serveraldevices being "copletely unstable" with the matter.js based server ... I hope he delivers more information thaan this sentence 🙁

[supple basin]

No issues here with 11 IKEA thread sensors on ZBT-2 on latest fw with OTBR beta and matter.js beta server

[deep drift]

Maybe I missed it - What do you mean by "ghost thread network on my android phone"?

[left rose]

do you have a link?

[rotund drum]

I am currently having issues pairing my nanoleaf lightbulbs with matter over wifi - I had them paired previously but I removed them. I cannot get them to re-pair. I have 4 of them and initially only paired 3. I confirmed that I was able to pair the fourth one just fine (which had not previously been paired and removed). Has anyone else ran into this issue? I'm thinking I should probably submit an issue for it on the github.

[fringe fjord]

https://github.com/matter-js/matterjs-server/issues/187#issuecomment-3961313680

[fringe fjord]

if they were paired already, did you do a proper factory reset? Try to also update them via Nanoleaf app to latest version. Else ... what error you get? How you try to pair?

[rotund drum]

did a factory reset - updated to latest firmware. Paired through the python matter server -> commission node -> pair new wifi device

https://github.com/home-assistant/core/issues/164241

[fringe fjord]

You use the python matter server and not the new one, so sorry but thats the wrong channel here ... I will still add a comment to the Issue

[rotund drum]

I just added the app to home assistant like a month ago... Is that old now? Sorry

[fringe fjord]

This channel - so introduction - is for the beta version of the matter server … so when you select beta then it are right here. When not you use the official version. Which is handled in other channels.

[hollow cloak]

The topic of this channel is:

Discussions around the development and testing of the new matter.js-based controller. Find details for this testing phase here: https://github.com/matter-js/matterjs-server/blob/main/ALPHABETATESTS.md

General Feedback Requests related to Home Assistant should be posted on GitHub Discussions here: https://github.com/orgs/home-assistant/discussions/12

[bright rune]

While testing and setting up I had vlans all the rules setup to make it work. I ended up changing to a flat network, annoying but hey ho. I tried to add a matter device but because the python matter server tried to use my disabled wireless interface the link up failed. But as that was the first thread network my phone saw it grabbed the details and put them into Google play services. Now my phone shows no thread networks or devices in its settings but every time I try to connect a matter device using the phone it chooses the old thread network which it's cached and which I can't see or delete. The only way I have found to stop this is to clear ALL my data in Google play services, I don't want to do that as it affects a lot of things. To get around it I'm using a Bluetooth device on the home assistant for the initial pairing not ideal but it works for now.

[analog sphinx]

are there plans to support scenes for lights?

[reef crane]

The upgrade to 0.4.3 caused one lost node label to be lost (Eve Dimmer). upgrade to 0.4.2 caused zero lost node labels, previous version update caused 4 lost node labels and the one before that caused 3 lost node labels.

[zinc cliff]

Any idea why it says that the thread border router is external when the thread border router im using is the HA ZBT-2

[deep drift]

Because it is not a Matter device. The visualization of matterjs-sever is only showing Matter devices based on the availability of their optional diagnostic information. The only reason the OTBR is showing at all, is due to the diagnostics of Eve Energy showing it as neighbor.

[fringe fjord]

yes, but thats not the topic of this channel 🙂

[fringe fjord]

Ok guys, sorry but I can only repeat: Node labels are ideally stored on the device! So when they get "lost" then the device did not stored them correctly or did not reported them back to us on next read! The server from itself will never write to this attribute from itself.

So to analyze this a bit more structured:

• set a node label
• verify that the label is still there in the BasicInformation cluster attribute data
• do a re-interview of the node
• is the node label still there?
• restart the device and wait till it is online again
• is it still there?
• do another re-interview
• still there?

if it is still there via all these steps, restart the matter server ... and check again

Lets see what comes out there.

[fringe fjord]

Read the Readme. The BR is not a commissioned matter node and the Matter server can only see data from the devices within the network that he has commissioned. So for now this is a known limitation which should be described in the UI and in the readme 🙂
Ideally please tell me where I should add the info additionally because you are not the first one asking. So , where you would have seen it better? 😉

[zinc cliff]

I did see this note and it is a thread border router. Matter is completely new to me and i havent fully understood how it works yet. This is the first time ive actually gotten to see a visualisation of it. i was reading that its quite similar to zigbee how all the nodes mesh together and it uses the 2.4GhZ band

[junior sedge]

You're talking about Thread, not Matter. Matter can run on Thread, but it can also run on WiFi and Ethernet too.

[bright rune]

Should we be able to do OTA updates on the new server. I have an Ikea Bulb which shows an Update is available but trying it throws an error

[fringe fjord]

Yes updates work. What error you get?

[grizzled burrow]

One of two Ikea buttons updated their firmware. The second gets to about 16% (very slowly....about 30-45 minutes) then appears to stop

[bright rune]

Apologies for the late reply: Added the log entry as a file as its too long to paste.

[exotic bone]

That, to me, kind of suggests it did do the update, I think I had a similar issue a week ago or two and I think maybe a reboot or re-pair fixed it

[fringe fjord]

If you like provide debug log. In general updates are problematic the more sleepy they are. How good is battery? Else just retry?

[bright rune]

The update is on a mains powered bulb so power shouldn't be a problem.

The update didn't work as its showing as available again, I will give it another go

Update worked this time, strange 🙂

However there was an error straight after: 2026-03-01 10:35:57.554 INFO Controller~ndHandler Node @1:a basic information changed, sending full node_updated 2026-03-01 10:36:22.793 INFO InteractionMessenger Error sending success after final data report chunk [retransmission-limit-reached] Operation timed out at MessageExchange.#retransmitMessage (/opt/matterjs/node_modules/@matter/protocol/src/protocol/MessageExchange.ts:551:45) at StandardTimer.callback (/opt/matterjs/node_modules/@matter/protocol/src/protocol/MessageExchange.ts:588:18) at Timeout.<anonymous> (/opt/matterjs/node_modules/@matter/general/src/time/StandardTime.ts:100:18) at listOnTimeout (node:internal/timers:588:17) at process.processTimers (node:internal/timers:523:7)

[bright rune]

I'm Adding a series of devices at the moment, not all are staying online after they are attached. Was attempting to add another bulb. I think i was a bit slow entering the code but got the following and the matter server restarting when I tried to restart the process

By restart the process I mean turn the bulb off/on 6 times and try to add it again

[fringe fjord]

hm ... and it shows you an available update for this node? According to this log the node does not support Matter OTA updates at all

Node @1:a has no OTA requestor and is currently not applicable for OTA updates

So please create an issue in the matterjs-server repo and add the log also from checking for the update. thanks.

[fringe fjord]

this is basically an error "t be ignored", we just log a bit verbose currently. basically that means that we did not got an answer to the acknowledgement of the final report data ... so basically we received all data we needed, "all good", but the devcie did not acked that we told "hey we go everythign" 🙂 so this error is basically an "we ignore it because we were done".

[fringe fjord]

duplicate https://github.com/matter-js/matterjs-server/issues/327 ... will be fixed in next version

[fringe fjord]

@bright rune Please send log where I see tha checking ... beause nodes wothout ota requestor should be sorted out before already... so I need to see this log

[bright rune]

I should clarify that by not staying online. I am taking them offline as I am just checking they are working I will eventually deploy them as time goes on.

[bright rune]

Trying to add the log but there is a large amount of data in it which is more than the 65k limit on the ticket.

[fringe fjord]

whatabout attaching as file? 🙂

Honestly: noone wants logs inline as text ... attaching a file is the best 🙂

[bright rune]

Thought I tried that will check, do you want the color code and ANSI stripping out ?

[fringe fjord]

all fine ... a can read it

[bright rune]

Raised, #330

[bright rune]

The log starts from the commissioning of the bulb. Let me know if you need anything else

[undone cypress]

Hello folks. Is there any way to merge a google hub and ikea hub to the same thread network? I've been pulling my hair out all day trying without any success, and it appears that some folks have done it, but I'm stuck.....

[junior sedge]

This is unrelated to the matter server beta; please start a thread in #1284966617670881350

[undone cypress]

@junior sedge thank you!

