new-matter-server

[compact root]

Welcome to the feedback channel for the new Matter Server!

You can see the full overview of the testing phases for the new Matter Server here, but for convenience here are the main goals of this channel during Phase 1 (our Alpha testing phase).

• A place where the community can help validate the basic functionality of the new Matter Server based on matter.js
• Ensure the storage format migration works as expected
• Verify that the server connects to all online nodes in the fabric, confirm that the dashboard shows the correct values for all nodes and works correctly, and verify that adding/removing nodes will update the <long-number>.json file properly

We welcome anyone to join in testing this during any phase!

Please review the Alpha Test Instructions

for instructions on how to get started testing. Much like how we notify chat on version updates in the #beta channel, you can expect updates in this channel when new versions are released and links to the changelog.

Phase 2 will be the official release for all "Beta" Home Assistant add-on users, once basic functionality is verified. We'll update this channel once we move to beta testing! 🙌

[reef smelt]

Can't wait for #newer-matter-server and #no-really-this-is-the-one-matter-server

[tired pond]

#final-matter-server_2_edited_v3.7

[compact root]

Don't tempt me. 🤣

[manic lichen]

https://tenor.com/view/gifaton-galicia-nothing-else-matterse-acordeon-metallica-gif-12958656

[heady token]

this for matter?

[manic lichen]

The new matter.js based server.

[heady token]

dark matter

[raw sigil]

I'd be curious: Why the choice for JS / TS ?

[fringe fjord]

Basically I am a JS/TS developer and my goal was to use it in some projects and ideally without native components.
So back then as I started (2,5 years or such) there only was the chip which is a big binary thingy. So doing that as node.js module would have been compiling chip sdk on user computers and such and create a JS binding for that. So I joined a project that already started building it with the goal for a "pure Javascript Matter implementation". And so we ended with interesting dynamics with matter,js and in the meantime me leading the efforts here as a OHF project.

Project chip is mainly designed for devices. So the Python controller basis is mainly developed and used in chip for testing ... so Marcel and the other often run into issues where very fast they would need to touch C++ code instead python ... and so thats. all limits the options to make it the best experience for all users. With the new Matter server based on matter,js we can all do this because we have full control and the JS/TS language is also very common (I think it fights year by year against python other who it place 1 and who 2 :-)) )

So to also repeat here: 0.2.3 of the Matter Server was released today, all currently known issues should be fixed.
Changelog see: https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#023-2026-01-15

[chrome oxide]

WOW!!! What a lot of work... Well done, even though I didn’t test it. 😉

[junior sedge]

One thing I'm curious about - at some point I suppose a vendor might implement a feature that can be used via binding but doesn't have any way to be used directly from a matter controller. (pretty unlikely at the moment given lack of support for binding in the matter controller ecosystem, but i'm thinking ahead here). Has there been any thought of the matter server supporting binding to itself (similar to how zigbee integrations support binding remotes to the coordinator so they can receive the commands the remote sends?)

[fringe fjord]

The matter server dashboard that it opens (check under addons) already offers an experimental way to add matter bindings when devices support that. So if you know what you are doing you can already experiment with it. But that’s not convenient for multiple reasons and so it is just a testing feature there. There will come a time where this is hopefully better exposeable to users a d then we will do. (Same for matter groups and scenes)

[reef smelt]

I really like that there's a dedicated dashboard, even if there isn't much you can do with it yet. I did get to remove a stale node already

But I also use Z2M and zwave JS UI, so I'm definitely biased toward standalone tools that integrate with HA 🙂

[fringe fjord]

The dashboard in 0.2.3 also has some very simple commands being displayed for OnOff and LevelControl clusters. We definitively do not want to mimic a full device view (I think thats better done in HA), but for low level testing if attributes behave as expected and such things it is maybe handy to have this on the page together with the attribute data

[fringe fjord]

I just released version 0.2.4 with some enhancements in the dashboard and latest OTA fixes (so, also try out software updates if you have any open!). This version also re-adds the Eve cluster engery polling for the devices that are not yet updated, but try to update if you have an update available!
Changelog: https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#024-2026-01-16

[fringe fjord]

We found something else to optiomize, so we have 0.2.5 now: https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#025-2026-01-16

[left rose]

Quick question regarding non main-DCL hosted matter device firmwares: We could inject the OTA json snippets into the old python matter server by placing them in addon_configs/core_matter_server/updates and hence be able to push beta (or any) firmware to a device. How would that be handled by the new server?

[kind sundial]

@fringe fjord Has .js been pushed to the Server add-on beta channel yet?

[fringe fjord]

We plan this for beginning next week. But we also have a version for that and will update the github page later how to manually tweak your addon ... if you like to try that 🙂

[fringe fjord]

@kind sundial https://github.com/matter-js/matterjs-server/blob/main/ALPHABETATESTS.md#replacing-the-matter-addon-docker-image-home-assistant-os

This describes to add a pre-version of the future planned addon we plan to release next week to your system. If you turn on "Beta" the JavaScript server is started, if you turn off beta you run the Python server as before.

It automatically installs the latest Matter server verison when Beta is enabled on start and runs Python Matter Server 8.1.2 which is the latest/final version of that variant.

[sturdy lance]

Updated mine and just can confirm everything works as expected 🙂 Well, I only have 4 Matter devices.

[fringe fjord]

Awesome, and I really like success postings 😉 (honestly a bit more than issue postings, but I can handle both gg)

[sturdy lance]

I can totally relate 😄

[fringe fjord]

Oh I near to missed that question. The new server supports that too. The directory can also be specified and defaults to same as before, but I decided to combine this with the test-dcl flag because in fact all such things basically are test things. Open for discussion.
So copy a file in that directory (just the OTA file, no extra JSON file needed, we grab all infos from the OTA file header) and restart the server. it will then be auto imported and should be available. It should be directly available in the dashboard as usual. We always additionally check the CSA DCL (prod, and if enabled also test DCL) and when we find a new version there then this is offered.

[left rose]

awesome!

[fringe fjord]

Ahh to complete above: When the file was successfully imported it is removed from that directory where you put it in and shows in the log that it was imported, so do not wonder when it disappears ... thats a good sign 🙂. json files will be ignored and stay in the directory (because I know in the past it was needed to create an extra json file with dummy details)

[left rose]

Since it's been mentioned in the documentation and during the Community Sync that the first start is going to take a bit -- what do you think: Will the internal storage of a HA Green easily survive the migration process for a ~160 devices set-up? I am inclined to test -- at the same time have to be careful (wife would be upset, if i break it)

[fringe fjord]

it would be bad if not ... in fact it is not about "take much disk space" it is more we will create many small files, but usually the OS is good in sorting that out, but because I do not have such a huge "real setup" I can not tell anything. But I would also have no idea why it should not survive it 🙂 Do you have doubts?

And said that: if this breaks it ... any other update would have broken it somewhen soon in any case too ... see it that way. So i guess the "risk" is the same then "running that system another some months"

As always ... have a backup 🙂

The migration is "just I/O"

[left rose]

just recalling that it is eMMC and not SD, so should be fine indeed..

[fringe fjord]

https://tenor.com/view/simpsons-homer-simpson-gif-13518564799186478937

Then good luck 🙂

[solid flare]

Just installed the beta add-on, the GitHub steps worked perfectly and everything appears to have migrated (mix of Inovelli, Eve, and arre devices for me). Migrations seems to have been quick (1-2 minutes) and successful. Nodes are currently resubscribing. Intel N100 system and 53 nodes. The only real thing I see is a lot of these messages:

  at PairedNode.#handleReconnect (/opt/matterjs/node_modules/@project-chip/matter.js/src/device/PairedNode.ts:557:19)
  at PairedNode.#ensureConnection (/opt/matterjs/node_modules/@project-chip/matter.js/src/device/PairedNode.ts:686:20)
  at PairedNode.getInteractionClient (/opt/matterjs/node_modules/@project-chip/matter.js/src/device/PairedNode.ts:853:21)
  at Nodes.interactionClientFor (/opt/matterjs/node_modules/@matter-server/ws-controller/src/controller/Nodes.ts:80:33)
  at ControllerCommandHandler.handleReadAttributes (/opt/matterjs/node_modules/@matter-server/ws-controller/src/controller/ControllerCommandHandler.ts:389:42)
  at Object.handleReadAttributes (/opt/matterjs/node_modules/@matter-server/ws-controller/src/controller/ControllerCommandHandler.ts:144:22)
  at CustomClusterPoller.#pollNode (/opt/matterjs/node_modules/@matter-server/ws-controller/src/controller/CustomClusterPoller.ts:245:41)
  at CustomClusterPoller.#pollAllNodes (/opt/matterjs/node_modules/@matter-server/ws-controller/src/controller/CustomClusterPoller.ts:217:28)```

[junior sedge]

huh, doesn't seem like something that would actually cause a problem (the poller is just saying "hey, i can't poll right now"), but probably still worth filing a github issue for it. Maybe the poller could be changed so it isn't started until the connection setup finishes?

[meager zenith]

Do you plan to implement a feature similar to z2m's "last_seen" diagnostic, allowing users to track how long a device has been offline?

[fringe fjord]

Good catch I will add a safe guard for the polling that it waits till the device is connected. But it should not do any harm or?!

[fringe fjord]

We could use the "last interview dat" for this that we return because I update this with any "new connection established" which automatically means that data are updated again ... so strange name but would fulfill the purpose. The Python server might fill that field differently

[meager zenith]

Do you update "last interview dat" continuously or only when a new connection is established? Ideally, the field would report the last time any successful communication was received from the node.

[fringe fjord]

ahh you mean "subscription alive" kind of ... we have that internally but it's not exposed. Feel free to open a feature request in the Home Assistant tracker because also needs a representation and such

[fringe fjord]

No issue needed this time. PR already up. will be in next version. WIl will release one in 20-30 mins

[solid flare]

It was just clogging up the logs. At least with Python Matter Server I always had a lot of CHIP timeouts so I'd have the error logging level set to error (I seem to be getting warning and info logs even when set to error right now). My Thread network still seems to have some instability, but it seems like devices re-interview and come back online much faster when it happens, as well as after restarting the add-on

[fringe fjord]

the old one had two loglevels - one for the server, one for the sdk! the sdk error log is ignored by the new one and only the normal loglevel counts. Might it that be?

[solid flare]

yep that is exactly it! Makes sense

This may be expected since I just restarted the add-on but I now have several instances of this error as the devices re-interview. 2026-01-16 10:47:52.146 ERROR Multiplex Error: [session-closed] Session •unsecured#7982b0e24a6b4d44 ended at UnsecuredSession.get channel (/opt/matterjs/node_modules/@matter/protocol/src/session/Session.ts:299:19) at MessageExchange.get channel (/opt/matterjs/node_modules/@matter/protocol/src/protocol/MessageExchange.ts:245:29) at MessageExchange.send (/opt/matterjs/node_modules/@matter/protocol/src/protocol/MessageExchange.ts:452:20) at async ExchangeManager.#onMessage (/opt/matterjs/node_modules/@matter/protocol/src/protocol/ExchangeManager.ts:268:17)

[fringe fjord]

ok for that I would need more log but likely timeouts or such on establishing a session to the devcie

[solid flare]

definitely loving matter.js so far - add-on restarts used to be 30-60 minutes for all nodes to come back online. 3-4 minutes after this restart all 53 are back!

[polar knoll]

That's strange. 🤔 Starting the "old" Python Matter Server took under 5 minutes. 71 nodes on my Thread Network paired to Apple Home and 59 nodes also shared to HA.

I'm seeing lot's of otaSoftwareUpdateRequestor with the new matter.js server. Seems that it cycles through all nodes. Is that normal? Server is running since 2h.

2026-01-16 17:56:52.259 INFO   ClientEventEmitter   Received event otaSoftwareUpdateRequestor.stateTransition on server.@1:b previousState: 2 newState: 1 reason: 1 targetSoftwareVersion: null
2026-01-16 17:58:10.837 INFO   ClientEventEmitter   Received event otaSoftwareUpdateRequestor.stateTransition on server.@1:c previousState: 1 newState: 2 reason: 1 targetSoftwareVersion: null
2026-01-16 17:58:11.757 INFO   ClientEventEmitter   Received event otaSoftwareUpdateRequestor.stateTransition on server.@1:c previousState: 2 newState: 1 reason: 1 targetSoftwareVersion: null```

[fringe fjord]

No, thats just logging evenst from the nodes that we receive when we subscribe to them because intiially we get all events that the devices have stored (and basically ignoreiot). After that we knopw the "max even number" that we last got and then we only request newer events next time. so should only be visible.
In that case you see that a device had tried to get a software update and returned to idlestate because none was there.
So yeah maybe we can change loglevel ... but it is just intermation

But on the other side ... roughly 5-15 mins after start of the server we announce ourself as the sopurce for software updates and when we do this devices might ask if an update is available ... so could also be that

[polar knoll]

Last question for today... 🤪 Some of my nodes are stuck at "Update in progress". But there is definitely no new firmware available for these nodes.

[fringe fjord]

reload the UI ... i would assume these "old" events might get the UI in strange status ... I need to check what the UI does when in this case. please open an issue

Ah can you check what the endpoint 0, cluster 42, attribute 2 has as value?

[polar knoll]

Yep... has value "3" at the moment.

[fringe fjord]

that means for whatever reason it got a "busy" response and is doing a delay before querying again fort an update ... could be us and strange effects or could also be another fabric ... so giuve it some time ... should sort out . If you have a debug log then share as issue please and I look

[polar knoll]

I'm seeing this mostly on battery powered devices from Bosch (RBSH-SWD2-ZB contact sensor). The Aqara bulb seems to be the only FTD.

I can open an issue for that if it doesn't settle in a few hours. But hard to find the exact lines in the log.