[craggy scarab]

Has anyone successfully connected the new matter varmblixt from ikea using the js matter server? While I was able to connect many other ikea matter devices, the varmblixt always fails to receive the bluetooth messages. I can provide logs if that helps. Not sure if the issue lies in the server or the device.

[fringe fjord]

add debug log and create an isseu please, i check

[craggy scarab]

I created an issue here and appended the relevant part of the log

https://discord.com/channels/330944238910963714/1478085268706693321

[silk plover]

Would be best to create the issue in the Matterjs-server GitHub and upload the log there.

https://github.com/matter-js/matterjs-server/issues

[craggy scarab]

Sure, makes sense.
https://github.com/matter-js/matterjs-server/issues/333

[bright rune]

@fringe fjord I updated the issue with some comments. If you want the full log putting up then let me know, maybe that will help explain whats gone on.

[tranquil agate]

I'm not sure what happened here, but I lost some sensors from my smart plugs. It appears to happen when a subscription times out. When the matter server reconnects, it removes a cluster (NeoCluster) where those sensors seem to originate.

But it does not happen to all of them. One of my three smart plugs still shows the NeoCluster (Node 1:85). That smart plug shows no subscription timeout.

I cannot evaluate whether this is beta-server related or not. How can I tell? And if so, should I file an issue?

[fringe fjord]

Interesting. I need to check whyt it gets removed, should not happen if all is correct but we also saw strange things like existing clusters that were not in the cluster lists and such ... Please oput the image and logs into an issue in the matterjs-server repo! Thanks Also include the information which server version you use

[tranquil agate]

Done, #337

[hardy juniper]

I’ve noticed the matter-js server seems very disk up sensitive. I operate on a nas / docker swarm combo. When a snapshot happens or snapshot pruning happens matter status to HA gets sluggish.

I solved for now creating a new storage pool using some small SSD drives and moving all docker swarm persistent storage to that pool.

I saw a post about files and folders being cleaned up at a later date. I didn’t know if not maintaining the 73,000+ files in the server directory would change that dynamic.

[fringe fjord]

Yes we know and will address this topic before we go stable with the server. On the list already

[proven marten]

I have an error with matter server, it was working OK, but recently after a reboot it is not starting. I tried everything already, resart, delete, clear, migrate to js beta version, nothing works

Can someone help me to resolve with this issue?

this is my log

Hello there, and thank you for ur support!

[warm bone]

can i run this in docker and use the matter integration? The documents only talk about add-ons. Also, I tried to import my python matter info from ~/matter-data amd it did not import. I only have one device. How long should it take?

[undone hull]

yes, I use it with docker (actually podman)

[warm bone]

sorry, that quesrion was more anout using the matter integration over the matterjs addon. I also run home assistant in docker so i do not have access to add-ons

[compact root]

-# *apps

[hardy juniper]

Yes, deploy as docker service, go to ha and when you configure matter point to your matter server service ip address and port

[fringe fjord]

This log is not showing the new matterjs based server , so this is the wrong channel ... As hint: somethings (for me as non python dev!) seems borked with the python in the addon. maybe try to reinstall it? But as said please use the openprotocols channel and create an thread there.
I also do not see any try to install the matterjs server, so seems there is no Beta flag enabled

[fringe fjord]

Sooo friends of the new Beta Matter Server ... After being stable and proving a lot of things the last weeks ... The new version is changing again some details in the Re-/Connection logic to devices ... Ideally we expect it to be even better than the last version, but there could be things to tweak in the details ...

So 0.5.0 (Changelog https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#050-2026-03-05) is there... please check and especially have a look that/how fast your devices come online again after the restart.

For the "many devices guys" I would loooove again to get your logs from the start of that new version to see and be able to compare it. We have many knobs to tweak settings depending on the results and tested around a lot, but noone of us devs have a system with that many devices than you 🙂

Please also report issues (for sure).

[deep drift]

Update was smooth. Reconnect speed is comparable to former versions (I only have 25 devices). The new visu search function is awesome! Great work!

Nevertheless, there is a new error in my log:
2026-03-05 10:35:23.909 WARN AttributeDataDecoder Error decoding attribute 1/0x130afc01/0x130a0014: Unexpected type 10, was expecting 4.

[fringe fjord]

hhmmmmmmmm .. now it becomes starnge ... another user reported exactly the other way around ... which firm ware ou hve? Most current?

[deep drift]

All my firmwares are the latest official releases. No dcl-testnet versions. Which device is it?

[fringe fjord]

I found the reasons ... the other user report was wrong ... will revert in next update

[fringe fjord]

As Info: There was an issue reported with Command Batching and that the 0.5.0 sometimes can try to combine invokes for devices that do not support this - only happens in rare cases. A 0.5.1 which fixes this will be planned for later today

[fringe fjord]

Sooooo clear. I prepare release and GitHub has issues. Sooo all a bit delayed.

[compact root]

[fringe fjord]

Sooo guys ... 0.5.1 which fixes the startip issue with uncommissioned old peers (including cleanup of them) and also the batch invoke is fixed ... https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#051-2026-03-05

[hardy juniper]

I have 137 node matter env. Just started 5.0.1. Captured the log right after the web ui came up. Is that what you are wanting to see examples of? Took almost exactly 60 seconds to start.

[fringe fjord]

Cool. Interesting would be log of the start till nodes went online. Just drop me the log here as DM

[grand kite]

getting some nodes (a small number) not comming up on boot. sometimes when I try to re-interview nothing happens, but I also just got this stack trace:

after the exception, a new session wass established with the node and it came back online

this is on 0.5.1

[grand kite]

somewhat bizarely when I try to manually reinterview nodes that won't come up, the WebSocket always resets:

2026-03-05 18:52:56.145 INFO ClientInteraction Read « @1:2a•a276⇵a2e9 27↔27 attributes: 804 events: 0
2026-03-05 18:52:56.509 INFO WebSocketC~erHandler [d] WebSocket connection closed
2026-03-05 18:52:56.956 INFO WebSocketC~erHandler [e] WebSocket connection established
2026-03-05 18:52:56.971 INFO WebSocketC~erHandler WebSocket server start_listening
2026-03-05 18:52:56.982 INFO WebSocketC~erHandler WebSocket server start_listening. Returned 72 nodes

[grand kite]

here is a full log for a restart in which 2 nodes haven't come back online

Node 36 & Node 41

Manually power cycling them doesn't bring them back.

[hardy juniper]

Log sent . I also had 2 nodes not come online. Ended up having to power cycle the switch. This isn’t really new. Since it ran for 5 hours, there’s some errors in there you may want to check out after it was up

[hardy juniper]

When I pull the air gap this is all the activity I get for that node.

[fringe fjord]

For issues when node do not come back would be great to get some logs in debug level - ideally from. Start but maybe also ok for the beginning when dynamically switched to debug for some time (min 10 mins) using the matter server dashboard.

[lusty linden]

This is just to say that the new Matter server has been rock stable. Thank you @fringe fjord !

[fringe fjord]

Hey guys, thanks for the logs and details, I will work through them today.

I also have a friendly reminder: Coming Monday (9th march 2026) at 7pm CET the next OHF Matter Communitiy sync (Google Meet) will take place ... so join if you want to stay on top of the development around the OHF Matter server, Matter in HA and matter.js ... Hope to see you all there!
https://meet.google.com/uch-rsxw-txh

[fringe fjord]

Ok, so in the first log you sent me as DM this node got connected. In the other logs here from later the first one I see that we sent the read ... but then not more ...

The second vry short log is interesting ... because it shows a device restart and the devcie that established a session and wants to re-esgablish subscriptions that we do not know .. that can happen ... so to see more I would need more log

[fringe fjord]

Ok node @1:2c ... 44

• trying to connect and got a "state issue" and node tells busy for 69s ... so we delay more tries, got a session at 19:13:25
• was online 19:17:47

Node 36:

• Session created 19:11:10
• sending read 19:11:48.720 .. then stalled somehow

Node 41:

• stalled on subscribe

[fringe fjord]

@grand kite could it be possible you try again and get a debug log?

[hardy juniper]

I restarted the docker service twice and both times all the nodes came back online. I’ll see if one falls off and try to reset it and capture the debug log and/or restart service later today.

Would be handy if there was a way to restart the matter server without having to stop and start the docker service like how you can restart HA in dev tools.

[fringe fjord]

all fine, if sall come up tghen all fine

In the logs I basically see that something can happen ahd n the read or the subscribe "hangs" somehow and so never get finished but also no other part catches up here ... but lets see ... I still look into having it get better and also have an idea on how to try to give you an optio to recover without restart
I am very sure e can also fix the root cause but for that I need to understand it 🙂

[grand kite]

Update: about half my nodes fell offline overnight. Haven't seen this with the beta since I started using it. Definitely some sort of regression.

[fringe fjord]

just grab the latest 10k of loglines and send over please

[grand kite]

Note that in addition to getting the new matter version, I did also turn on ipv6 in the docker instance that HA apps run inside. I don't think this should be a cause, but it was another change I made at the time. ha docker options --enable-ipv6=true

I belive this is standard on new installs as of 6ish months ago.

[fringe fjord]

I check it ... interesting .. but I think I already found things

[fringe fjord]

@grand kite Did you filter this log anyhow or is this really the full plain log as it was logged?

[manic lichen]

https://discord.gg/home-assistant?event=1479499800914165951

[grand kite]

This is just the log entries from when the network lost all its nodes. But it happened again so I can give you a full log now. Give me one sec.

[fringe fjord]

Ok checking ...

[craggy veldt]

I was asked to post my log files here (i hope those are those files)

I was having issues with the Matter server and got suggested to try the beta knowing the update would take roughly half an hour. Tho it seems now that my matter server isn't properly starting cause when i press "start" and come back 20 sec later i can press "start" again. Suggesting it doesn't start properly.

Did something maybe go wrong when updating or is my network the issue?

this is what my device show (Eve Thermo Matter over Thread)

[fringe fjord]

your log is only this one line? or was that a broken try to drop the log? 🙂

[craggy veldt]

uhhh

probably a brokem try

should i choose to download 10000 lines?

my HA is only a week old anyway

better? idk what I'm doing here.

Long story short: was using homekit for years and wanted to get into HA. Was having issues with my matter thread devices on HA but not in HomeKit. heard about thread issues and that HA is trying to fix the matter server and tried out the beta. And I think the matter server never came up again properly

[silk plover]

Click the circled download button choose 5000 or 10,000 lines in the pop-up then the log should download to your desktop or phone. Upload that file here.

[fringe fjord]

still not the file you wanted to attach

[craggy veldt]

okay i swear i just did that with 10000 lines

instantly exported

to here

did i do something wrong?

i...i'm sorry. I must have earlier not safe the change to "debug" log

This should be it

[silk plover]

That looks like what apollon can use

[fringe fjord]

@craggy veldt did you just created this also as GitHub issue because there is anopther such case. Please make a backup of your matter app and send me as DM this backup file

[craggy veldt]

no i did not

how do i do that?

(make a backup of matter app)

[fringe fjord]

Versuion 0.5.2 will be reverted!!

[silk plover]

What will be reverted in 0.5.2?

[fringe fjord]

I just released a 0.5.2 but reverted it to 0.5.1 on npm again, so noone will get it ... i updated my installation after release and somethings happened ... sooooo need to check that first

I guess noone has updated so fast, if someone got the 0.5.2 ... then restart!

[fringe fjord]

Can anyone help @craggy veldt on how todo a app-only backup ... I can also not find it (uups)

[craggy veldt]

found it! Now to figure out how to share only the one file

(only could do full backups?

[elfin hull]

@craggy veldt App-only backup: Settings -> System -> Backups, + Backup now, Manual backup, deselect History, then deselect Home Assistant settings (which is insensitive until History has been deselected), then change the Apps dropdown to Custom, select Matter Server (or any other app if you need something else), Next, give it a name and choose locations and you're good to go

[undone hull]

My docker container was updated, and as I use latest, it stays like it. Could you please release new version?

[fringe fjord]

Just restart. Will downgrade

At least did on my system

[undone hull]

I use container, not add-on

systemctl --user start podman-auto-update leaves it like it is

http://localhost:5580/ does not even open

I use ghcr.io/matter-js/matterjs-server:latest

[fringe fjord]

Then use 0.5.1 and not latest please.

A fix comes tomorrow. Too late that I could support it.

[undone hull]

ok, but you could just push 0.5.1 as latest

[fringe fjord]

I am already in my bed. Sorry. Tomorrow. I did for npm but forgot the docker. Sorry. Promise better if that hopefully never happens again 😉

[undone hull]

ok, no problem, downgraded, thanks!

[elfin hull]

I've been running the matterjs-server since it was exposed behind the beta flag in the Home Assistant Matter Server Add-on App (and since I haven't had a chance to say this, thank you so much for this project! It has really improved my experience with Matter and Home Assistant). I believe that I'm seeing a noticeably higher latency in receiving updates from sensors since upgrading to 0.5.0/0.5.1. My easiest case to observe is a Myggbett or Aqara door/window sensor. I have a few simple automations for the case "door opens -> turn on light", and since the upgrade yesterday it seems like things take just a bit longer, nothing extreme, just long enough for my brain to trip over.

I have been able to compare doors where the controlled light is a Matter/Thread bulb against another door where the controlled light is a Z-Wave switch, and they both seem delayed compared to 0.4.3. The only other change recently was upgrading to HA 2026.3, but no change to my Thread network (OTBR App in HA is my only border router). The lag persisted past restarting HA and even rebooting the Proxmox VE host.

Trying to just watch sensor entity in HA while opening the door, the delay appears to be before the automation triggers. I see the results of the automation at the same moment I see the entity update state in HA.

Has anyone else seen anything like this? Is there any way for me to temporarily roll back to matterjs-server 0.4.3 to see if this behavioral change is real or just in my head? The Matter Server App pulls the latest matterjs-server release on startup, so my backups aren't useful here as far as I know.

Cheers, and again thank you @fringe fjord for all of this work!

[silk plover]

Apollon should have a 0.5.3 or so out tomorrow to fix some issues with recent optimizations in 0.5.0 & 0.5.1, so maybe try that update and see if your problem persist. I don’t know of a way to roll back the install from the HA Matter server app either.

[grizzled burrow]

On the 0.5.x I've observed "random" Inovelli dimmers being permanently offline until I do a full HAOS VM reboot. For example, yesterday two white switches were marked as "offline" no matter what I did to the switch (power cycle) or a matter server restart. I did a full HAOS VM reboot, and those two came online, but then a different one is now stuck "offline". This never happened with releases prior to 0.5.x.

[flint grove]

I'm having the same problem with my Inovelli switches and my Eve Motion Blinds, they go off line for several hours and they come back for some time.

[hardy juniper]

I am seeing same behavior. Sending a big debug log via DM. 1 of the nodes is an inovelli.

[grand kite]

Ya, I've seen the same. Have sent @fringe fjord several logs. Hopefully helpful. Perhaps 0.5.3 will fix it

[flint grove]

But the strange thing is that I have a similar setup in my main house and it is working perfectly.

[silk plover]

Same, reported to Apollon soon after 0.5.1. with DM startup logs. he responded he found something to fix. Think 0.5.2 included that, but was had issues and was reverted immediately. We should gat a new version asap once he gets a chance.

[deep drift]

Yes, with 0.5.1 I also experience a noticeably increased latency with my BILRESA buttons.

[fringe fjord]

I am preparing the 0.5.3 just now (will post when available) .. please try there and if the issue is still there polease provide a debug log ideally from such cases. when "only" events are delayed then it should be enoight to enable debug temporariely via the matterjs server dashbord (settings icon upper right) and do some actions to show the delays, notre down times please so that I can compare ... then create issue in the matterjs-server repo and I have a look. But maybe we are lucky and it is gone in 0.5.3 again but the stuff there was mainly other topics but lets see

[fringe fjord]

Yes, as already written 0.5.x changes the waa to connect and reconnect to devices "dramatically" ... and 0.5.0&/1 had some issues in details. cominng 0.5.3 should fix those ... at least then we can see whats left

[fringe fjord]

The issues we had were triggered by devices that stopped responding to messages in the middle of a read or subscribe where potentially multiple packages are transferred ... so it really also was about the devices and the network quality/issues if it hit you or not. but 0.5.3 will fix all stuf I found after reviewing logs yesterday

🚩 Soo hey guys ... 0.5.3 is out to replace the borked 0.5.2 (https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#053-2026-03-07) ... Please tell me if your connections to the devices are now reliable again

[tranquil agate]

Wow, thanks for the quick follow up, again. Even going on during the weekends is real dedication! 🙏

[craggy veldt]

could someone please help me. How do I update my Matter Server? I could only find the 2026.3.1 update for HA

[flint grove]

Thanks Apollo77 , can some guide me to update it by the git hub, I am new in HA.