[fringe fjord]

Asd said I would need to see a log of that to follow what exactly happened on the node to bring it there ... else bumb all log as zip into the issue and tell me the node id ... I will figure out what I need

[fringe fjord]

Hello everyone, all today mentioned things above here, in GitHub and an other optimization are included in v 0.2.6 just released. CHangelog: https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#026-2026-01-16
just restart the addon and it should automatically install this latest version from npm. If you run docker the new version should be also there within the next 60s

[fringe fjord]

0.2.6 should display you the real state here and not always " Downloading update" 🙂

[solid flare]

restarting the add-on pulled it in for me. So far so good

[fringe fjord]

it all depends on the mix of thread and non-thread devices and such, so can easiely be different experiences depending on things. Also different kinds of thread have different behavior sometimes, espeiclaly when under load

[solid flare]

In my case I suspect I have some Thread devices running an older version of the standard that are causing issues. Things are more stable and responsive with matter.js but I can see where every 30 min or so I still have a specific node that goes offline for 30-60 seconds then comes back

[fringe fjord]

would be interesting to get info on what kind of devices that are, maybe disgnostic log ... if thats e.g. ICD decices then one of the next planned things might help too ... but yeah depends on exact reasons ... so logs and infos help..just opennissues and we have a look

[meager zenith]

I created the discussion: https://github.com/orgs/home-assistant/discussions/2518

[solid flare]

it seems to mainly be the Eve Energy Outlet (US) and sometimes their light switches. Possibly the Inovelli fan switches and canopies (which they commited to updating to Matter 1.5 last week on their 2026 roadmap). It used to be much worse when all my Inovelli 2-1s were on the factory firmware and before the Eve Energy plugs got the 3.5.0 firmware. My hope is with some of the devices getting CSA certifications for new firmware with 1.3/1.4 back in the fall things will continue to improve (plus Apple finally bumping HomePod OS 26 from 1.3 to 1.4 too, whenever it finally happens)

the one I've seen flap 2-3 times this morning is Eve 20ECN4101 (US Outlet) running 3.3.0. Hard to say if it is the device or the route to it having issues though

[fringe fjord]

Yeahh thats always the "fun" to find that out. (not to make it smaller) When the issues also happened before you (and I) know that it is not because of matter,js. 🙂 ... mainly in order to know if it is a specific issue or "just the normal Thread weirdness, what we want to make better but maybe just the Server switch is not "the final solution".

In fact when you see "subscription timed out" or such with matter,js and also had such cases before the. it gets more likely that it is a device or route issue because matter,js waits even longer before it declares a subscription timed out ... so there really was nothing coming in.

[polar knoll]

The new matter.js server is sooo blazing fast! 🎉 It took under 2 minutes after a restart (with downloading the new version) to get 58 nodes back online. Only one Ikea BILRESA needed a bit longer to get back. Nice work @fringe fjord ! 👍

[fringe fjord]

Thanks, great to bring some improvements

[heady token]

but, i mean, the NEWEST CHANNEL??

[sleek crane]

I’m a little confused on the state of this integration. Is this ready to replace the current matter server for most users? Will there be a migration option for seemless transfer to the new one? I read documentation and was still confused on the state and readiness for mainstream adoption.

[silk plover]

Not yet for most HAOS users to install, still a manual process

@fringe fjord finalizing a HAOS addon Matter server beta that will be be a drop replacement. once the beta option is selected from the Current Matter server add on config.

Yes to seamless migration

[fringe fjord]

I currently work through the reported things and I expect a new versio of the addon for all users (8.2.0) by next week. The matter,js variant is then being used when you set the addon to "beta". If youdo not use Beta you still use the python one. with this we then try to get more testers.

Somewhen later and when all are happy then we will update the addon to 9.0 which means that python server is gone completely. But thats a later step 🙂

And yes the migration should be seamless and automatic - and in 8.2.0 also allows you to switch free between python and matter.js server when you need to (hopefully not too much because this can have other effects, but other story).

[runic scaffold]

can somene clarify something for me, can i use this to add non matter devices and make them appear as matter devices?

looking to replace emulated hue

[analog sphinx]

No, this is to connect matter devices to HA, you're looking for matterbridge, however it stopped working for me without a physical hub

[runic scaffold]

isn't appletv or alexa studio considerd a physical hub?

[analog sphinx]

I believe so, but I have neither, no Google hubs either

[runic scaffold]

just installed and exposed 1 light, seems to be working.. added to my home app

[analog sphinx]

You sure you don't have any Google hubs? Nest? Speakers?

[runic scaffold]

no google devices at all, just a AppleTV and Alexa Studio

[analog sphinx]

That might be why it worked, anyway glad it did, wrong channel for this tho

[heavy dirge]

I added about fifteen Matter devices to my test environment. So far, it's stable using the the server.

[heavy dirge]

Are there any new feedback reports on the new server?

[fossil topaz]

I'm not sure if I'm on the new matter server or not but I'm running into some new matter HA issues
was the new server update pushed to main?

[manic lichen]

No. The new matter server is beta.

[fossil topaz]

hmm, must be a different issue then, at least I know it's not the matter server

[warped nexus]

hey everyone... i'm not sure where to pop this question or seek for help. i have latest haos installed in a proxmox vm. i have the skyconnect1 usb passed through to the ha vm and ha does see the usb. i installed all related to matter/thread, but when i try to pair a device, i constantly get pairing failed... not sure what might be wrong... i have no other hardware where i would install haos in order to skip the virtualization/usb passthrough (in case that might be the issue here)

are there any best practices when haso is installed in vm that might help eliminate thread-related issues?

[silk plover]

Try starting a new topic with your questions here. https://discord.com/channels/330944238910963714/1284966617670881350

[deep drift]

Is it already a Beta and available via the Addon-Beta-Channel?

[silk plover]

No still Alpha, no beta directly thru HA yet as far as I can tell. hopefully this week

[compact root]

Keep an eye on the pinned posts for when we switch it over to the beta channel - I'll be sure to watch for that. 🙂

If you want to get involved with this first phase, though, you can follow the steps using the links in the current pinned post. 😄

[modern olive]

Can this new server be deployed without affecting the existing one, or will it take over the devices?

[junior sedge]

the intent is for it to replace the existing one - one thing specifically being tested is ensuring that it imports the configuration from the existing matter server correctly.

(you should be able to switch back to the old matter server tho; the new one keeps compatible configuration updated, or use a backup)

I'm actually kind of curious whether you could add a fresh/from scratch config of the new matter server running parallel with the old one using the "Add Hub" option in the Matter integration?

[fringe fjord]

Hey All, I just relewase 0.2.7 of the new server, so please restart the addon (when you use the tweaked one from Julian) and verify if still all is fine. If there is no negative feedback I tend to release the 8.2.0 addon tomorrow. Changelog: https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#027-2026-01-22

[fringe fjord]

see post below, if all is good likely tomorrow

[fringe fjord]

It will take over the devices because it takes over the commissioning. But basically you can switch between python server and the new one, just not have them run in parallel.

[fringe fjord]

Sure so you can, but then you need to commission all devices again

[deep drift]

Awesome! Looking forward to tomorrow!

[modern olive]

Ah cool, thanks for the clarification. So while I have the new server up, I'll lose control over the devices but then I can just switch back to the Python one and not lose anything, not have to re-pair devices, etc? If so, I'll try to find some time in the next few days to give it a go.

[fringe fjord]

No, the new server migrates all nodes and data from the python one and so offer the exact same functionality than the python server. so you "Loose" nothing

[undone hull]

Could you please put the Changelog link to the new github release announce message(https://github.com/matter-js/matterjs-server/releases/tag/v0.2.7)? It currently is: "Official release, check CHANGELOG.md for details". CHANGELOG.md could be a link (https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#027-2026-01-22). The same for matter.js releases.

[fringe fjord]

need to see if I cn automate that, right now you need to click manually 🙂

[undone hull]

You could just put https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md instead of https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#027-2026-01-22 , it woould still be better 🙂

now I have to click a few times until I get to the Changelog.

[fringe fjord]

will be adjusted with next release

[fringe fjord]

I just release 0.2.8 of the server package which fixes BLE commissioning. https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#028-2026-01-22

[silk plover]

Matter-Server 8.2.0 addon just showed as available on my HA update notifications.

Jumped in feet first, installed and set to use latest beta in add on configuration menu. Matter over wifi devices come up first, 93 Matter over thread devices more slowly coming online, expected this to be the case, with first start.
took about 4 min for all online from first start.

Would love a link in the Matter server node info page that takes us to the the actual device in HA. Very difficult with large installs to know/remember which Node # equates to which HA device.

[deep drift]

Jumped in too, 22 of 22 Matter over Thread devices online. The first start has been slower than Python - as expected. IKEA's MYGGBETT door/window sensors were a bit reluctant and the retries cost one sensor almost 10% of battery capacity - strange. But at the end the connection was restored.

[sleek crane]

Just tried beta. Nice work.

[deep drift]

I have only one error in my log:
2026-01-22 16:04:47.410 ERROR AttributeDataDecoder Error decoding attribute 1/0x130afc01/0x130a0014: Unexpected type 10, was expecting 4.

[silk plover]

have same error reported, unsure of any impact

[buoyant nova]

It's known, but won't cause any issues.

It's an unused (and unknown) Eve attribute that fails to migrate properly. Will work shortly when it's re-read.

[fringe fjord]

faster then I can write here ... So YEEESSS we decided to release the Addon officially already today. Have fun.

One note: OTA is currently not working, Fix in progress

[deep drift]

The self-destruct switch 😅

[compact root]

Surprise!

XD

[fringe fjord]

yes the first start takes a bit longer ... any follow up start or restart should be much faster

[fringe fjord]

Yes was reported and queued as fix for next version already

[fringe fjord]

No in this case it is a known ione but had wrong datatype in the details ... fixed on Github already ... version soon

[deep drift]

There is an info about a timesync failure:
2026-01-22 16:09:01.836 INFO ClientEventEmitter Received event timeSynchronization.timeFailure on server.@1:1b undefined

[fringe fjord]

Yes, thats coming from the device because the python and also the matter,js server do not have TimeSync currently, so we can not give current time to the devcie. Thats on the Todo list

[buoyant nova]

Ah right, should have looked at the attribute ID here. 0x130a0014 is the Eve pressure one.

[silk plover]

Thank you for the the work @fringe fjord and the other devs to get here. Very much appreciated. big step forward.

[deep drift]

Yeah, Eve Weather - Pressure does not get any values anymore.

[buoyant nova]

When Ingo releases a new Matterjs-server version with the fix, you can just restart the add-on and it'll pull the latest release with the fix and then work again. (Alternatively, you can temporarily also switch back to the Python server if you really need to.)

[deep drift]

Yes. I would like to chime in. Great work and almost no issues! Hat tip! 👏

[raven grotto]

Migration worked well on my small environment of ~ a dozen devices split pretty even between mains powered routers and child devices

[deep drift]

Yeah! Lightning fast! The second start just took about 50% of the time the Python server needed. Well done! 🚀

[haughty stump]

+1000% on linking from matter server to devices. I always struggle to track down which device is offline to an actual device in HA.

[silk plover]

seconded all 93 back up more quickly than prior with a restart.

[junior sedge]

huh, the "LegacyDataInjector" step seems to be running pretty slow for me, taking 10-30 seconds per device.

looks like it's IO bound, lots of small slow synchronous writes maybe?

[fringe fjord]

no we write async but need to wait for the FS once per node too ensure all is persisted, so yes bound to your FS speed ... but only the first time

[junior sedge]

hmm, might want to see if there's some way to speed that up like doing a batch transaction with all the migration stuff at once particularly to help people using HA on slower storage with large numbers of matter devices

(iirc it's using file based storage rather than a transactional database, that would maybe involve writing all the individual updates without fsync, then doing a single sync at the end, prior to doing an fsynced update to flag that the data is valid?)

[junior sedge]

with the new beta add-on, how do you get matter.js updates? Does it pull updates whenever the add-on is restarted? or only when a new add-on version is released?

[buoyant nova]

It installs the latest version whenever it’s restarted

That’ll change when the official add-on release is made eventually

[junior sedge]

right, i'd expect then it'll have a locked version in the add-on at that point (but setting the beta flag will still make it pull in the latest stuff)

[fringe fjord]

exactly. SO right now "beta" for ,atter,js behaves like "beta" for pythonbefore ... always pulls latest. When we end the beta and move tomatter,js completely then the "Non-beta" will be a defined version and beta is the "latest"

[fringe fjord]

but thats really slow ... slowest I saw was from a raspi 4 it was 3s/node

[hollow cloak]

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

[junior sedge]

just to confirm the units, that was about 17 minutes for storage conversion of a bit over 150 matter devices?

[fringe fjord]

The "Simple re-establishing session did not work. Reconnect ... [discovery] @1:1f2 has no known address and No discovery was requested." are completely expected on the first start because we do not know former IP addresses, so we can not connect directly. At second start this should be gone when we already had it connected

So basically the first start basically does

• Convert all data you had before
• Then we need to MDNS discover all devices because we do not know their IPs
• We need a full session establishment because we do not have resumption details
• We fully subscribe to them including reading all data

On restart:

• we directly try to connect to the last known address
• we only request data that were changed since last "connection lost"

[left rose]

Also, the addon seems to report thread-mesh internal ip addresses to the integration. For example:

[fringe fjord]

please open issue ... then I need to check how to sort that out ... need to check what python did

We report what is returned by the device 🙂

[fringe fjord]

aahhh i seee .... The Python server only returned the IPs that were returned via MDNS ... We return the IPs from the device "GeneralDiagnostics.networkInterfaces" ... Ok ... then let me adjust to MDNS

[left rose]

That probably somehow needs to be matched against what's exposed via mDNS. In this case fdf8:cced:c480:1:19dc:527d:e5ae:8f9e is the routed address thats also the only discoverable address via mDNS

[fringe fjord]

Seems I missed that slight difference 🙂 Will be fixed

[left rose]

https://github.com/matter-js/matterjs-server/issues/96

[junior sedge]

anyways, i guess i'll hit the restart addon button since i think there was a new matter.js release since i first set it up, and i want to see how quickly the reconnects work :)

[left rose]

Another observation: For me the matterjs add-on uses 2.67 times the amount of ram as the python version did -- which is in total 16.4%. This is still the first start though.