[manic lichen]

Are you using the matter.js testing version or the regular Matter server?

[craggy veldt]

js testing. Use latest beta version is on

[flint grove]

same here

[flint grove]

yo have HAOS?

[sick pond]

matter-server/0.4.3 (matter.js/0.16.11-alpha.0-20260225-033797f3c) My install is fresh on an HA Green, with both the ZWA-2 & ZBT-2 usb gadgets setup... I don't have a listing for a testing version of matter, so I'm guessing that a url needs to be added ... not sure if I'm that game just yet! but ask me in a few hours and I bet I'll be keen to try!!

[craggy veldt]

yes

[flint grove]

same here , But not experienced to update with github

[craggy veldt]

same :/

[hardy juniper]

5.3 still has devices going on and offline a lot. They are recovering, but they keep showing up in ha as unknown then back …

One of them is a sonoff device on WiFi. I know it isn’t disconnecting cause I can see its log on my router. Will send files

[fringe fjord]

To update to 0.5.3 just restart the matter server app/addon. then it pulls new version.

[fringe fjord]

Checked the WiFi details in the matter server dashboard? There is a graph for it. And this uses the data from the device. We already had users that were surprised because the routers showed complete other (better) numbers.

But you send me log. I will check

[flint grove]

already restarted , but still some devices thread devices ofline. should I give them some time?

[warm bone]

I think I borked something trying to get this running. Any ideas?

matterjs-server | 2026-03-07 12:07:59.945 INFO MatterServer Command line: (no arguments) matterjs-server | 2026-03-07 12:07:59.955 INFO ConfigStorage Storage location: /data (Directory) matterjs-server | Error: EACCES: permission denied, mkdir '/data/config' matterjs-server | at async mkdir (node:internal/fs/promises:855:10) matterjs-server | at async StorageBackendDisk.initialize (/app/node_modules/@matter/nodejs/src/storage/fs/StorageBackendDisk.ts:65:9) matterjs-server | at async StorageService.open (/app/node_modules/@matter/general/src/storage/StorageService.ts:154:9) matterjs-server | at async ConfigStorage.open (/app/node_modules/@matter-server/ws-controller/src/server/ConfigStorage.ts:65:25) matterjs-server | at async Function.create (/app/node_modules/@matter-server/ws-controller/src/server/ConfigStorage.ts:44:9) matterjs-server | at async start (/app/node_modules/matter-server/src/MatterServer.ts:151:14) { matterjs-server | errno: -13, matterjs-server | code: 'EACCES', matterjs-server | syscall: 'mkdir', matterjs-server | path: '/data/config' matterjs-server | } Container matterjs-server Stopped matterjs-server exited with code 0

[flint grove]

Did you updated ? How did it go ?

[craggy veldt]

yes i did. It updates itself. I'm sure yours is too on 0.5.3.

I added the Matter Server to my side tab and opening this i saw a page i never found before and in there i could see it's on the latest version

my issues also still persist

[flint grove]

Me too, still divices off line

[fringe fjord]

Without having more logs but that seems like aa subscription timeout ... so peer stopped responding ... we should try to reconnect ... but likely such things are more device or network issues. if you like grab a log and send me and I tell you wahat I see

[fringe fjord]

you use docker directly? Please see changelog!!! There was a change for the "pure docker" usage some releases ago, maybe this is the case for you?

🚩 There is currently one issue with the 0.5.x line which was not there before but this only affects "new users" that just switch from the Python server now. These should ideally wait for 0.5.4 I will fix this issue there. There is a small change in initialization order which leads to issues when you just fresh migrate. All users that migrated before 0.5.0 are not affected and also 0.5 works ...
Thats basically also @craggy veldt 's issue

[fringe fjord]

No, not "me too" because you have devices 🙂 So when they are offline as said above, try to restart them or send me a log and I check what I can see.

[flint grove]

it's ok like these? Sorry I'm new even in Discord

You'rre right sory , jajajaj

[silk plover]

Change the log level of Matter server to debug.

Then Air gap those disconnected Inovelli switches for 10-15 seconds and power on and off the eve. See if they come online. if not send the debug log for Apollon.

[fringe fjord]

Ok whioch nodes are offline/relevant in the log?
@1:22 for examply is not responding to the read and so retried ... and this constantly ... so retry time went up to 1h 🙂
@1:32, @1:7a same timeout on re-subscribe

... so all I see are network or device side issues

Maybe I could see more in a debig log (thanks @silk plover ) ... but i general seems you have con ectivity or network isues

[flint grove]

ok, changing to debug

I have to restart MAtter server after changing to debug?

[fringe fjord]

if you changed just in the dashboard that should have immediate effect ... and you can also turn back. mno restart needed

look at the logs. when it worked you have now also "DEBUG entries

[flint grove]

yes I have Debug entries

[silk plover]

if you see 'matter-server/0.4.3 (matter.js/0.16.11-alpha.0-20260225-033797f3c) ' then you are already on the beta/testing version of matter server. Just go to the Matter Server 8.2.2 app in the HAos/settings/apps/matter server/info and hit restart. Matter server will automatically update to latest 0.5.3. no need for a URL or other change.

[grand kite]

all of my nodes successfully came online with 0.5.3 🙂

[fringe fjord]

I prepare 0.5.4 with a fix for "all new beta migrators" and also I found that one "we need multiple reads because devices are flaky"-optimizations is not optimal working, soalso this should make things better there.

[grizzled burrow]

with 0.5.3 now I have an Eve energy offline, but the Inovelli white switches are back

[fringe fjord]

try to repower ... else as usual .. loooggsssss

[flint grove]

So we restart the server?

[fringe fjord]

??

[flint grove]

The Matter server

[hollow cloak]

@elfin hull I converted your message into a file since it's above 15 lines :+1:

[fringe fjord]

if you mean for a 0.5.4 then wait until I post that I released oit ... i jhust said "I work on it 🙂

[flint grove]

Ahh ok , thanks

[fringe fjord]

Ok log states that the server receved the event 08:23:30.983 and delivered it to HA at 08:23:31.502 ... and likely because minimum subscription interval we use there shpuld be a 1s delay ... so "delayed" in which regard? Server wise I can not see any delay from device triggering the info and we giving it to ha ( ok maybe 600ms) ... or do you mean these 500ms as the delay?

@elfin hull Please create an github issue in the matterjs-server rpo for this .. I think I have an idea how to optimize this

[elfin hull]

600ms matches my perception. It feels like it's consistent and longer than I experienced with 0.4.3 or any of the previous releases, but I'm not sure.

[elfin hull]

Will do!

[fringe fjord]

should be no change compared to 0.4.3 but can still nbe optimized I guess

[grizzled burrow]

0.5.3 is a total disaster....numerous devices falling offline. Debug logs sent via DM.

[silk plover]

All 92 of my nodes online for 1-2 hrs and no issues here with 0.5.3, fixed my issues with earlier versions. About to commission a couple more that came in. Will report back if issues.

[hardy juniper]

has anyone been able to set the LOG_FILE env variable for docker? I have set it (tried as directory and as a file, file fails) and server starts but nothing is logged to the directory, still going to console in docker.

`environment:

• TZ=America/Chicago
• OTA_PROVIDER_DIR=/data/updates/
• PRODUCTION_MODE=true
• LOG_LEVEL=debug
• USERNS=keep-id
• LOG_FILE=/data/logfile/`

the directory is mapped correctly, I can go in the container and write out to that directory and see it on my nas.

All 138 of my nodes come up on startup. chasing down some falling off and back on, but they all come up on startup.

[deep drift]

My 7x Eve Energy are all online and stable. 🤷‍♂️

[fringe fjord]

I see many subscription timeouts ... and also see these eve pollings for energy ... no firmware update available or are these those US devices?

[deep drift]

I agree. 0.5.3 still feels a bit laggy compared to 0.4.3.