[junior sedge]

… hmm, well, i have a broken automation somewhere which turns my living room lights off when the matter server is restarted. gonna have to figure that out :)

(i'd bet i have incorrect handling of a mix of available and unavailable lights)

oh, there's an issue with how the events entities are being restored when buttons come online

i dunno if this is actually a bug, and whether it's in the new matter server or an existing thing, but the button event entity goes from unavailable state to restoring the previous state, which causes my automations to get re-triggered as if the button was pressed again.

[lost jasper]

@small totem Hello. Can I ask You something about Your HA dashboard? How to install it on my own instance?

[kind sundial]

@fringe fjord I'm not sure if it's been reported already, but I'm getting this error, it keeps restarting:

Warning: --paa-root-cert-dir is deprecated and no longer supported. This option will be ignored.
2026-01-22 17:28:55.290 INFO   LegacyDataLoader     Loaded legacy chip.json with 1 fabric(s)
2026-01-22 17:28:55.311 WARN   LegacyDataLoader     Error loading server file 12902487403059095876.json: TypeError: Cannot convert undefined or null to object
TypeError: Cannot convert undefined or null to object
    at Function.keys (<anonymous>)
    at start (/opt/matterjs/node_modules/matter-server/src/MatterServer.ts:132:38)
[17:28:55] INFO: matter-server service exited with code 0 (by signal 0).```

[junior sedge]

@kind sundial as far as I know, all previously known errors like that have been fixed, so it's probably something new.

[manic lichen]

No. This channel is not about dashboards and the user has been inactive since 2023.

[lost jasper]

Thanks

[junior sedge]

ah, looks like retriggered events is unrelated the matter server itself, already reported in https://github.com/home-assistant/core/issues/145941

[crisp raven]

hi, I'm having the "Cannot convert..." error as well. Makes the server loop-restart. Happy to help debugging.

[digital nova]

Just installed the new addon version and enabled beta to swith to matter.js. Seems to be working, but I got an assert regarding BLE comissionnning (which I have enabled in the addon settings).

2026-01-22 09:53:00.353 ERROR  ControllerBehavior   Disabling BLE due to initialization error: Cannot find module '@stoprocent/noble'
  Require stack:
  - /opt/matterjs/node_modules/@matter/nodejs-ble/require/require.mjs
    at Function._resolveFilename (node:internal/modules/cjs/loader:1383:15)
    at defaultResolveImpl (node:internal/modules/cjs/loader:1025:19)
    at resolveForCJSWithHooks (node:internal/modules/cjs/loader:1030:22)
    at Function._load (node:internal/modules/cjs/loader:1192:37)
    at TracingChannel.traceSync (node:diagnostics_channel:328:14)
    at wrapModuleLoad (node:internal/modules/cjs/loader:237:24)
    at Module.require (node:internal/modules/cjs/loader:1463:12)
    at require (node:internal/modules/helpers:147:16)
    at loadNoble (/opt/matterjs/node_modules/@matter/nodejs-ble/src/NobleBleClient.ts:22:13)
    at new NobleBleClient (/opt/matterjs/node_modules/@matter/nodejs-ble/src/NobleBleClient.ts:42:9)

Not sure if this is an addon issue or a matter.js issue.

[fringe fjord]

When Whats the absolute numbers? % is always very misleading 🙂

[fringe fjord]

Could it be that maybe events that happened during server being offline are fired on start . If you see that in tze logs then I needd to filter them out. Please create an issue.

[junior sedge]

no, this is specifically that the last event from prior to the matter server being restarted is restored after the device comes back online.

so automations see the unavailable -> {event timestamp, event_type=some_event} change, and unless you explicitly filter out transitions where the previous state was unavailable, that will retrigger the automation.

[fringe fjord]

yeahh I checked, we do not send these out to my knowledge. so should not be the case

[junior sedge]

I think after the device is reconnected, the event entity state should always be reset to "unknown". Since you don't know what events might have happened during the time when it was unavailable :)

[fringe fjord]

is that file maybe broken? because seems so. Please check it and maybe rtestore backup

[fringe fjord]

I could add to ignore that, but that would then mean that the server starts empty ... no idea if thats the best idea then.

[fringe fjord]

same answer for you, seems the json file is corupted... in fact he checks for a "nodes" entry in the json and thats not there, so only we could do it come up emoty ... but also honestly your python server could have the same issue on start

[deep drift]

Where do I see the absolute numbers? "top" and "ps xa" are not showing the addon

[fringe fjord]

If you like pack up the json file and send it to me here as PN or such and I have a look

[buoyant nova]

It’s fine to restore the existing state for event entities. The issue is that the "state" trigger doesn’t work nicely for event entities. You‘ll need to add not_to to the trigger with unavailable and unknown.
With the new trigger types (currently partly in HA Labs), there will be a proper trigger for event entities that do not have any of these restore issues.

The behavior is not specific to Matter.

[fringe fjord]

then likely the BLE could not be compiled on install and so BLE is basically not available. interesting. To know more we would need to see the startup log of the addon basically ... aahh but I have idea ... maybe we need to add some more deps for BLE ... U prepare an addon update for next week - will be addon 8.2.1

[junior sedge]

the thing is, it really only makes sense for the "event" entity to change state when an event happens on the device. restoring previous state isn't an event happening on the device :/

having a distinct state for "no longer unavailable, but no new events" seems quite useful

and the "unknown" state is already used for an event entity that has never had any events

but yeah, not specific to matter

(tho it might need some integration specific work if a decision is made to change the event entity behaviour on reconnects)

[buoyant nova]

The state for event entities will stay as is. It’s aligned with all other HA entities. Changing the behavior just for event entities isn’t ideal.
The main issue is that the event entity doesn’t make a lot of sense to use with the State trigger in automations. That’s why a new trigger type will be introduced to solve this use-case in automations.

[junior sedge]

right, as long as the new triggers make sure to not re-trigger on state restore, that would be fine :)

[buoyant nova]

Yep. It’ll only trigger if a new event actually happens (e.g. clicking a remote button).

[junior sedge]

i'm still not sold on restoring the last event tho, since you don't know if that was actually the last event or if you might have missed some events.

but i guess from the HA point of view it's the last event that was seen by HA

[buoyant nova]

That‘s the case for a lot of integrations. With many integrations, a motion sensor can fire but if you restart HA during that time, it’ll only change state the next time it’s fired when HA is online.

[buoyant nova]

Yeah. I think it’s more useful for the user, too.

[junior sedge]

I wonder how often people are actually checking the previous state in automations from event state triggers (except to filter out retriggers). In most cases the only thing you care about the new event, not which event happened previously.

[buoyant nova]

The event entity can also fire when a device detects low battery, for example. If we wiped that during restarts, it’s not very useful anymore.
You don’t know if and when the device last fired a "replace battery" event.

It would also be inconsistent that the last state randomly disappears for event entities when restarting HA. At the moment, it’s consistently restored for all integrations in the base event entity implementation.

But I agree that previous event entity state will rarely be used in automations.

Still useful for the user to look at that state though.

[silk plover]

@fringe fjord can you make a similar change to the matter-js/matter.js release page and notifications? So that we can check the changelog quickly. Thank you!

[left rose]

Hmm, given that HA Green has 4G of ram that would be 640MB.

[fringe fjord]

0.2.9 of the server published ... https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#029-2026-01-22 ... just restart to get fixes and updates, mainly OTA and mDNS for IPs fixed

[fringe fjord]

planned for next release too. I already added it but git then confused by an error ... but in the end was only GitHUb weirdnessm, so all good

[kind sundial]

Not sure what happened, but reinstalling did the trick, thanks! 🙂

[robust crypt]

Updated to 8.2 and switched on beta (0.2.9 was installed), happy to report seamless transition for my 6 node fabric

[deep drift]

Updated the beta server. OTA update checks still seem to have an issue. All of my Eve and Ikea devices have the firmware status „unknown“.

[fringe fjord]

Where? What you mean?

[deep drift]

The „Firmware“ status should be „up-to-date“, shouldn’t it?

[silk plover]

Eve, IKEA and Inovelli all correctly showing Up-to-date for me. On latest Beta.

EDIT: Aqara and Leviton also correct

[junior sedge]

firmware from a variety of vendors is showing as up-to-date for me. Did you just start the matter server? it doesn't run the first firmware update chck immediately at startup, has a bit of a delay

[deep drift]

Ok, I have just updated and restarted matter.js. Maybe it needs some more time to update the status.

[junior sedge]

the logs should have some info about the delay before the update check, and i think it prints a few messages as it downloads the DCL stuff to get the firmware info.

[fringe fjord]

From my knowledge HA queries the server for updates. No clue exactly when he does that. But basically when an update is checked we validate. So would be interesting what the servers ui tells

[opaque panther]

2026-01-22 16:51:26.444 INFO   LegacyDataLoader     Loaded legacy chip.json with 2 fabric(s)
2026-01-22 16:51:26.448 ERROR  LegacyDataLoader     Expected exactly one fabric with index 1, found indices: [1, 2]
2026-01-22 16:51:26.450 WARN   MatterServer         Legacy data error: Expected exactly one fabric with index 1, found indices: [1, 2]

2026-01-22 16:51:27.363 INFO   Controller~ndHandler Found 0 nodes:

oops

Restoring a backup as this is my production 😬

[lusty linden]

Hi all! I'm getting this after upgrading:

2026-01-22 16:46:03.603 WARN   LegacyDataLoader     Error loading server file 5586628661939700736.json: TypeError: Cannot convert undefined or null to object
TypeError: Cannot convert undefined or null to object
    at Function.keys (<anonymous>)
    at start (/opt/matterjs/node_modules/matter-server/src/MatterServer.ts:132:38)

I've extracted /data containing that 5586628661939700736.json - let me know if you're a dev and would like me to send that to you

[fringe fjord]

Yes please make zip and send me as on or as issue in GitHub. Thanks.

[fringe fjord]

Would love to get the chip.json.

[grizzled burrow]

I'm running the beta..have about 71 Matter/Thread devices. Took about 15 minutes from fresh HAOS VM reboot to bring all devices online. Pretty much on par with the legacy server reboot times. I was hoping for much faster, but hey, at least it's not worse.

[fringe fjord]

Was that the very first start or already a restart after first one?

[modern olive]

@fringe fjord I just flipped the Beta flag on the addon to test this out. It's mostly working, but quite a few nodes are still offline and I'm getting this crash repeatedly in the logs:

2026-01-22 22:13:43.567 ERROR  ExchangeManager      @1:6•ec2c⇵52e3✉082af0d2 Error: [implementation] Behavior cluster$125dfc01 is not supported by server.@1:6.ep1
  at Endpoint.stateOf (/opt/matterjs/node_modules/@matter/node/src/endpoint/Endpoint.ts:201:15)
  at PairedNode.#handleNodeChange (/opt/matterjs/node_modules/@project-chip/matter.js/src/device/PairedNode.ts:787:32)
  at BasicObservable.emit (/opt/matterjs/node_modules/@matter/general/src/util/Observable.ts:377:26)
  at CommissioningController.#handleNodeChange (/opt/matterjs/node_modules/@project-chip/matter.js/src/CommissioningController.ts:732:23)
  at BasicObservable.emit (/opt/matterjs/node_modules/@matter/general/src/util/Observable.ts:377:26)
  at ChangeNotificationService.broadcastUpdate (/opt/matterjs/node_modules/@matter/node/src/node/integration/ChangeNotificationService.ts:47:22)
  at ClientBehaviorBacking.#broadcastChanges (/opt/matterjs/node_modules/@matter/node/src/behavior/internal/BehaviorBacking.ts:422:30)
  at ClientBehaviorBacking.#onChange (/opt/matterjs/node_modules/@matter/node/src/behavior/internal/BehaviorBacking.ts:408:14)
  at Object.store.externalChangeListener (/opt/matterjs/node_modules/@matter/node/src/behavior/state/managed/Datasource.ts:451:43)
  at Object.externalSet (/opt/matterjs/node_modules/@matter/node/src/storage/client/DatasourceCache.ts:56:28)

[fringe fjord]

Please ideally pack the whole data dir and send me as tip as on or as issue. I need to check what’s up there. Does it also happen if you restart?

[silk plover]

first time startup of the beta or was this a restart?

[grizzled burrow]

First time. I also maybe found a bug? One of my Aqara door P2 sensors went offline (prior to upgrade..unrelated). So I had to factory reset and recommission as new in HA. It got added as node 198, for a total of 72 nodes in the new Matter server. However, going into devices -> services -> Matter, it's stuck at 71 and node 198 is MIA.

[modern olive]

That was first time. Got a different error on restart. Where do I find the data dir on HAOS?

[silk plover]

First time start is expected to take much longer than normal. Restarts and reboots after that should be much faster.

[grizzled burrow]

Ok let me do a full HAOS VM reboot, and also see if that fixes the missing device.

[grizzled burrow]

Ok did a full (second) HAOS reboot after beta matter. In roughly 5-6 minutes all 71 devices appeared online. The full HAOS reboot also solved the 71 vs 72 devices in settings -> devices -> Matter.

[silk plover]

not going to happen from the Matter Server Webui. https://github.com/matter-js/matterjs-server/issues/94#issuecomment-3787125614

[woven spade]

So far so good on js. I'll start adding in extra thread border routers and see if everything explodes in !!fun!! New ways

[deep drift]

I just triggered the check for updates manually via the HA UI. Now all my devices have the Firmware state „up-to-date“.

[fringe fjord]

Just.Make an addon backup. Then download this. It should include these files

[deep drift]

Eve Thermo 5. Gen bindings to Ikea Timmerflotte are also working. Matter.js has been rock solid over night - no offline devices, no errors in the log. Really awesome, @fringe fjord!

[fringe fjord]

Thanks for the flowers ;-)) (but bindings go completely aside any controller beside the initial setup) 😉

[deep drift]

The flowers are yours anyway.💐

[deep drift]

There is a little layout issue with the button icons in the Matter.js UI of the iOS App. There is not enough space in the bar as it seems.

[fringe fjord]

Please create an issue. Responsiveness. Yeahh

[left rose]

Unfortunately matterjs-server lost all sleepy deveces over night, so I had to navigate my house in the dark this morning...
Collecting the logs and creating an issue.

[fringe fjord]

I thought again and that can mainly happen when devices report data for a cluster but not the cluster descriptor which is weird. So if that happens please provide log - ideally debug level.

[fringe fjord]

Issue with log please.

[left rose]

sigh, is there a way to get to the logs except for the ui download? It does let me scroll back only to the point 8 minutes before it all started....

[fringe fjord]

when downloading logs I gues you can select up to 10k lines to download ... I guess there are other ways but maybe someon ethat knows the addons better should answer that

[left rose]

yepp, and I need let's say 12K of lines.

Another question: From where would I know the exact version I am running if I cannot the beginning of the logs?

[fringe fjord]

tell version best you can. all good

[fringe fjord]

Ok, checked this with @lusty linden and in fact his json file had no nodes in it and also a reset "lastnodeid" to 3 while the backup file was 110kB bigger and had nodes.
I checked and in fact also the Python matter server only uses the backup json if the normal json is missing or "not json parseable". missing nodes lead to an empty python server on start too ... The matterjs-based Matter server works the same (or ok, I will not make it not crash but log a warning on issing nodes but thats still interesting.

[left rose]

Did my best to extract as much logs as I could https://github.com/matter-js/matterjs-server/issues/102
Going to restart the sever now to see, if it brings back the devices (and check the version)

[left rose]

Going to attach the re-start log as well, becuase I see

2026-01-23 08:41:41.063 ERROR  ExchangeManager      •unsecured#99d84518c5faeaa4⇵bb0a✉0d6a678b Error: [flow] Incorrect ack received. Expected 115939201, received: 115939154

and

2026-01-23 08:41:41.094 INFO   PairedNode           @1:1d2 Simple re-establishing session did not work. Reconnect ... [flow] The previous message •unsecured#4a1b9bb57d244b62⇵bb16✉06e9178b has not been acked yet, cannot send a new message.

which might(?) be related...

[fringe fjord]

These things can happen when messages arrive later as expected or such ... in fact in such cases the node should get reconnected which in worst case creates a new session and so is a clean start ... or it should sort out on next usage ... but would need more details andlog to understand real reasons

[left rose]

Now here is an interesting observation: it seems that for some reason all my thread devices seem to have changed from fdf8:cced:c480:1::/64 to fd2e:d139:e716:1::/64 -- and this might be related to the issue. I only noticed now after restarting the matter server that one device did't come up online again, and matter server still believes the device lives in fdf8:cced:c480:1::/64

[fringe fjord]

We first try to connect to the last known addresss! When this is not working then we do mdns dicsovery and try all addresses we get there

Nwo the problem might be that thread mdns is usually veeery long lived (ttl wise) so the question is what comes back on mdns and if thats the correct stuff ...

And if not the correct stuff but wrong comes back ... how should we determine?

[left rose]

But also no, thinking of it, the offline sleepy devices were all pingable under their new address which I got from the device info page

[fringe fjord]

ping !== udp ... i had that fucking often 🙁

especially ping on thread? what is the OTBR really doing when getting a ping?

so debug log should show how natter,js did , which ips we tried, which ips we got back from mdns and such

[left rose]

it should pass it through to the device, I guess

[fringe fjord]

i honestly would doubt that 🙂 but no expert and so also only assumptions

[left rose]

at least that is what I would expect from a router connecting two adjacent networks 🙂

[fringe fjord]

If you like check the debug logs for the IPs we tried ... were the new IPs part of it or only zthe old one=? If yes then some mdns returtned the old and not the new ... then question is why

[left rose]

Yeah, as stated just, there is only one device I could observe showing this behavior, all the other affected offline devices had already had the new IPs (at least thats what the integration showed)

[fringe fjord]

also please add these infos to the issue. I have some other stuff to finish first then I look into it

[left rose]

I can, but I tend to think that this one observation doesn't fall into the "rule of three" category...

I will just copy this dialog into the ticket if OK

[fringe fjord]

Yeah if integration shsows these ips then these came from mdns. So still when devices change matter,js will loose connection and do a new mdns and should then (maybe with a bit delay) find the new IP)

[left rose]

Yeah the drop occured around half past 3 and lasted to now when I restared the server (like 4+ hours)

[fringe fjord]

that should not be the case ... wtf ... yeah will be fun logs reading.

[left rose]

but then the good news is: restating brought all but one (the one lost in the old cidr prefix) back.

[modern olive]

What's explained in that issue makes sense. Where can I find the node information in Home Assistant to tie it to a specific node in Matter?

I have a few nodes that are offline and I'm trying to troubleshoot them. I found one of them in HA, got its MAC address, found it in the DHCP table and tried to ping it. Got a response, so I know the device is online and responding. If I attempt a ping from HA, it says there is no IP for the device. What would be my next step?

[deep drift]

Another UI question: is it only me or is the „jigsaw puzzle piece“ the standard icon of the Matter Server in the sidebar? Should be the Matter logo, shouldn’t it?

[modern olive]

It's been up for 15 minutes since the last restart and no errors in the logs. I'm still trying to troubleshoot a few offline devices, but whatever did come back online did so pretty quickly and seem to be working well.

[left rose]

I looked into the one lost device. It is probably not related to the massive loss of sleepy devices at ~ 3:25am. Going to create another issue.

[fringe fjord]

can you restart the device?

[fringe fjord]

create issue for the addon please (not in the server repo, ut in the HA repo please)

[modern olive]

Yes but it's really awkward because it's an embedded spotlight and I have no physical switch for it as it stands. 😆 I'll try that later but I'd really like to understand better why it doesn't just come back and if there's anything that can be done remotely to fix it.

[deep drift]

I created the issue in HA "Frontend". I hope this is correct: https://github.com/home-assistant/frontend/issues/29145

[fringe fjord]

my guess is more the addon repo, but I think people will tell or move

[left rose]

sigh , actually the lost device is somehow related to the other case. After a device IP address change, at first, matter server and device can perfectly talk. But the new address doesn't get persisted. After a server restart, server cannot resolve new IP, uses old. New IP is mDNS resolvable though. Device tries to talk with matter server (from new ip), somehow fails.
https://github.com/matter-js/matterjs-server/issues/105

[fringe fjord]

I check logs... interestuig because usually we always rediscover anwe via mnds any time

[left rose]

just added the note, that HA's zeroconf does see the device with its correct IP. server keeps stating that "no". (maybe timeout of 3s too short?)

[fringe fjord]

matter.js logs should show what we see there and such ... the 3s are just when IPs are requested, but yes maybe too short

[raven harbor]

hi, may I ask if Bluetooth for commission is working on this version

[left rose]

because I spotted the device trying to talk to the matter server: I wonder if the server regards the device IP as part of its identity, hence communication fails. Espsecially asking because this kind of identity check is already being taken care of in threads SRP/advertising proxy components. The SRP Spec requires this kind of identiy check from an SRP server, so we wouldn't even see the updated IP in mDNS if that ckeck in thread failed. OTOH matter server would have to know about the prefix realms reigned by the pan's border routers in order to delegate trust.

[spark hamlet]

Hmmm, after switching to matter.js (via the Beta check) not a single device is discovered, probably because of 2026-01-23 10:52:32.287 INFO LegacyDataLoader Loaded legacy chip.json with 2 fabric(s)
2026-01-23 10:52:32.295 ERROR LegacyDataLoader Expected exactly one fabric with index 1, found indices: [1, 2]
2026-01-23 10:52:32.297 WARN MatterServer Legacy data error: Expected exactly one fabric with index 1, found indices: [1, 2]

[fringe fjord]

in the addon 8.2.0 likely not because packages were missing. I guess we have addon 8.2.1 monday and there it should work.

[fringe fjord]

Yes already known. I never had that situation, but already fixing it ... maybe I have an update later today. Right now a bit blocked because need to replace a disk in one of my home servers because was failing 🙁 syncing data takes ages

[raven harbor]

thanks for the answer

[fringe fjord]

We do not care about details. we use the operational device matter service and do MDNS discovery. and we try what comes back there

[left rose]

I am talking of this:

2026-01-23 09:53:10.062 INFO   CaseServer           Received pairing request « udp://[fd2e:d139:e716:1:d387:8518:ea3a:df7f]:5540
2026-01-23 09:53:10.070 INFO   ClientNode           server.@1:1c4 is online
2026-01-23 09:53:10.279 INFO   CaseServer           @1:1c4•e3b6 Resumed session with @1:1c4 address: udp://[fd2e:d139:e716:1:d387:8518:ea3a:df7f]:5540 fabric: 3134315bc7a9175a (#1) SII: 3.5s SAI: 2.06s SAT: 1.5s DMRev: 18 IMRev: 12 spec: 0x1040100 maxPaths: 1 CATs:
2026-01-23 09:53:10.681 INFO   ClientSubs~onHandler Ignoring data report for unknown subscription ID 4bbb4509
2026-01-23 09:53:10.975 INFO   Session              •unsecured#a2733b4ab3c1699e Session ended
2026-01-23 09:53:15.171 INFO   PairedNode           @1:1c4 Still not connected after new session establishment, trying to reconnect ...
2026-01-23 09:53:15.172 INFO   PairedNode           @1:1c4 Reconnecting now ...

The device reveals its new IP in the incoming datagram...

[fringe fjord]

Problem with mdns and thread (but mainly for "device went offline") is that thread mdns have ttls from up to 1-2h ... so they are announced very long while device already dead , this then leads to an increasing backoff on reconnect tries

[fringe fjord]

would like to get that as a debug log :-))) it should take up that new session wehen the device "pushes "one

[left rose]

it is part of the debug log in the issue 😄

The same log shows the server trying to resolve the IP, failing, defaulting to the old one. While HA's zeroconf (and local avahi) can resolve the IP.

Now, maybe we're mixing the two issues here:

• https://github.com/matter-js/matterjs-server/issues/102 (mass offline after prefix change)
• https://github.com/matter-js/matterjs-server/issues/105 (server doesn't persist new address between restart for single device)

The above quoted log is from issue #105. In the latter case, server and the single device already recovered. Then I restarted the server in order to bring back the many devices. Server reverted to old IP for specific device, states it cannot resolve new one albeit being discoverable by HA and other tools.

[woven spade]

Two border routers. Restarted matter server and it all just came right back even battery devices. Oh boyyyyy

[fringe fjord]

perfect... I am currently distracted because I need replace 3 nvme's in my server cluster i use here and as it needs to be one broke whoile copying, so getting that back into the needed state is meehhhh 🙂

[fringe fjord]

all good ... the more logs I have the more I can check ... and when infos what the log show from your persoective is also in or pointers like nodes to foxus on or such thats even better

[left rose]

Good luck!

[lusty linden]

Thank you! Replacing the json file with the backup at least makes the nodes show up, but they seem to be offline. No idea what happened there!

[fringe fjord]

hm ... have a log from the server?

[lusty linden]

I’ll restart it in verbose, but the bulbs are actually not on the WiFi anymore

[fringe fjord]

ahahh wait .. ok if they complketely lost network then the server can not do anything. fix that first 🙂

[deep drift]

I just realized that my OTBR log shows today for the first time ever a real timestamp rather than just counting up the time passed since starting the OTBR server. Could this be connected to the Matter.js server, eg mDNS?

[fringe fjord]

I would not know how/why

[deep drift]

I don’t know either where the timestamp suddenly comes from. I didn’t change a thing but installing matter.js. 🤷🏼‍♂️

Which runs rock solid since a day btw.

[deep drift]

My mistake - the time counting in the log accidentally matched the current time. What are the odds. Oops. 🤦🏼‍♂️

[raven harbor]

I just set up a new HAOS instance in PVE with scripts to test the new server. Seems to be stuck in a loop "Error loading server file" when installing the new Matter components

[fringe fjord]

If it is about https://github.com/matter-js/matterjs-server/issues/107 then check your json file. It seems broken. Restore config from backup and update again

I will fix this soon but still the python server also starts up empty in that case. Right?!

[raven harbor]

nah, it is in a loop but thanks, i think that might be my issue

[junior sedge]

if you have a case where the python matter server was working, but the new matter.js server fails to start when loading the config from the python server, then that is a bug in the config import - the reason for the beta is to find issues like that so they can be fixed :)

[fringe fjord]

… and we have that issue that sometimes the python server json file is somehow valid json but misses the nodes.

[raven harbor]

My primary setup upgrades and starts just fine, waiting for devices coming back online. I have dozens of nodes🤦

[silk plover]

First start takes quite the while with large node count. Much faster than prior after that initial start-up/mdns discovery though.