[fringe fjord]

already on an optimization ... but should have had been the same before

[deep drift]

I also have subscription timeouts with Eve Energy. Those errors haven't been there before:

2026-03-07 17:40:28.835 INFO ClientSubscription Subscription a19d8482 to peer @1:9 timed out after 1m 23s
2026-03-07 17:40:28.835 INFO Controller~ndHandler Node @1:9 availability changed: true -> false (state: Connected -> WaitingForDeviceDiscovery)
2026-03-07 17:40:28.835 INFO WebSocketC~erHandler [0] Node @1:9 availability changed to false
2026-03-07 17:40:28.835 ERROR ClientSubscription Replacing subscription to @1:9 due to timeout
2026-03-07 17:40:28.836 INFO ClientInteraction Read » @1:9•a2f7⇵f749 probe
2026-03-07 17:40:28.934 INFO ClientInteraction Read « @1:9•a2f7⇵f749 (empty)
2026-03-07 17:40:28.935 INFO ClientInteraction Subscribe » @1:9•a2f7⇵f74a min: 1s max: 1m attributes: 1 events: 1
2026-03-07 17:40:29.551 INFO ClientInteraction Subscription successful « @1:9•a2f7⇵f74a 2↔2 id: cb3fbfab interval: 1m timeout: 1m 23s
2026-03-07 17:40:29.616 INFO Controller~ndHandler Node @1:9 availability changed: false -> true (state: WaitingForDeviceDiscovery -> Connected)
2026-03-07 17:40:29.616 INFO WebSocketC~erHandler [0] Node @1:9 availability changed to true

[fringe fjord]

I have a look. We could ait a bit longer on subscriptions but usually ... yeahhh

[fringe fjord]

Yes in fact this was a change introduced by 0.5.0 line, so yes. "fix" will come in 0.5.4

[deep drift]

Perfect. Thanks!

[grizzled burrow]

US devices

[hardy juniper]

Rebooted all my thrad border routers. Normally that caused a 2 hour storm with python matter server. All nodes came back up. A few took about 15 min to back out of it's old config, but they all came up. I had noticed this before, but there is 1 node which is my hue hub with matter integration turned on, that get's a continuous warning / error. Log below and picture of the device.

[fringe fjord]

Info: getting late,,so 0.5.4 likely tomorrow

[fringe fjord]

lol ... yeahh ... can you send me the version details of that hub ... i contact hue guys. in fact thats full valid. the endpoint 4 can not have a basic information cluster

[hardy juniper]

Software version is 1.75.1975170000 hardware model is 3241312018A. if you wanted something else let me know

[fringe fjord]

So yeh the warning iis really a devcue bug in my eyes, so valid 🙂

[warm bone]

I basically used the python matter server data folder for my matterjs docker compose volume and that’s when the issue seemed to arise

[fringe fjord]

thats also the supposed way, so not your fault ... there was a change and I need to adjust something. Fil will come in 0.5.4 tomorrow

[warm bone]

Thank you

[craggy veldt]

oh my you're fast

[warm bone]

it looks like the settings are the same for the matter integration in hass also? So no need to edit that? Should just show devices and info like the python matter server?

[fringe fjord]

exactly ... new server is supposed to be a drop-in replacement

[hardy juniper]

LOG_FILE is not being picked up as environment variable. Can't see it in list of ENV variables in container, it isn't kicking out any info messages when it configures the log file location (success or fail). Would like to redirect to a file when using debug, can you validate the LOG_FILE is being picked up the build

[fringe fjord]

if you have issue screate an issue in the matterjs repo please and I will check next week ... then it does not get lost

[vapid timber]

Hm. mine says "No update available" and is stuck on matter v1.3.0, firmware v1.8.5

[silk plover]

Have to enable test-net, but be aware this is not a stable release channel and ikea firmware has bricked at least one users bilresa when it did not complete.

[vapid timber]

thank you. I'll wait for it to come down the normal channel

I'm definitely struggling with https://github.com/matter-js/matterjs-server/issues/187

[fringe fjord]

So Matter friends, 0.5.4 is just released and should improve the experience and ideally also behavior for flaky devices on startup issues. It also should alow new migrations when coming fresh from the Python server.
https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#054-2026-03-08

[warm bone]

I am still getting the same error messages and container will not start

[fringe fjord]

please provide me with a debug level log of the start ... did you sent me your backup?

[fringe fjord]

Ok thats not the startup issue ... i think I already wrote above ... please check Docker readme and instructions, especially when upgrading from earlier versions there was a breakign docker change from 0.x to 0.4! ... especially about permissions. The docs also contain commands to fix permissions. please try this first

[warm bone]

where is rhe change log?

[fringe fjord]

Read my release announcement above? 🙂

likely relevant for you is
https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#040-2026-02-19

BREAKING: (schildbach) Only for Docker/Podman users: run server as an unprivileged user. Use chown -R 1000:1000 /path-to-data-volume once to migrate permissions! If you're using rootless Podman or Docker and user namespaces, UIDs and GIDs will be remapped to a different value and the previous command needs to be adapted accordingly. If you want to avoid that, use the --userns=keep-id option when running the container.

[warm bone]

adding

environment: - userns=keep-id

didnt help

[fiery dune]

Few quick questions;

1. Is it okay to start fresh with the matter-js or a migration from py is recommended?
2. I run a containerised version of HA. matter-js in a docker is discourged, I read. What is the recommendation for non-tech users using HA container?

[warm bone]

ok, chown -R 1000:1000 /path-to-data-volumeworked

server,is up and running

Anything I need to worry anout with changing those permissions?

[fringe fjord]

However you like, the server is a drop in repelacement, so just changing the contaioner you use and all should work the same and migrate the data on first start. Also switching back to the "python matter-server is possible. So if you start fresh or migrate is up to you.
The Python server will stay on the current version and will not be changed anymore, but. is there as "the baseline".

I do not know what you mean with "matter-js in a docker is discourged" .. the matterjs-server is basically the same as before ... there is just a special case when you want to enable BLE inside the container then unprivileged does not work, but the docs contain infos ion this.

I guess "Non-tech users" should use HAOS and just use the Matter-Server App/Addon as is ... any own dockerized setup of HA is never "non-tech" 🙂 And for App/Addon users they do not need to do anything but enable the Beta flag in the addon/App and you will use the Beta server.

[fringe fjord]

Honestly ... Should not but I am not really a docker pro ... so that likely can only be answered to a full detail by docker experienced users.

[warm bone]

@fiery dune I run docker for both HAss and Matter. I just copied my data folder from the python matter server into the matterjs server. changed the volume and followed rhe docker docs for matterjs-server I had to run that command a couple posts up to get everything working. stopped the python matter server, started the matterjs server and didnt have to do a thing inside hass

[elfin hull]

Happy to report that with 0.5.4 I don't see any of the extra latency, seems just as snappy as 0.4.3 and before was. Thanks!

[fiery dune]

Thanks for the promt replies!
Since I am starting fresh with thread and matter, I will then jump straight into the js container.
Post that, will I start adding whatever few devices that I got.
Thanks again.

[warm bone]

Yea. I currently only have 2 thread devices, both the same and only one matter device

[flint grove]

Apollo77, thanks for your support and dedication on the proyect. I already updete it and they came back slowly .

All my divices are back on line again Is it normal that the matter/WIFI come in faster and thread is more delayed? Because the strange thing that in my main home everey thing is working perfectly.

Well now both are working great, Thanks again .

[fringe fjord]

The very first start needs more time in general because we migrate data from the python config ... this is a one time thing.
But yes in general thread is limited currently by a queue to not overload the thread network, so yes thread will be slower. When it is not stable in general again with the new networking wedid in 0.5 then we will start optimizing more ... if you like just grab 10k log lines and send me as DM here so that I can have a look how the startzup behaved and such ...

[craggy veldt]

I can also report improvements! My devices so far don't become unavailable and after a retsrat of HA/Matter server they are all back online within 30 seconds. Only issues are still with connecting new devices (pairing mode from HomeKit). Either it works quick (like 10-20 seconds) or it doesn't work at all and I found it to help restrating HA as a whole. Then it usually works.

But thank you it's useable now for me!

[silk plover]

Matter over Wifi always come back online first for me as well, so expected. Thread slower to all get back online, once all online I don't notice any difference in commands or automations using either.