[raven harbor]

2026-01-24 03:03:14.085 ERROR  WebSocketC~erHandler Failed to handle websocket request Node 4 does not exist
  at Function.nodeNotExists (/opt/matterjs/node_modules/@matter-server/ws-controller/src/types/WebSocketMessageTypes.ts:526:16)
  at Nodes.get (/opt/matterjs/node_modules/@matter-server/ws-controller/src/controller/Nodes.ts:49:31)
  at Nodes.clusterClientByIdFor (/opt/matterjs/node_modules/@matter-server/ws-controller/src/controller/Nodes.ts:89:27)
  at Nodes.clusterClientFor (/opt/matterjs/node_modules/@matter-server/ws-controller/src/controller/Nodes.ts:114:21)
  at ControllerCommandHandler.setAclEntry (/opt/matterjs/node_modules/@matter-server/ws-controller/src/controller/ControllerCommandHandler.ts:1067:36)
  at WebSocketControllerHandler.#handleSetAclEntry (/opt/matterjs/node_modules/@matter-server/ws-controller/src/server/WebSocketControllerHandler.ts:880:43)
  at WebSocketControllerHandler.#handleWebSocketRequest (/opt/matterjs/node_modules/@matter-server/ws-controller/src/server/WebSocketControllerHandler.ts:384:41)
  at WebSocket.<anonymous> (/opt/matterjs/node_modules/@matter-server/ws-controller/src/server/WebSocketControllerHandler.ts:242:31)
  at WebSocket.emit (node:events:519:28)
  at Receiver.receiverOnMessage (/opt/matterjs/node_modules/ws/lib/websocket.js:1225:20)

Deleting a Matter Binding related to a removed node won't work. Is that an expected behavior?

[junior sedge]

yeah, depending on how fast the storage on your home assistant machine is, i've seen people report anywhere from 5-30 seconds per device to do the configuration import.

[raven harbor]

Mine's on a PVE running with SSD, it took like 10 mins. That's good enough. My Thread network is a mess with too many interference sources. Wi-Fi and Ethernet devices are faster

[modern olive]

I'm trying to troubleshoot devices that never came back online. How can I map devices on the Matter Server UI to Home Assistant devices?

And in fact, what about the other way around? I have a device that's been duplicated with the new server. The new entities are fine but the old ones are unavailable.

[sleek crane]

Do I need to troubleshoot further or is there an issue with OTBR integration and the new matter server? Never-mind it resolved itself.

[modern olive]

I've got one device that I'm unable to add to Matter. It's a SwitchBot floor lamp, and at the point where it fails, I get this in the logs:

2026-01-24 00:27:14.013 ERROR  WebSocketC~erHandler Failed to handle websocket request [commissionable-device-discovery-failed] No device discovered using identifier {"longDiscriminator":2288}! Please check that the relevant device is online.
  at <anonymous> (/opt/matterjs/node_modules/@matter/protocol/src/peer/ControllerDiscovery.ts:65:23)

[junior sedge]

That probably means that it was unable to resolve a _matterc mdns service corresponding to the provided commissioning code. How are you doing the commissioning?

[raven harbor]

I have one in the box, can help to test if the issue persists

[fringe fjord]

How we should delete a binding when node is deleted? Or what’s exact case? Report issue please if there is one. Maybe there is a big sure but need more context.

[fringe fjord]

The server should not duplicate it unless you actually did a repairing. These check the dashboard of the server especially offline nodes?

[fringe fjord]

They should not be directly connected. Same as with python server

[fringe fjord]

Then the device could not be discovered. So the usual mdns troubleshooting applies?

And as usual for commissioning check which code you use and do not wait too long and such.

[modern olive]

Thanks, that's useful information. I'm commissioning it by pressing 2 of the buttons until the light colours cycle as the manual says, then adding it via the QR code on the HA app on an iPhone.

[modern olive]

I'm literally trying to add a new device. I don't understand the question in this context. Or did you mean to reply to a different message?

[modern olive]

There was no re-pairing. There were 2 nodes on the Matter server though. One was offline, the other wasn't. I deleted the offline node from the Matter Server GUI, at which point HA deleted all entities but kept the device. I then had to delete the device from HA, clicked on Share on the Matter Server, then re-added it in HA and it's been fine since then.

[modern olive]

Where can I find documentation on "the usual mdns troubleshooting"?

[modern olive]

What do you mean by "which code"? How many codes should there be? I usually just scan the QR code.

[brave ginkgo]

I don't think the .js server is picking up my secondary nic.
Does it respect this advanced settings param --primary-interface?
enp1s0 is primary for the server but all my matter devices are on enp3s0.

[deep drift]

Just successfully commissioned my first Matter over Thread device (Ikea Alpstuga air quality sensor) via Matter.js and the HA iOS App, and it went flawlessly. No errors in the log. The firmware update from 1.0.13 to 1.0.15 was also smooth. The known issue with missing time sync is already being addressed. However, there’s a small glitch with updating the “Device info”: even though the firmware update was successful and the correct version appears under “Configuration,” the “Firmware” field in “Device info” still shows the old version. Could this be related to Matter.js?

[fringe fjord]

Yes. But depending on if the device is already commissioned in an ecosystem the QR code on the device itself is not the relevant one anymore. So you need to go to the relevant ecosystem to get the code to use. And when device is new and you run in addon ble is currently not working for the new server, will be fixed next week with addon 8.2.1

[deep drift]

Another finding: the node number the device got is not just +1 of the last node number commissioned, but +10.

[fringe fjord]

Yes for me the primary parameter works. You should see also in log. If not working make debug log sandpits in GitHub issue please.

[fringe fjord]

After the update it should update. No clue when/ how the dashboard here loads the new value. Would be interesting how log looks. What does the dashboard ui shows as version? Will it get updated when doing new interview?

[fringe fjord]

Yes because of reasons we do an initial increase of 10 when we migrate the data. In the end irrelevant. Once we are out of beta node ids will get random anyway so do not rely on that number. It is „just a number“.

[deep drift]

Here is the log part with all firmware update related messages.

[modern olive]

Ok thanks. The device is being reset each time, and indeed has never been successfully added anywhere so the QR code is likely the right one. I'm not sure if it uses BLE but I'll try again after the next update to isolate that.

[fringe fjord]

Please create issue and update there. Here I loose track too fast of that tbh @deep drift

[deep drift]

Ok, in the Matter.js Addon?

[fringe fjord]

If it is in no ip network and was reset then it should require ble. Then you need to assign a ble adapter in settings of addon

[fringe fjord]

No. In the matterjs server repo. See changelog links above (on mobile right now)

[modern olive]

I've added at least 4 other devices (different device types and brands) after factory resetting them with no issues. How does that work?

[modern olive]

Also, the pairing procedure on the phone does move from Connecting to Setting Up before failing. I don't know if this means much. In any case, I've been watching _matterc._udp during that, but nothing showed up for this device.

[fringe fjord]

Factory reset does not automatically mean the devices forget network credentials. So could be the they are still on ip network. Then it works. If also network creds resented then ble is needed. Saw devices with a factory reset and a „real reset“ with difference how long you pressed the button or such. So all a clear „depends“ 🙂

[modern olive]

Ok thanks. I'll check the network while I do one of these resets to see which is the case. I've been watching mDNS announcements while I paired one of such devices, and it's got a few additions and removals. Does this look normal?

10:40:22.180  Add        2  14 local.               _matterc._udp.       4EA0A536CC5D1331
10:40:22.180  Add        2  14 local.               _matter._tcp.        42ED8032F72EA41A-00000000AF145D84
10:40:28.704  Rmv        1  14 local.               _matterc._udp.       4EA0A536CC5D1331
10:40:28.704  Add        2  14 local.               _matterc._udp.       C1CCD4930E04ECD3
10:40:53.645  Add        2  14 local.               _matter._tcp.        8C3B30073DF55484-0000000000000087
10:42:27.034  Rmv        0  14 local.               _matter._tcp.        42ED8032F72EA41A-000000000AC682C8
10:42:27.139  Rmv        0  14 local.               _matter._tcp.        8C3B30073DF55484-0000000000000035
10:43:41.584  Add        2  14 local.               _matter._tcp.        42ED8032F72EA41A-000000000AC682C8
10:43:41.837  Add        2  14 local.               _matter._tcp.        8C3B30073DF55484-0000000000000035

[fringe fjord]

Usually such ui change mean it was seeing it. But also here might depend on app/ecosystem. really hard to say.

[fringe fjord]

There are two matterc ones visible

Others are operational announces from two fabrics

When it is XX..XX-YY..YY then X is fabric Identifier. Y is node ids in the fabric. So likely the 8C.. is HA

[modern olive]

This is a Matter-over-WiFi device. I thought fabrics were a Thread thing. Have I got that wrong?

[fringe fjord]

Fabric is matter and describes basically devices controlled by one controller. In fact devices in one fabric share the same cryptographic root and can talk/interact with each other. So it is matter and unrelated to thread

[modern olive]

Thanks! It's really difficult to find good documentation on this stuff. Why would I have 2 fabrics though? And from what I can tell, both of them are in fact adding the same device, but here's an interesting aspect of it. If I unplug it, the IP in the 42... fabric stops responding, but the other one stays up.

If I remove the device from HA, reset it and re-pair it, both entries are removed and new ones are created, which indicates to me that they are indeed for the same device.

I appreciate that "why would I have 2 fabrics" is likely a tough question to answer, but any indications could help. I really just use HA for Matter, and I'm only adding the device on it.

[fringe fjord]

Do you use the ha app to commission? With Apple or Google? Then likely the second fabric is that other ecosystem and is normal when using via Apple or Google.

[modern olive]

Ok makes sense. I use the HA app on Apple. Is it normal for the device IP on that other fabric to keep responding when the device is unplugged?

[fringe fjord]

No normally not but maybe some mdns repeater answered. In fact mdns entries have a ttl and when you just unplug the device no expiry is sent, meaning it will stay potentially till ttl expires.

[hollow cloak]

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

[left rose]

I also think that this cannot be explained by TTL (and backoff (hopefully)) as

$ sudo ot-ctl 
mdns srvresolver start 3134315BC7A9175A-00000000000001c4 _matter._tcp
Done
mDNS SRV result for 3134315BC7A9175A-00000000000001c4 for _matter._tcp
    host: 5AB109884A33D82B
    port: 5540
    priority: 0
    weight: 0
    ttl: 7200
    if-index: 2

2 hours have passed since 3am yesterday morning.

[fringe fjord]

Faulty repeater?

[left rose]

no. as said, home assistants own zeroconf browser can see the device perfectly.

I have now re-commissioned the device, but unfortunately matter server is missing the "open door" (Falle ziehen) property now. It was there before though and worked with the new matter server already...
So it imported the device correctly, worked correctly before IP address change, then lost it, and now after re-commissioning "open door" is missing.
https://github.com/matter-js/matterjs-server/issues/114

[modern olive]

I don't think it's a repeater, these entries have different IPs - and like I mentioned, it's not the fact that the entries remain on the mDNS cache, but rather that one of the IPs responds to pings, so as far as I can tell, both entries are generated when I pair the device (this is reproducible and happens every time I reset and re-pair it), and one of them remains online and responding even if the device itself is unpowered.

[left rose]

@fringe fjord where do I find the nodes.peerX.endpoints.* files in HA Green, so I can add them to the ticket for the faulty and the good nuki.

[fringe fjord]

Data dir where the json files are is a new sub dir called „server“. In there. But we basically read all data from the device and use them. So only difference that could be is that the data are in the migrated data but no longer when we read. In fact need to be a feature. Or we respond features wrong. So maybe enable debug and ask for a reinterview and Post the log. (There should be a node_updated event in the log)

[left rose]

I will enable debug and do a re-interview for the good (imported) and the bad (freshly commissioned) nuki

[fringe fjord]

by chancel I have annuki too ... let me check

And yes ... seems we convert the featuremap wrong in the websocket reponse. Issue please I will fix it.

[left rose]

https://github.com/matter-js/matterjs-server/issues/114
No, the "converted" device works correctly. The newly commisioned one is the broken one.

[left rose]

Updated the issue with re-interview information

[fringe fjord]

@left rose can you access these peer files ? interesting would be the ciontent of both "Nodes for

nodes.peerXXX.endpoints.Y.257.65532

[left rose]

There must be something inherently wrong: Now after a couple of matter server restarts (for toggling log levels), and after talking to devices for nearly two days correctly at the fd2e: addresses, the matter server pulls the old fdf8 addresses, and more devices lose connection. So it definitively doesn't persist the "last seen working" address between restarts...

[fringe fjord]

same for nodes.peerXXX.endpoints.Y.257.65529

My nuki interestingly reports as featuremap that unbolting is NOT supported but has the commands

[left rose]

No, it seems I cannot access them via the regular ssh add-on. Would have to prepare my HA for root level ssh access.

[fringe fjord]

you can do a backup of the addoing .. they should be in there

[left rose]

can try in an hour or so. need to head to the grocery store quickly.

[fringe fjord]

on my list to check

[fringe fjord]

no stress ... ainly family duties today, so not that much time to look into that in depth before monday ... but all good

[fringe fjord]

just came back from there 🙂

So I verified for my Nuki ... the unbolt feature is not et but the command is provided ... so in fact ... yes the server and HA is reacting right. you could check also in the "longnumber.json" file whats the value of "1/257/65532" of that relevant node. if there a value different to 0 is in then it seems that the feature was set in the past which would describe why it was migrated. if this node was now never connected. to matter,js or such then it could be that there is still that "old" value.
But when the new device is read out officially and the feature buit is missing then thats basically correct ... For me it feeld s like nuki updated something on their firmware and "fucked that up". I can try to ask them

Also: HA is doing "discovery" when a device is added, so I would assume that a disappearing feature would simply leave the "unbold" logic there and not remove it

[fringe fjord]

interesting.. btw also here the storage files are interesting. there are nodes.peerXXX.endpoints.0.commissioning.addresses files ... there the current address should be stored and is initialized from on start ... I will also check the internals for that logic again

[left rose]

might be an explanation. At least I updated the firmware recently. interestingly enough, they claimed to introduce matter 1.4.1 with that release. (I think, it is firmware 5.6.0)

[fringe fjord]

my nuki matter firmware shows "1.4.0" (Smart Lock Pro) ... i have auto update enabled I guess, so no clue ... but matter version claims to be 1.4.0 ... ok update from november ... I update to newest now too ... so when, seems broken since then already ... ok same in newest firmware ... or is there any setting I need to do to get unbolt activated?