[fringe fjord]

And first start also needs to wait for all devices to announce via MDNS and such so also this might delay a bit ... subsequent starts should be even faster. SOó if you like grab the log from that initial start, then do a restart and grab that log again. I can then analyze it and show details

[fringe fjord]

thats strange ... which error you get when it does not work?

[fringe fjord]

And also as for the others would be cool if you could drop me a lot of that 0.5.4 start ... to check things ...

[craggy veldt]

Either no info at all or "failed to add device"

[craggy veldt]

will do

[fringe fjord]

There should be anything - at least in the log

[craggy veldt]

Also just noticed...one device became unavailable. let's see if it persists. All other devices all stable (aroudn 15)

[hardy juniper]

Everything started back up perfectly from .5.3 to .5.4. All nodes came back online and control snappy.

[flint grove]

Same here, everything running smoothly. 👍

[fringe fjord]

..That was the hope 🙂 Glad to hear

[grizzled burrow]

0.5.4 appears to be good so far.,...didn't have any straggler offline devices after a restart..out of 73

[deep drift]

0.5.4 - smooth and snappy for me. Many thanks, @fringe fjord!

[warm bone]

Did you ever get this figured out? I can’t even get it to generate the folder

[fringe fjord]

Ever tried providing a path that exists folder wise? I bed to look in code if we would create the folder when missing or assume it exists tbh

Yeah code assume path exists

[hardy juniper]

I created the folder inside the volume I have mapped to data and then pointed log to that. But it doesn’t even seem to try and write a file. I’ve logged a bug on git

[warm bone]

same experience here

[fringe fjord]

please ensure to include the info which env var you set how or which parameter you used

[fringe fjord]

Ok I got one issue and the log contains this line

Failed to set up file logging: Error: EISDIR: illegal operation on a directory, open '/data/logfile/'\u001b[0m\n","stream":"stderr","time":"2026-03-08T14:02:54.291355004Z"}

and yeahh ... you should define a log FILE name ... not a directory

So for me it clearly tries to use it nd also set it up but you configured it wrong

[hardy juniper]

Ok, I updated the issue log to say it is a documentation error. The documentation says.

LOG_FILE Log file path (none) Any valid file path

And file path is defined by several sources as:

A file path is a string of characters that represents the unique location of a file or directory (folder) within a computer's file system.

Would be better if it was actually a directory and then the system created a new file with a timestamp, otherwise when trying to log outside of the console, it's just gonna make 1 huge debug file of all system starts.

[undone hull]

How to enable test-net in docker version of matter server?

[hardy juniper]

Environment variable on startup.

https://github.com/matter-js/matterjs-server/blob/main/docs/docker.md

[fringe fjord]

I will updte docs ... more I will comment onnthe GitHub issue

[hardy juniper]

Does matter binding helper work with matter-js? I see an issue to integrate UI in their git repo. I tried but get an acl error.

[fringe fjord]

lets say it that way ... I do not know ... but the error handling of the python server when it came to "Write ACLs" was incomplete and when writung 10 reconrds (example) it only checked if the first one was written successfully leaving you in the feeling that it worked ... the matter,js server checks now if the write is REALLY successfull for all data and returns an error if not ... and the cases I saw for an error were all errornous because too many results where written. The Beta server also returns the correct error code and the dashboard of the matter server also shows the correct error message depending on the error.
It seems noone everyreally was able to reach the developer of that helper and he also never responded to the question if he want to integrate it into the new server now that we have stabilized ... soooo .....

TL;DR: If used right then it will work. If invalid write messages are sent you get an error.

[fringe fjord]

So guys, today no update of the Matterjs server, but the HA Addon was updated to 8.3.0 but mainly to
enhance some advanced settings... more about them when needed 🙂

And remember ... Matter community sync in 4h from now 🙂

[compact root]

It's in the Events section if you missed the link before! https://discord.com/events/330944238910963714/1479499800914165951

[left rose]

Does the 8.3.0 update get announced as a regular update (even if currently in beta mode), or does it update to 8.3.0 on next server restart?

[silk plover]

Announced thru normal HA update mechanism for me this morning.

[left rose]

Thanks for clarifying. It just appeared for me a few minutes ago as well.

[craggy veldt]

what's Matter community sync?

[craggy veldt]

Is it worth installing? will it make a difference for beta users?

[silk plover]

HA devs; matter, thread in general and Matterjs Server, discuss where they are and where they are going next in video call open to interested users.

[silk plover]

No difference for beta users today that I can tell, assuming apollon has made some changes in the Matter HA app that will allow him to expose additional options/settings in Matterjs Server before long.

[flint grove]

@craggy veldt did you solve your devices?

[deep drift]

I updated two Ikea Bilresa to the latest testnet-dcl firmware yesterday. I needed 2-3 tries each, because the update procedure broke off. Never happened before with all the official firmware updates of Eve (Eve Weather aside) and Ikea devices. Is there something different about testnet-dcl firmwares?

[craggy veldt]

yes! Everything is running solid now. Tho i did set up my HA again from scratch (i had nothing to lose cause i had just started anyway).

I then instantly opted into the beta and could mostly without issues add my devices and they were all solid with occasional but rare unavailables.

I'm happy and confident enough to rely on my automations.

[deep drift]

Currently I am fighting with the Myggspray update to testnet-dcl 1.1.2 firmware. This one is especially nasty. After breaking off the second time HA is stuck in an infinite update loop. Restarting HA only won't help, I need to restart the whole HAOS (Proxmox VM) to get out of it again. I am logging on debug level and creating an issue.

[fringe fjord]

that 🙂 #1346944149303066711 message

[fringe fjord]

will add some options to better tweak settings for future testing plans ... but not mission critical

[fringe fjord]

could by battery level originated, network originated because remember it is a battery powered device ... so in general saw that some times already ... but yeah if it worked then it worked

[deep drift]

I now moved the Myggspray 10cm next to my ZBT-1 dongle to rule out signal quality. And yes, it worked.

[deep drift]

The devices have a medium signal quality in the -70s dBm. Shouldn't this be sufficient for a reliable update procedure? I understand that an update would take more time due to low bandwidth, stall, retries. But it shouldn't break off.

[hollow cloak]

@deep drift I converted your message into a file since it's above 15 lines :+1:

[fringe fjord]

yes ... ok? 🙂 SOrry but with this I can not dssay anything ... no context

[deep drift]

Sorry! These are the two lines when the Myggspray OTA update broke off.

Could it be that the OTA download used too much bandwidth, so that the device lost its session?

[fringe fjord]

I need log line from before to answer that ...

[deep drift]

PM or issue?

[fringe fjord]

lets do PM for now because in fact that can anytime happen, so not really an issue "per se"

[fringe fjord]

So also for the channel. Yes the device stopped responding during OTA, and @deep drift was not patient enough to have the device time out officially on the transfer to report this back ... sooo thats why HA felt "stuck" in a loop

[flint grove]

@fringe fjord still having issues with my devices , any clue What should I do?

[modest hedge]

well i switched over to this today and am getting further in the pairing process

its taking forever at checking network connectivity to thread network

[silk plover]

Beta testing OTA firmware before release, with latest matterjs server, is there a way to delete an OTA that I manually uploaded and matter server has already moved into its hidden storage location? I see mention of:

@matter/nodejs-shell
Adjustment: ota add stores files as "local" mode; ota list, ota delete, and ota copy support "local" mode filter

in the matter.js 0.16.10 changelog but unclear if that is internal only or somehow user accessible.

[fringe fjord]

basically stored in the data/ota directory as vendorid.productid.version.*

[silk plover]

Sorry for my ignorance, but using a file Browser from the root of HAOS what might that file path be? I see a /data, but no other folders there. assuming HAOS container storage is buried somewhere else.

[flint grove]

Can somebody guide me , Im having devices coming and going off line. Aparrently is not the matter server .

I have Thread border router with Apple tv ,a 2 home pod minis, I have the 2.4 ghz in channel 1 , Zigbee 11 and thread in 25

[compact root]

This is more likely an #1284966617670881350 or #1284966582375813201 support issue. This channel is specifically for the beta testing of the new Matter Server. 😄

[flint grove]

Thanks Missy

[silk plover]

answering my own question for any future users looking to list or remove Manually added OTA.