[modern olive]

Trying to chase down a device that's been offline since I switched to the new Matter Server. The device is online and responding on the same IP Home Assistant has on file, but it shows as offline on the Matter Server and all entities are unavailable on Home Assistant. Where should I start looking?
The device is still online on the Tapo app.

[fringe fjord]

Debug log ... because if it is shown offline you should see tries to connect it at least every 10 minutes or such ... note down the "hex node id" and then look the log. then you see whats up ... it could get no response or it could get a response but something is not ok

[modern olive]

Thanks, I'll enable debug and keep an eye.

[fringe fjord]

BTW ... not sure if you saw in changelog ... you can also enable debug just temporariely without a restart

Just go to the Matetr server dashboard and use the cog wheel upper right and change loglevel there. SHould have immedisate effect without a restart

I thought especially for such cases where a restart might be too much such a way is efficient

[modern olive]

Yep that's awesome, I had restarted it already but I'm sure it will come in handy soon enough. 🙂

[modern olive]

When you say "note down the hex node id", where exactly do I find that?

[fringe fjord]

from the dashboard UI ...

Node 5000 @1:1388

(the 1388 in that case)

[modern olive]

Right. So literally the node ID converted to hex yeah? I can't find it in the logs, though it's a bit difficult to search for the number 66 as I find lots of false positives. 🙂 Is there a specific string I could search in the logs to help find it?

[fringe fjord]

you should be able to use "@1:1388" to search because thats the "peer address" which is the internal unique peer id in. the server

[modern olive]

Thanks, that's helpful. It's failing to discover it

2026-01-24 17:16:22.740 INFO   PeerSet              Resume failed, remove all sessions for @1:66
2026-01-24 17:16:22.740 INFO   PairedNode           @1:66 Error waiting for device rediscovery, retrying [pair-retransmission-limit-reached] Failed to connect on any discovered server
  at Function.iterateServerAddresses (/opt/matterjs/node_modules/@matter/protocol/src/peer/ControllerDiscovery.ts:237:23)
  at async <anonymous> (/opt/matterjs/node_modules/@matter/protocol/src/peer/PeerSet.ts:608:32)
2026-01-24 17:16:22.741 INFO   PairedNode           @1:66 Reconnecting in 10m ...

[fringe fjord]

No ... it discovered ... but "Failed to connect on any discovered server" ! And this means either not reached (maybe the invalid IPs tried issue that @left rose already saw ... or device was answering but declining. debug log would show that

[modern olive]

This is the debug log. Doesn't seem to have anything else on it.

[fringe fjord]

Then this was before that ... depending on debvcie and network type could ba 1-2 mins even for that connection tries ...

look out for lines from CaseClient ...

[sturdy ivy]

I currently have about 40 Matter devices across 2 Thread networks, mostly Nanoleaf bubs and Tado thermostats, intending ultimately to expand to around 60 devices of different types. I have had a lot of problems with stability with the Python server, with numerous timeouts and devices going offline. In particular, switching light groups (chandeliers, downlights etc) has been very unreliable, often resulting in the whole Thread network hanging for several minutes. The new matter.js beta server has improved stability dramatically. There are now almost no timeouts in the logs, and switching is reliable for small groups of lights. There are still issues with switching large groups, but I am working on a script to throttle these sorts of operations. Overall, a massive improvement though. Thank you!

One oddity I’ve noticed. There are numerous log entries looking like this, although they don’t seem to cause noticeable symptoms:

2026-01-24 06:12:28.640 INFO ClientSubs~onHandler Ignoring data report for unknown subscription ID 64a3059e

2026-01-24 06:12:33.918 INFO ClientSubs~onHandler Ignoring data report for unknown subscription ID 1b11b158

2026-01-24 06:12:34.819 INFO ClientSubs~onHandler Ignoring data report for unknown subscription ID 45090504

2026-01-24 06:12:43.819 INFO ClientSubs~onHandler Ignoring data report for unknown subscription ID 6d12a740

[fringe fjord]

Could it be that these reports come together with you "wire-powering-on" (so real power loss and restore) or a device? If yes I know what it is. 🙂 And yes then you can not really do that much against it but it also does not hurt too much.
I will improve logging for next version so that we might see which node was causing it.

[sturdy ivy]

That could be it. I've still got physical light switches which inevitably get switched sometimes. I'll keep an eye on it and see if I can confirm this once everyone else has gone to bed!

[modern olive]

5 min after those messages appearing again, there are no other logs for that node ID, and no logs whatsoever that match CaseClient.

[fringe fjord]

I said you need to look "before" these lines ...because these lines mean" we got dome IPs and tried all and are now done" .. so when it is it comes before them ... but effectively because we try every 10 mins they should come again later 🙂

[modern olive]

Yeah nothing before either as far as I can tell. I'll keep an eye for a few more minutes until the 10 m cycle happens again.

[fringe fjord]

and just to be sure ... you see "DEBUG" messages? 🙂

[modern olive]

Yep 🙂

[fringe fjord]

If you are unsure just let it run 30-60mins and upload the log as issue and I check

[modern olive]

Can do, but 10000 lines is the most I can download via HA web UI, and it covers about 5-7 min only. Where can I get full logs?

[fringe fjord]

maybe one of the pro's here can answer this ... I can't till now ... would also be interested

[modern olive]

HAOS makes a lot of things easier but how I'd love to be able to easily SSH into the host...

[left rose]

unbolt is always available/on in the Nuki app by default. And I don't see anything to activate it separately for matter there.

[keen breach]

Adding Matter device ignores the HA border router, insists on using Nest border router. Hubs are unplugged and Google Home disabled. Any suggestions?

[left rose]

Groceries took a bit longer...
It's getting strange and stranger. Now there is a dangling entity for the re-commissioned nuki, but it cannot be deleted from the UI...
Updated the issue.

[modern olive]

I've got a bit over an hour of debug logs, would you like me to send it to you? How would it be best? It's a bit under 2 Mb gzipped.

[left rose]

Created another issue for "server cannot reconceil with new IP address after restart / doesn't persist new IP" this time with DEBUG enabled.
https://github.com/matter-js/matterjs-server/issues/115
Funny thing seems to be: When can resolve to the new device address, it simply seems to ignore that, but when it cannot resolve to the new address, it tries to ping the device on the old address (getting no answer of course).

[grizzled burrow]

Just wanted to say that for my 72 Thread/Matter devices, the new beta is fantastic! For me the < 5 minute restart time for Matter is a dramatic difference..down from 15-45 minutes. I look forward to even more stability improvements!

[lusty linden]

On running docker images I see that the Matter server is nearly 1GB. Is that known and necessary? Should I look into what can be trimmed?

homeassistant/aarch64-addon-matter-server 8.2.0 52c7dc90dc31 2 days ago 944MB

[grizzled burrow]

701MB for me

[lusty linden]

You have the new js server installed?

[grizzled burrow]

I do

[junior sedge]

is that

[fringe fjord]

ZIP it and create GitHub issue please

[fringe fjord]

No it is not. You need to select the right lock type in Nuki app. That door type screen (German) and the one with white border needs to be it. So I got unbolt feature activated.

[fringe fjord]

That’s planned. When we can not resolve we still try old address before we get a new one or it works.

[fringe fjord]

Is that ram or disk? Ram sounds a bit much. But maybe there is something. Would be interesting if it increases or stays stable and such. Any over-time-charts?

[left rose]

Oh wow! I have all my nuki configured to "Drücker", and all but one show "Tür öffnen" in the nuki app UI! But I will try of course, because it makes sense. Do I have to re-pair the lock to matter in order to get that function back?

[fringe fjord]

No. Just change setting. Try if it updates - worst case needs a new interview.

[left rose]

OK, hang on...

[fringe fjord]

Or worst case server restart.

[lusty linden]

Disk

[left rose]

Have you seen the updated logs? Because it shows some interaction with the new IP, but then seems to ignore, trys to dicover again, fails, and defaults to old address,

[fringe fjord]

Currently we add nodejs and the npm packages to the image when you activate beta - plus all python is also there.
When beta is over and the python server is gone should get smaller again. For now needed that way.

[fringe fjord]

Not yet had time to. But all logs help. Together we can come to the bottom of that I am sure.

[fringe fjord]

But yeah will be interesting. Were the interaction with the new ip successful or not. But Yeahh will see in logs

[left rose]

It didn't need a re-interview! Wow, good catch -- thanks a lot! Updating config for all locks now. Wonder where that got out of sync...
Closing the ticket, because the ones I opened later have better information regarding the IP address confusion.

[lusty linden]

I wonder if the python or js version takes up more space

[fringe fjord]

I assume Js needs more in the end. But I did not really compared.

[fringe fjord]

Yeah inward also confused and figured that out for my test lock I have

[lusty linden]

My HA green has 32 gb of storage total, so it adds up 😅

[modern olive]

Created issue and attached logs. Thanks! https://github.com/matter-js/matterjs-server/issues/117

[hollow cloak]

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

[timid oxide]

Running latest beta without any issues.

[fringe fjord]

Hm ... thats not how it is designed, so a debug log would be great. When eve devices support the new matter attributes we should not poll them. so would be interesting to see for which nodes we get normal infos and for which we poll. Please let debug log run - fine with all 6 devices and then add that debug log as isue. You can also start without debug log and then enable it via the matterserver dashboard UI

@left rose I did some code digging and yeah I found two race conditions. Trying to write tests for that because repro is hard but yes it seems that

• When a new session was pushed by the device and a discovery was running this was not stopped leading to having a new session but not using it but waiting for rediscovery and potentially trying old addresses in that timeframe
• When a new address was discovered there were cases that did not updated the stored address for next start, so making issues there
  These two together could then lead to fun situations

I still need to look through some of your logs but what I see and also what you posted could point into this direction. I expect to have an updated version by tomorrow because need to manually test a lot to ensure it did not broke other things.

[left rose]

Sounds totally reasonable. And yeah race conditions... Probably also takes a larger testbed to get triggered. 😁 -- good there are some here in the alphabeta community.
Happy to try out once you release something. I can always force an onmesh prefix change by enforcing leadership on different BRs.

[willow crescent]

I'm new to matter and just have setup matter and thread om my HA system. I run HA as a docker container and then also matter and OTBR as docker containers. Using this image, ghcr.io/matter-js/matterjs-server:latest. First i ran the python-matter-server image, and now switched to the matter js server. I don't have any devices so for me I can delete the old volume, but when starting the new container i get these log error:

2026-01-25 13:25:54.300 INFO LegacyDataLoader Loaded legacy chip.json with 1 fabric(s)
2026-01-25 13:25:54.317 WARN LegacyDataLoader Error loading server file 17065309897990109356.json: TypeError: Cannot convert undefined or null to object
TypeError: Cannot convert undefined or null to object
at Function.keys (<anonymous>)
at start (/app/node_modules/matter-server/src/MatterServer.ts:132:38)

So my question will be, should this work? and is the matter js image ready to be integrated i same way as the puthon matter server was?

[silk plover]

Anecdotally I and several others have noticed having several eve energy plugins is big negative to thread network reliability. Removed/replaced all of mine and have had little to no repeat complete degradation since.

[left rose]

still have a few of these. In general I am looking not to buy anything <1.4 these days. Be it thread or matter. I generally find devices with a more recent stack more stable and beneficial to the whole setup. Unfortunately there is also a large base of nanoleaf bulbs in my home...

[silk plover]

Agree, switched to Aqara T2 from Nanoleaf first. Some improvement, specifically the T2 bulbs would not fall off and fail to respond individually like the Nanoleaf, but was still experiencing complete thread network decompensation every so often. That has stopped (so far 🤞) since replacing Eve energy with Onvis matter 1.4/thread 1.4 smart plugs.

[grizzled burrow]

Which Onvis smart plug is that? I'm looking in the CSA database and I only see the unreleased Onvis T20. Is it the US version?

[silk plover]

https://www.amazon.com/Onvis-Matter-Thread-Outlet-Google/dp/B0FRFPV5FK

This one, update to the Onvis 1.4.0 firmware uses Matter OTA in your fave ecosystem and its matter 1.4/thread 1.4 compliant. Yes US version.

[grizzled burrow]

Perfect! Just ordered 3 to replace the crappy Eve smart plugs.

[grizzled burrow]

I found a problem with the beta server. I had an Aqara P2 motion sensor that I was using, had on the shelf a while, and repurposed in another location after a factory reset. HA had a memory of the prior instance (even though it has a new Matter node number), and now I have duplicate sensors/config/diag items. One set is online, and the other is 'Unavailable' but I'm unable to delete them. I did a full HAOS reboot, and they are still in limbo.

[reef crane]

Here is an add-on log taken with only 2 Eve Energy Outlets running for about 12 hours. The system is stable but heavily loaded with traffic like I described in the original post. I have a few other Thread devices dropping off and coming back (likely because of the excess traffic), about 90% of my Thread devices are online.

The DebugLog includes some earlier times in the day when I had some troubled 3rd reality Motion sensors chattering. you should look at the last 1.5 hours of the log. I removed the troubled motion sensors, so the whole system quieted down a lot. There is no time overlap between the debug log and the add-on log, but the problems shown in the add-on log were also happening during the period of the debug log.

[deep drift]

Really? I have only Eve Energy plugs (7x EU version) in my Thread mesh and they have been rock solid ever since. I recently added two Ikea Alpstuga, which also act as routing end devices and there hasn't been any degradation of the mesh stability yet (all in all 23 Thread devices).

[reef crane]

Every Thread device came back online and has stayed up for several hours. I love stability!

Let me know if you need add-on and debug logs that are time synced together.

Looking forward to your next version. Would love to plug my 6 Eve Energy back in.

[silk plover]

yes, at least among a couple of us with 30-50 inovelli matter over thread light switches on Matter 1.4 firmware.

things got better when I removed the HA OTBR and zbt-1, using basically only Apple TBR and a aqara hub to get access to aqara Matter over thread device setting, stayed there for a week or 2 at a time to test, but removing the Eve Smartplugs was the last step and it helped significantly as well.