1. used the Advanced SSH & Web Terminal app with protection mode off
2. docker exec -it addon_core_matter_server bash (to enter the Matter server Docker container CLI)
3. ls -lh /data/ota (to list the files, can delete from there as well)

[flint grove]

One quick question. I have my network through Apple. If I set up another Thread network through the ZBT-2, can I transfer my devices without losing them, or do I have to add them again as new devices?

[manic lichen]

This channel is not about Thread, but about testing the new Matter server.
Please use #1284966617670881350

[rigid bone]

Just another datapoint.... It took me around 6 to 8 re-tries each to finally get the OTA updates to succeed for my two MYGGSPRAY devices (~1 to 2 hops away from OTBR). Seems most of the times were BDX Transfer errors/Peer no longer responding anywhere from 8% to 48% through the transfer.

[silk plover]

Yup if the prior firmware has trouble staying connected then it can be a mess to update. Not much matter server can do to force a device to respond.

[fringe fjord]

I have an update tomorrow that could improve things here

[craggy veldt]

which issues are being targeted specifically?

[fringe fjord]

I had the same at 8% and needed to remember your post ... and I learned something after checking with Ikea ... They potentially update the chip FW separated from the matter stack or such. So the current MYGGSPRAY is basically executed in 3 chunks which means ... it loads first 8%, do a reboot, request the whole file again, update the second part (likely 48%ish? still need to do that second step in my device), do another reboot, then again for the third.

I think no system is really propared for this "with a user sitting and triggering".
I will tweak matter,js in another version to handle that better

ANd that behavior also seems to potentially trigger a bug in the update state engine in matter,js ... lets see

[loud raptor]

I posted this in the #otbr-beta channel, but this channel seems to be more suiteable for the question.
In the meantime I also had reply explaining the issue. But first the initial question:

When looking at the mesh, the ZBT-1 I use for my Thread network is listed as unknown/external (as is another router, a Nest Hub).
This cannot be correct..

It appears that my ZBT-1 is not known as a Matter device itself, and thus unknown.

[fringe fjord]

Yes and this is exactly the way it is from the veiew of the Matter server ... the Matter server does not know your ZBT-1/2 or the OTBR addon. so it is unknown.

[loud raptor]

I'm comparing this type of setup to ZHA, where my Conbee stick is the 'Coordinator' (and indeed also a Zigbee device)
Why can't coordinators of the Matter network(s) (both Thread and WiFi) be listed as such? And also show up in the Mesh charts.
It would make perfect sense IMHO

[fringe fjord]

because there is not "one" coordinator ... from the view of the network it is a big mesh ... and some nodes (Border routers) are also able to translate "thread radio <--> IPv6" .. but thats not exposed as role.

And because these routers are no matter nodes we just know "yeeahh there is a node with address XXXX which is a router ... but it is no matter node so we do not know anything else about this node

[fringe fjord]

🚩 0.5.5 release with several optimiuations 0.5.5 https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#055-2026-03-11

I especially hope that the "we do not declare a session as dead when device "just" stops responding while read/subscribe ... lets see if that maybe helps for some of the Inovelli cases that had issues to connect.

[silk plover]

Interested in first restart logs on this one? You have enough on your plate, so dont want to spam your DM, if changes are not expected there.

[fringe fjord]

yeahh doesn't hurt 🙂

[elfin hull]

Oh is that useful? I can send you a first restart log as well if you want. Thanks to some combination of you and IKEA, I have gone from 7 Matter devices I was always frustrated with to 39 nodes and growing that are very reliable 😁

[deep drift]

<off-topic> Since there is an ongoing exchange of @fringe fjord with IKEA - the latest BILRESA testnet-dcl firmware v1.9.11 is still reporting wrong battery levels (%) if LADDA NiMH batteries are used. It seems the firmware uses the discharge curve of Alkaline batteries. </off-topic>

[elfin hull]

0.5.5 report, 38/39 nodes were back up in around two minutes, one straggler bilresa remote dragged it out to 4 minutes and change but everything made it up in the end

[silk plover]

Inovelli has released a version 2 of the 1.1.5 firmware for vt31 switches to Main-Net, I am on the Version 1 of the 1.1.5 firmware from Test-Net. Matter Server correctly sees this new version as an update if I select Update from the node details page; however, the HA Updates page does not report/show this as a new update option. Is it correct to aasume this is a HA core issue to report and not a Matter server issue?

[deep drift]

0.5.5 installed smoothly. All 25 devices came back online. However, 15min later two TIMMERFLOTTEs went offline - the first time ever. Now I waited patiently for an hour, but they are still offline. I can provide an Info level log. HA still doesn't show them as "unavailable" although Matter.js shows them "offline". After 2 hours also HA shows them "unavailable".

[flint grove]

my Inovellis are in Firmware 1.1.5 and still going Offline

[rigid bone]

Oh wow! yeah like 8%, then 18%, then 48%, then it finished. Who would have thought they did it that way.

[junior sedge]

huh. if you have limited ram and flash so you can't stream to an alternate partition in an a/b setup i guess that might make sense to try… kind of a tricky thing to do to save a few cents bom tho :(

[fringe fjord]

And BTW also for the matterjs server I just analyzed my behavior ... yes I can optimize but we were managing the situation ... it is just in between looking strange because progress gets back to Querying and 0% .... but yeahahh lets see

[fringe fjord]

lw<ys intersteing to look at startup logs ... showed up interesting things. info loglevel is suufficient. wehne you do updates, start the server and wait till all nodes are online and grab the 10k logs and drop as DM ...

[fringe fjord]

wait a bit. HA is asking for updates ... there is currently no "push" way ... so when you click in Matter server dashboard you force a check but HA needs to ask for updates to get that info ... so should appear within ??? hours 🙂

[silk plover]

If I am not mistaken this Main-Net update was released yesterday or the day before, so its been 12 or more hours and nothing in the HA updates sections even if I force a check.

[fringe fjord]

That would be a qeustion for the HA integration side of things tbh ...

[silk plover]

That was my assumption, just wanted to confirm as you respond much more promptly. thanks

[fringe fjord]

Ok Ikea wise, as written above the update also progresses when once started via HA if you keep reeeaalllyyyy patient and let it just run then it will finish through all cycles ... I work on an optimized version

[fringe fjord]

They are aware and planning fixes

[buoyant nova]

If a check is forced, HA sends a CHECK_NODE_UPDATE to the Matter server and expects a MatterSoftwareVersion JSON back if there's an update.

[fringe fjord]

And when not forced @buoyant nova whats the default interval to check normally?

[buoyant nova]

HA seems to send CHECK_NODE_UPDATE every 12 hours by default

[strong hazel]

Thanks for your work on the new matter server.

I have 3 pieces of Aqara Hub M100 which are matter controllers and yet they do not show in the new matter server.

[elfin hull]

I'm not familiar with the M100 beyond what I just googled, but what you are sharing looks correct to me. The M100 looks like its primary connection is over Wi-Fi, so you're seeing it as a node in your HA fabric but you won't see it as a known node in the graph, which is explicitly a Thread Network Mesh. The M100 seems like it can connect Zigbee or Thread devices. Do you have any devices bridged to your M100s of either type? I think what you would expect to see in your mesh graph would depend on that.

[strong hazel]

Aqara Hub M100 can act as an Aqara Zigbee bridge. It is also a thread border router, matter controller, HomeKit, and so on. There is no Aqara Zigbee connected to them even though I have some Aqara devices.

I use Aqara Hub M100s as thread border routers and as matter devices connected to both Apple Home and Home Assistant. Additionally, I connected them as HomeKit devices in Home Assistant.

Sometimes, my Schlage Encode Plus(es) connects to the Aqara Hub M100 as end devices.

[warm bone]

So, I am not 100% sure what I need to do here

[warm bone]

I created the directory and file and put both in the env var and i am still getting Failed to set up file logging: Error: ENOENT: no such file or directory, mkdir '/home/user'

LOG_LEVEL=debug #critical error warning info debug verbose
LOG_FILE=/home/user/docker/.matterjs-server/data/logs/matter-server.log

[fringe fjord]

if you run that in docker the path you provide needs to be the path inside docker ... It says that there is no /home/user dir

[warm bone]

that did it, and makes sense. thanks

New Question: I have two Homekit Thread devices and one Matter device. The only thing that shows up on my Thead Network Mesh is the matter device. Is that intended? Guess the verbiage may be a bit confusing but I don't fully understand the differences between the two.