[quick leaf]

How do you see which Matter version an accessory uses? For Thread I use Home Assistant’s built in mDNS discovery.

[reef smelt]

I believe that it's called "Firmware" in HA on the device page:

it would be nice if it was visible in the matter-server dedicated web page

[left rose]

It is not. Not in general at least. Firmware numbers are not in in any way related to the matter version. A few vendors do it though.

[heavy dirge]

Find your device here to get the Matter version: Certified Products Search

Example:

[reef smelt]

Fair enough. It seems like Nuki does, as they just updated the firmware to 5.6.10, which enabled Matter v1.4.1

[grizzled burrow]

Is there a different local firmware update procedure for the new matter server? I dropped an Inovelli JSON file in the core_matter_server/updates directory like the old server, but a restart did not pick it up.

[silk plover]

#new-matter-server message

Copy then actual OTA file to addon_configs/core_matter_server/updates and restart the matter server should do per the linked post.

[tranquil agate]

I updated after trying to find a solution to my Thread related issues. Upgrade went realy smooth and easy. Thanks for that and all the hard work!

Unfortunately, my Lidl Smart Plugs (basically Tuya) sensors all just read '0'. I am able to switch them, though. Log is attached. Is there something I should do about this, or should I file a bug?

Now with the logs 🤦🏻‍♂️

I guess it's got something to do with these lines:
2026-01-27 09:56:42.152 ERROR AttributeDataDecoder Error decoding attribute 1/0x125dfc11/0x125d0022: Unexpected type 4, was expecting 10.
2026-01-27 09:56:42.153 ERROR AttributeDataDecoder Error decoding attribute 1/0x125dfc11/0x125d0023: Unexpected type 4, was expecting 10.
2026-01-27 09:56:42.154 ERROR AttributeDataDecoder Error decoding attribute 1/0x125dfc11/0x125d0024: Unexpected type 4, was expecting 10.

[fringe fjord]

There are currently issues in some reconnection cases we work on. Also polling should only happen when devices do not have the matter clusters and in my test sthat work because the others are scipped. To validate please craete a debuic level log and upload in issue. YOyu can also enable debug log temperariely via the dashboard of the matter server so you can just switch loglevels during runtime

[fringe fjord]

The next version will start up also when "Nodes" are not existing and log a warning. So wait for that or yes just deplete the "longnumber.json" and should also be fine when you have nothing in there right now

[fringe fjord]

Thats not an issue of the beta server this is normal behavior. matter commissions unique node-ids. So when you reset a devcie and not delete it also from the server then the old node id keeps there and when you then recommission the devcie jsut gets a new node id. there is no reliable way to map that between "recommissioning". You can go into the matter server dashboard and delete the node there and that should remove it in HA too.

[fringe fjord]

Please create an GitHub issue as requested ... it does not make senser to handle that here and I will loose track faster than you can write the next post 🙂 also: thats no debug log 🙂
(and the decode error will be fixed in the next version too)

[fringe fjord]

Jo this is really the devcie firmware. The device matter version is currently not exposed in HA. This could be determined accurately starting Matter 1.3 .. for before there are ways to guess if it is 1.0/1.1 or 1.2

[fringe fjord]

ALso not really because thats just the certified version. Yes in a perfect world this matches the latest firmware ... but .....

[fringe fjord]

How about reading Github infos? 🙂 https://github.com/matter-js/matterjs-server?tab=readme-ov-file#importing-custom-ota-firmware-files download the file and put in the file. no json needed. BUT you need to have enabled the test dcl flag

[fringe fjord]

seems the infos in the python server about datatypes were wrong. I will fix for next version

[tranquil agate]

Thanks for fixing this 🙏

Do you want me to report this as a bug on github, or is there no need for that anymore?

[fringe fjord]

there is already an issue on Github, so "covered". thanks

[jolly quest]

hey, I am new to Thread/Matter and tried (without good success) to pair my first Matter sensor with HA in a proxmox VM. I might have IPv6 problems.

Is there a more specific guide than https://www.home-assistant.io/integrations/thread ?

(the speicifc problem I am currently having is that my Android app tries to use a old ha-thread-xxxx network, that I had created hours agof, but delete since then....)

[silk plover]

Wrong channel. Repost in https://discord.com/channels/330944238910963714/1284966617670881350 or https://discord.com/channels/330944238910963714/1458060881970856051

[fringe fjord]

Alright, we just now released addon 8.2.1 which should ensure all packages are in docker to correctly allow enabling BLE with the new server. There is not yet a server update ( I finalize that currently, likely tomorrow).

[deep drift]

Just installed it. Lots of gyp errors.

[fringe fjord]

grmppfff 8.2.2 on it's way 🙂

[deep drift]

I have to add that I don't have a BLE adapter connected to my Proxmox VM, only ZBT-1.

[fringe fjord]

all good it is not even coming that far 🙂

[crisp raven]

the server is still unable to startup, not sure if I should report a bug?

[18:29:21] INFO: Using 'enp0s25' as primary network interface.
[18:29:21] INFO: Successfully send discovery information to Home Assistant.
Warning: --log-level-sdk is deprecated and no longer supported. This option will be ignored.
Warning: --paa-root-cert-dir is deprecated and no longer supported. This option will be ignored.
2026-01-27 18:29:24.588 INFO   LegacyDataLoader     Loaded legacy chip.json with 1 fabric(s)
2026-01-27 18:29:24.603 WARN   LegacyDataLoader     Error loading server file 11623776939177563688.json: TypeError: Cannot convert undefined or null to object
TypeError: Cannot convert undefined or null to object
    at Function.keys (<anonymous>)
    at start (/opt/matterjs/node_modules/matter-server/src/MatterServer.ts:132:38)
[17:29:24] INFO: matter-server service exited with code 0 (by signal 0).
[18:29:25] INFO: Starting Matter Server...

(also note the weird timezone on one line)

[junior sedge]

try reverting to a backup and make sure that the python matter server (non-beta) start up properly.

if switching back to the beta fails again in the same way, then file an issue and include the json file that it's complaining about.

[crisp raven]

Reverting to a backup would mean losing recent data. Since I don't have any actual Matter device, is there another way?

glad to help if there is

[reef smelt]

you can just restore a backup from that addon

if you don't actually have Matter devices, there's nothing to lose

[junior sedge]

... has the use case of someone installing python-matter-server but never adding any matter devices, then switching to the matter.js server been tested? :)

[fringe fjord]

Yes, this issue is also still not solved (and noone told it would) 🙂 I hope to have annew version tomororow that fixes this.

[fringe fjord]

@crisp raven if it fails this way then your python file is strange OR you have two fabrics and we try to start the wrong one. If you like do a backup or send me the latest one and I check. YOu can send it here as PN and I validate that next version works as expectd

[crisp raven]

no rush. Just wanted to report it, so I'm happy to know it's already identified.

[fringe fjord]

bit the backup would still be interesting to validate next version works ... so I'm up for it ... just send as PN if you like

[crisp raven]

sure, can you hold my hand through the steps? Not really used to this.

[fringe fjord]

no because I never hadd such a case 🙂 but all good next version fixes that

[fringe fjord]

I can also only slightly ... still learing these HA thing. Basically no need to create a new backup. just grab the latest one and download and send me. Maybe one of the others can guide you? Would be great (and I also learn something)

[crisp raven]

ok I have it, I'll send it to you @fringe fjord

[quick leaf]

Is this something that could be implemented in the the Matter Server add-on?

[grizzled burrow]

Updated 30 Inovelli switches via local firmware update (1.1.2) and all went perfectly well! Would love a sequential updater to automate it all, but manual worked too.

[sullen bobcat]

hello! the second channel appears read-only to me. any idea why? I'd really like to be able to post there 🙂

[candid pollen]

you need to self assign the developer role in the channels and roles section.

[sullen bobcat]

ah, thank you!

[candid pollen]

no worries, that projects channel is for development discussion though, if you have questions about using then the standard support channel is better.

[sullen bobcat]

understandable, I do think it might be of interest to thread/otbr dev folks as it's buggy behavior with official, fresh out of the box gear and software, but if it's not the right channel and they say so I'm happy to stick with the support channel 🙂

for reference: #1458060881970856051 message

[fringe fjord]

If interesting open an issue as feature request (I will reflag the issue). Matter.js basically supports auto-update and queued-update capabilities. It is basically a ui and user experience topic.

[quick leaf]

Where would I post a feature request for this? Directly in matterjs-server or in Home Assistant feature requests?

[fringe fjord]

The question is what exactly you would do with this information? Right now it shoukd not matter which matter verdsions devices are unless you want to hammer on the manufacturer to release updates 🙂

[quick leaf]

Hmm okay. Guess it’s just nice to know. 🤷

[deep drift]

gyp errors are gone. 👍

[fringe fjord]

lol, fast than I can type 🙂

Soo guys, Addon 8.2.2 is released which should enable the BLE commissioning when a BLE adapter is configured

[deep drift]

thats me 😅

[fringe fjord]

New version od the matter.js is alsoin finalization and should come within the next hour hopefully

[left rose]

OK, so has anyone already updated their add-on to matterjs-server 0.3.0 on HAOS?

[fringe fjord]

Soo .... before Boris finish typing ... Version 0.3.0 (soon 0.3.1) is released 🙂

shit, too slow

So, guys, beside the 8.2.2 addon we now also have the Sevrer in version 0.3.1 (ok, now .0, but in some minutes .1) ... changelog see https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md

For all Addon-Users this shoukd work seamlessly and the "Breaking" note can be ignored because thats only relevant f you run the server standalone or via docker directly or such.

Included are many fixes, especially around mDNS, discovery and IP changes that hopefully fix the issues that were seen.
It now also starts with empty nodes and also with multiple fabrics in the migrated configuration.

I will now close all issues addressed by this version. If the issue persists please open a new issue with new Logs to not mix everything together.

So, have fun!

[rugged orbit]

hey, why is make and gcc needed for the Matter server?

[left rose]

Update to 0.3.0 went through smoothly here. I will be able to induce a thread net split after business hours...

[deep drift]

Update to 0.3.1 was flawless, the reconnection of all devices super fast. No errors.

[silk plover]

Second @deep drift report on 0.3.1. updated with add-on 8.2.2

[silk plover]

Would really like to be able to queued-update capability as well, like @grizzled burrow updated 32 inovelli switches with beta firmware manually one by one, which worked well. Likely too early to say, Would or could this sequential/queued update work via the standard HA update notification option on the HA Settings page or would it be a new Update page in the Matter-server UI?

[grizzled burrow]

I’d prefer a larger update queue mechanism that covers ESPhome, Z2M and Matter.

[silk plover]

Agree 100%, but suspect that's well outside the scope of what @fringe fjord can accomplish alone and unlikely from the broader HA devs in the short to intermediate term.

[junior sedge]

I'm hoping that at some point matter.js will be able to do BLE commissioning via HA BT proxies, that would be quite neat :)

[fringe fjord]

The matter server also allows to do commissioning via BLE. BLE functionality is using a native mnodejs dependency which needs make and gcc/g++ to build during install.

[junior sedge]

yeah - ideally it wouldn't be using a native nodejs dependency, but rather talking to the ha integration which would handle doing bluetooth stuff on its behalf.

[fringe fjord]

That would be a topic to discuss. In theory we could have a "auto update" entity that you could set and if it is set updates happen automatcially when becoming available somewhen - makes sense for some device types, maybe not for sockets 😉 ... Queued updates I would not see in HA, but could be added to the dashboard ... but if we have the "auto update or manual update flag" the question is if thats needed at all anymore ... but I guess thats a discussion then when we look into this. lets get the server stable and "out of beta" first

[brave ginkgo]

where is the best place to upload logs to help with debugging? 8.2.2 python works for me, but switching to the beta has errors

[fringe fjord]

We had a first discussion somewhen on that ... I fI know the protocol then it is "just code" because in matter.js these network variants are relatively good abstracted, so replacing Noble with something like that would mainly be implementing a variant of the two Noble classes in https://github.com/matter-js/matter.js/tree/main/packages/nodejs-ble/src ... so if anyone want to do a PR ... :-))

[fringe fjord]

the python matter server allowed it natively, so to replace it we need to be feature par ...

[junior sedge]

yeah, native always felt like a hack in the python server, since it required a dedicated bt that wasn't being used by home assistant itself.

but was better than nothing i guess :)

[reef smelt]

I've built out a bunch of BT proxies (it helps that even the ZWA-2 can be configured as one), so yeah, they need to be used for this

[fringe fjord]

Open an issue at https://github.com/matter-js/matterjs-server/issues and attach logs, ideally debug level. If the errors are about "after initial migration errors are logged about devices that can not be connected directly thats "Normal" and can be ignored. because we do not have collected "former IP addresses", we need to discover yall devices once ... so during this there could be strange logs yes ... but post the log and describe your issues and we see

[brave ginkgo]

Cool. will report there. 👍
And to have here to, here's a little snapshot. Everything comes online and then it seems one thing errors and it brings down all the others.

[junior sedge]

Is that on HAOS? Or running in a container?

[brave ginkgo]

in a vm

haos 17.0

[silk plover]

Agree waiting on a stable release is the best course. The question with auto-update will be whether it is a one after another sequential process with the ability to retry because updating numerous thread devices results in at least some degree thread network instability when thread routers are updated.

[fringe fjord]

I will also comment that on the issue ... you chose a network interface as primary which does not exist? Please verify that the network interface is the one to be used and also existing where the server runs. In worst case leave the setting empty to get auto chosen

[fringe fjord]

it would be queued for sure

[fringe fjord]

And to clarify the log: All gets initialized but then it crashes because MDNS network configuration is somehow borked, so "Nothing can come online" 🙂

And yes there might be a chance that the python server ignored such wrong config, but matter.js seems to require it right ... sooo not sure if I want to adjust that or better havit fail to show such a mis-configuration.