[fringe fjord]

honestly ... 🤷‍♂️ ... we use what the thread diagnistic data tell ...

[left rose]

The Matter controller can(1) only see/ask devices that are commissioned to its fabric. HomeKit (only) devices don't speak matter, so they cannot a) be part of the controller's fabric and b) expose the thread network diagnostics matter cluster. If the matter device doesn't know about its thread siblings the matter controller can't do anything about it. It is the same reason why border routers don't show as part of the matter fabric. Because they aren't.

(1) at this time

[fringe fjord]

But thats the part I wondered Boris ... basically Thread is transport layer ... so for "thread nodes" it is normally irrelevant if they are matter or not ... thats why i would also have assumed that the "pure Homekit" are there

[fringe fjord]

🚩 Matterjs Beta Server 0.5.6 is available and should optimize more strange edge cases and especially handle the Ikea "4 restarts in one update" case much better. Sure the progress. will reset to 0 in the middle and maybe also jump status but it should be all "time wise near together", so still feel strange but noch stuck 🙂 Also includes some more optimizations that ideally benefit in spontan device reboot cases. More details: https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#056-2026-03-12

Please report any issues. I do not need restart logs this timebecause most optimizations are for "reconnections. during run", but if you see strange things sure send a log

[left rose]

My assumption here is, that the Matter device just doesn't expose the relevant information in the cluster or the cluster at all (e.g. my Aqara contact sensors don't have the thread network diagnostics cluster).

[fringe fjord]

Could also be

[lilac temple]

I was having some issues. Logs below, context here:
Initially upgrading to the matter.js server did not work. It would download, attempt to install, and give me an exit code 137. Over and over, for days even after uninstalling everything, rebooting, etc. I had a device that was failing to upgrade with the old server.

When I enabled debug logging, the matter.js server finally installed. It's been running since, until it crashed this morning and immediatly attempted to download a new instance, which failed once or twice. Logs below:

[grand kite]

2026-03-12 09:32:19.591 INFO   ClientSubscription   Subscription bd14fc83 to peer @1:49 timed out after 1m 38s
2026-03-12 15:31:52.333 INFO   ClientSubs~onHandler @1:49•571d⇵d5d2 Ignoring data report for unknown subscription ID bd14fc83

some interesting logs related to my "every 6 hours" problem

the "storm" is always proceeded by a flurry of "Ignoring data report for unknown subscription ID" messages

looking at the ID in question for one of them, I found these 3 log entries that I thought were interesting

as they all happened around the time of a storm

[lilac temple]

It shows up for me. Go to Matter Server, click on the node, and select "update" from there.

[silk plover]

Yes, it shows up for us all from matter server, my issue was with the HA update mechanism and how much Matter server contributes to what is displayed there. After @buoyant nova response still not clear if MatterJS server is giving an appropriate update available json to HA or if HA is not seeing the firmware as materially different even though Matter server obviously does.

[fringe fjord]

Yes. In general that seems like device restarts. Reason unknown. Some devices then remember old subscriptions (why ever, bug) and try to reposition them. My suspicion is that there is any kind of network issue or reconnect or such which caused that. So as already written in DM something happens to your thread network ever 6h. Find out what.

[lilac temple]

They provided an update with the same version number. Should have been 1.1.6 or 1.1.5.1 or something.

[grand kite]

Ya, it's very weird that something is causing it to go back to the old subscription id.

[silk plover]

Yes they did, it was public release 1.1.5r3 versus the privately release pre-existing 1.1.5r1, Matterjs server can differentiate the 2 HA can not. They did not want to change the version number and have to go back to the CSA for new certification.

[fringe fjord]

So when I see it then this is the essence:

./run: line 102: 180 Killed npm install --foreground-scripts "${npm_package}"
[15:23:59] ERROR: Failed to install Matter Server from npm. Aborting startup.

How much ram is free?

Something is killing the process.

[lilac temple]

I can check. I'm running HA green with nothing major on top.

[fringe fjord]

As said if my assumption is right then these devices support a so called feature „persisted subscriptions“ and keeping old sub ids is a bug of that in past versions ;-))

[lilac temple]

Huh, 3.7 GB / 4GB.

[fringe fjord]

But what that error means and it gets killed is a good question and maybe better asked in the normal H addon released channels too.

[fringe fjord]

Free or used ;-)))

[lilac temple]

used I believe

studio code server was taking up more than 2GB of ram.

[fringe fjord]

Then I guess you know what killed the server potentially 🙂

[warm bone]

So, and again this is at the 1st grade level, while the protocol is advertised/promoted/propoganda as thread/matter - they are really two different protocols? I understood they worked together?

[silk plover]

Thread is just a network on its own. Apple and nest used thread to transmit data before Matter.

[elfin hull]

Not a 1st grade level, but Matter is an ontology, but you could also call it a standard or language maybe? It formalizes nouns and verbs.

[elfin hull]

Thread is a network, and is one of the two network types that Matter may be "spoken" over

[grizzled burrow]

@fringe fjord I got the Aqara G350 connected to Home Assistant. Zero controls. Forwarding the logs via DM.

[elfin hull]

Is that a Matter 1.5 camera? Is Matter 1.5 even supported in matter.js yet?

[grizzled burrow]

Yes and No. He wanted a dump from the camera.

[junior sedge]

everyone always forgets that matter works over wired ethernet too :) https://www.smartwingshome.com/pages/the-worlds-first-poe-matter-over-ethernet-motor

[elfin hull]

the moment I committed to two I realized I was almost certainly wrong because why wouldn't it 😂

[grizzled burrow]

The first SmartWings Matter over Thread blinds (about 2 years ago) were a complete disaster with broken firmware and non-OTA updatable firmware. But about a year ago with firmware 1.3 and new motors that support OTA, they have been flawless. Zero manual intervention on my 10 blinds over the past year.

[elfin hull]

I had given up on their matter-over-thread motor after that terrible first generation, so my most recent purchases from them were all z-wave. Then I learned there was a new matter-over-thread motor and they replaced my first gen blind motors for a reasonable price, and they are fantastic. I'm a bit annoyed but it's true, the new ones are absolutely top notch.

[grizzled burrow]

They replaced 8 of mine TWICE for free. Third time was a charm in early 2025.

[fringe fjord]

Please open an issue for the HA matter integration and add a json diagnostic dump of it.

[fringe fjord]

Matter is an application layer protocol that can be run on several transport/network layers. Transport is currently Thread, Ethernet, WiFi and for commissioning BLE. Matter is basically transport agnostic.

[fringe fjord]

Matter Community Sync recording and notes -> #1346944149303066711 message

[fringe fjord]

🚩 FYI: 0.5.6 has a bug in the event dta reporting. I missed a detail, fix will come in 0.5.7 later today

[left rose]

So, and again this is at the 1st grade

[fringe fjord]

🚩 Matterjs server 0.5.7 released which should fix the Event payloads - https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#057-2026-03-13

Matterjs Server Networking Optimizations

[fringe fjord]

According to GitHub it is verified fixed.

[fringe fjord]

So guys as informed I will be out starting tomorrow for a week and in Japan. So my answers will be a bit different timing wise and need to see how much time I have 😉 but I will read and process anything afterwards.

[craggy veldt]

Sicheres Reisen :)

[raw veldt]

I bumped into an annoying quirk: battery thread devices odten don't seem to update their ThreadDiagnostics cluster data unless you power cycle them. I've seen this on both IKEA and Eve battery devices. Maybe they time out eventually but 12+ hours after nuking their old neighbor and the data wasn't updated. Powercycle = instantly fixed and reported on the thread graph.

[fringe fjord]

The chart has a reload button for these data. Select a node ode klick the circle reload button on top of the node details. It allows you to choose what to load.

[raw veldt]

Yes, I've been using that. For whatever reason, it doesn't remove neighbors unless I power cycle the device by removing its battery for a few seconds. I've seen this on ALL my ikea devices, and at least one Eve device (eve motion).
A reload appears to be successful (according to the log, there's a read and reply) but the ghost is still there. If I power cycle it, the ghost disappears within seconds.