[left rose]

So I initiated a prefix change. For that I took the leader router offline. 85% of the devices came back within 5 minutes. I guess after that some sort of Back-off kicked in. We are now at about 40 minutes, and 10% offline devices. I did a few samples, and they show one of two behaviors:
a) Not resolvable via mDNS. This might be, because the leader also had the SRP server role, the other routers are 1.3, and that is just a totally different kind of hell...
b) Not pingable, so simply not back online in thread
In both cases the matter server would just not be to blame.

For the reverse prefix change I would create a debug log.

One thing I observe though: For devices that cannot be resolved after an IP address change, the device info page will never show the device info section. That makes it hard to know the node ids for the affected devices.

[fringe fjord]

That should good ... but still also not that good ... but yeahh such stuff is hard to blme anyone ... i still wonder why MDNS is not working there? Would nbe interesting ... usually thread MDNS is an hour or two, TTL wise .. would be interesting if somewhej after that MDNS would iick in again?

For the device info section what exactly you mean? There where the IPs are usually listed?

[left rose]

Exactly. Normally, if the IP cannot be determined the background process will time out, and then the section appears, and I can know everything but the IP address. But not in this case.

it just stays like this forever:

[fringe fjord]

ok, but what it shows beside IPs if we have none?

[fringe fjord]

Also for a) is it still resolving the old IPs or no IPs at all? If you tztry to look the records up on another host using low level mdns tools would that return anything or also not? tell me whicz tool (dns-sd or avahi-browse) and I tell you the cli if needed

[left rose]

For me most importantly the node ID because then I can create an mDNS lookup manually.

[fringe fjord]

what the matterserver does on @1:a (should be node 10 id i guess) .. what you see there in the logs?

[left rose]

It doesn't resolve at all. And the root cause for this is probably that the take-down of the leader in this case also initiated an SRP server role re-negotiation. And that's something that is unfortunately a bit broken in openthread 1.3 when more then two BRs are in place, especially with mDNSResponder on the platform level. It's an intricacy burried in the SRP specification.
So this case is just not solvable in matterjs. It can take 2h for the device, before it can re-register on one of the BRs (using SRP) to publish the host or service in mDNS.

[fringe fjord]

ok, "coool" (for me :-))) ... not for your network tho) 😉 thanks for the information and deep analysis

But that sounds very promising

[left rose]

When I do the reverse prefix change (taking the 1.4 BR back online and making it the leader), the SRP servers should stay stable. That will reduce the complexity, and that will be the time when I enable debug logging on the matter server.

[fringe fjord]

@left rose for above:
in your original message you differentiated between "not resolvable" and "not pingable". If it's "pingable" then presumably you knows its address from when it was (formerly) resolvable. So matterjs server should too. ... If it's pingable, is it at an address we previously talked to or a new address?

[heady token]

@fringe fjord busy man=doing good 😉

keep it up

[fringe fjord]

LOL. I am very done … I read first your name as mDNS and asked myself if someone want to trick me ;-))

[left rose]

The ones that are resolvable and pingable were all collected by matter server with their new address. That's the best outcome we could have hoped for.
For the ones that are not pingable, they also don't show in matter server, but yes, they have new IP addresses resolvable in mDNS and these also show in the device info section of the device info page. (that's case b)
Since I don't have any matter logs for this first split I can't tell if the device tries to talk to the matter server though. We will see later when I do the reverse prefix change.

For case a) the ones that cannot resolve as discussed above, either the device info section on the device info page does not show OR it shows with the old address (as that's the persisted one, and a new one isn't known)

[fringe fjord]

Also no Info Level log?

[left rose]

no, unfortunately not. I realized that only way too late.

[fringe fjord]

But if they are mdns resolvable then we should see them … hm … ok depends … Yeahh

[heady token]

dont panic lol

[left rose]

I actually think that it's a good thing, because we would have had these SRP server cases to be sorted out. We will get cleaner logs tailored to the original issue in the second phase.

[left rose]

but we do 🙂 -- at least they show up with the new addresses in the device info. The fact that the devices are not reachable on the network layer can't be blamed to the matter server, I think. That's maybe something to be addressed in network-manager? Idk, honestly...

[fringe fjord]

That’s exactly the interesting question if they are reachable.

Do you have it still running and still some offline?

You can open the dashboard ui and in upper right you have a settings icon where you can temporary adjust the loglevel. Set to debug and let runs bit … min 15 mins and check if anything for these offline. Nodes appear

[left rose]

I am going to do it, but the situation has changed meanwhile. Going to create a ticket listing all seven still matter-offline devices with their node IDs. Four of them have an IP. Of those all are pingable by now.

[fringe fjord]

Yeahh I think i have one idea ... maybe wait till the switch back till new version tomorrow

ok no bullshit ... but a debug log would be cool

whe there are open queries we should query in intervals and log (debug)should have

Sending ${queries.length} query records for ${this.#activeAnnounceQueries.size} queries with ${answers.length} known answers. Re-Announce in ${Duration.format(this.#nextAnnounceInterval)}

We start at 1.5s and double on each iteration

[rugged orbit]

this seems like something that should be pre built rather than taking on large dependencies

or at least be optional

[left rose]

The situation changed while I was writing, I am just now creating an issue with logs...

[fringe fjord]

Yes it will as soon as we switch to the matterjs servr as onlx option after the beta. right now we decided to do the dynamic install to save disk spae for the users that do not test the beta

[fringe fjord]

for the good or bad? ;-)) But yeahh interested. I validated that we do discovery queries in any case ... so either thats not responded (which would be strange) or we get IPs but can not send UDP to .. or ... lets see

[rugged orbit]

I am not sure how I ended up as a beta tester! But thanks for the info

[fringe fjord]

🙂 surprise 🙂

[left rose]

The good thing is that the matter server somehow caught the new IPs, even in cases that are currently not resolvable... They still show offline.

Ticket: https://github.com/matter-js/matterjs-server/issues/148

[fringe fjord]

commenting

responded

[left rose]

How would I send a command for such a node, when it's not in the logs and offline in the controller? I can power-cycle some of them though

[fringe fjord]

the matter dashboard can do some basic commands in the mean time and does not care of offline. clikc through to onoff cluster and see "devcie commands".

that dialog should not care if offline or not 🙂

[left rose]

But is not displayed for the offline device...

For online sister it is very well displayed.

[fringe fjord]

grrmppff ... seems I was too perfectionist .... can you trigger a command from HA? or also "blocked" because node offline?

[left rose]

What do you mean by "trigger a command from HA" -- I cannot interact with entities of offline nodes in the UI.

[fringe fjord]

ahhh can you klick interview?

try that

or share

are these available?

innfact they all do real interactions

[left rose]

The Interview seems to trigger something

Going to upload the logs

https://github.com/matter-js/matterjs-server/issues/148#issuecomment-3813967601

You tell me when I can start the second prefix change with logs for the whole transition 😄
(I am not sure what would fit in 10000 lines though)

[fringe fjord]

Ok, We had a session which would describe why we did not queried, but basically had no subscriptio which is strange.
So what happened now is that we tried sending the "Read" request using that session which failed nd so we also sent a discovery 3134315BC7A9175A-0000000000000203 and got the IP fd2e:d139:e716:1:59e4:6201:8372:1a7e ... but it was no IP change so we must have had this IP already before

Ok then this communication failed ... all ok ...

Interesting is

2026-01-28 22:11:47.505 INFO PairedNode @1:203 Session disconnected while Node is WaitingForDeviceDiscovery

Which means that a part of the system wait waiting for it ... so there must be a strange situation happend

The read went tghrough partially but in the end unsuccessfull because ended in resubmissions again. But THEN n next try wo got this

2026-01-28 22:11:47.964 INFO CommissioningClient Operational address changed for @1:203 from udp://[fd2e:d139:e716:1:59e4:6201:8372:1a7e]:5540 to udp://[fd37:5639:3134:1:65:4da6:2eac:219b]:5540

I just have no clue where it came from. 🙁

something strange happened in the state engine that handles the reconnections and such ... but I do not get what

Ok, I will dig throgh that log again but so or so it seems all a bit unstable to these devices

[left rose]

I can just restart the matter server to see if it resolves, but then the debuggable state is lost.

the question is if this is enough logs to debug for you. The best I can probably do is invoke an interview on all of these devices.

[fringe fjord]

you can try to interview all others too then lat that run 20 mins , grab the log and do a restart if they still did not updated - and ideally set loglevel to debug and tehn we see how it comes up

update all that ... i check tomorrow

[earnest juniper]

I bought two new devices or whatever they may be called (power oulet and a motion sensor), one being Zigbee and I later noticed the other was Matter. Do I need to run a whole other installation of HA to use Matter, or how does that work?

[analog sphinx]

It should just work ™ unless it's thread not matter over WiFi then you need a thread dongle

[lusty linden]

Hi! I'm getting a crash on starting the newest Matter beta:

[crashed-dependency] MatterController unavailable due to initialization error
      at crashedError (/opt/matterjs/node_modules/@matter/general/src/util/Construction.ts:623:32)
      at handleRejection (/opt/matterjs/node_modules/@matter/general/src/util/Construction.ts:325:69)
    Caused by: [implementation] Fabric with index 1 is already present but not under the control of the Fabric Authority
      at <anonymous> (/opt/matterjs/node_modules/@project-chip/matter.js/src/MatterController.ts:383:27)

On top of that, the container keeps running after this crash, and the watchdog doesn't try to restart

On top on top of that, this error doesn't get shown in the log unless I set logging level to verbose

[fringe fjord]

hm ...no clue how that happened ... please delete in the data directory the server-* directory and try again (or restore a backup from before so that this is done via this.

[fringe fjord]

I think thats a question better for the normal Matter discord channel -> https://discord.com/channels/330944238910963714/1346944149303066711

[deep drift]

Yesterday, the OTBR Addon/App also received a Beta toggle switch to test Thread 1.4. Is there also a public Beta channel like this one to discuss the topic? https://discord.com/channels/330944238910963714/1458060881970856051 is read-only and only for developers to discuss.

[fringe fjord]

we are discussing that. I proposed it ... i inform

[deep drift]

It would be great to make, if technically possible, the exact Matter and Thread versions visible a vendor uses in each device’s firmware. With so much ongoing development in these areas, it’s already clear that some vendors aren’t prioritizing sustainability. While it’s up to them which features they adopt from the references, every release includes improvements and bug fixes to the base layers that can - as we are already experiencing - significantly impact network performance and stability. Giving users an easy way to identify the Matter and Thread versions in their devices would highlight the “lazy” vendors and allow users, bloggers, and the press to put pressure on their update policies.

[left rose]

While a new spec version in matter or thread doesn't immediately improve the device in itself, it is true that many manufacturers just use the SDKs they get from the chip makers committed to this market -- and ideally also release updates based on those SDK updates. And chip makers improve their stack foundation as they follow and align to new matter or thread versions and release updates. I have seen a number of matter devices, WiFi and thread, that went from wonky to solid over the course of being updated and updated from 1.0. to 1.4. Even if they didn't get any new features over time. That's why, yes, I'd also think that these numbers can be an indication of general product quality.

[flint compass]

Unfortunately, I'm not having the same luck with the new matter.js that so many others are. I have about 102 matter over thread devices. I was experiencing every-6-hour offline events for many devices that coincided with restarts. I don't usually like to install beta software, but I couldn't take the instability anymore. Installed the matter.js this morning and after a half hour or so, everything came back online again. But an hour later about 50 devices went offline again and the logs where churning with devices going online and offline repeatedly. I decided to restart the matter server to see if that would make a difference. but now I'm hovering around 80 devices offline after about 10 minutes. Any ideas where to start? Should I revert?

[silk plover]

You can revert, but would be helpful to the developer if you started an issue in GitHub and uploaded debug logs so that he could make the product better. Who knows something may jump out to him as the cause for your issues and point you toward a real fix. Something is obviously wrong with your thread network above and beyond either matter server implementation.

[flint compass]

I think you are right. And I'll do that, but my experience with github bug reporting is that a few people comment and then the issue goes stale. I don't seem to be able to attract developer attention. Any suggestions on that?

[silk plover]

Not really, completely dev dependent. Can only say that @fringe fjord has been very responsive to making this beta and upcoming stable Matter-js server better. Now is a great time to engage in my opinion.

Would post here with the GitHub issue link as well to maximize the chance of getting eyes on the issue.

[flint compass]

Will do. Also, my matter over thread network was stable for about 1.5 days prior to me updating to matter.js this morning. My every-6-hour "matter storms" only happen for about 2 days after a reboot, on the dot starting with the reboot time. I've examined my network every which way, and I think if I had network issues, the problems would be pretty regular. But since my storms are tied to HA or Matter Server reboot times, I think i'm seeing issues with the software somewhere. Maybe a dev will be able to help out.

[fringe fjord]

I was able to spot some stuff in logs and fixed more things. There will be a new version tomorrow … maybe revert to python till then and try the new version directly. Sure I love to dig through logs so I leave it to you if you open issue with log or wait.

[flint compass]

Thanks for the reply. I just created an issue in github. I also uploaded logs to review: from when I first upgraded, after the first "matter storm" started, and after I restarted the matter server again. Maybe you can see something in there? https://github.com/home-assistant/addons/issues/4355

[grizzled burrow]

While it's a pain to do, you can usually get that info from the CSA database. That's where I look up which Matter SDK a device uses. Unless there's standard metadata that a Matter controller can query for a device which stores the Matter SDK version, I think the best we can do is the CSA site. I think it would take the CSA to make any progress on a standard way for a device to report such internal metadata.

[silk plover]

From the post below, @fringe fjord seems to state that the version is now discernible for devices that are matter 1.3 spec compatible and above. If that is the case it would be great if that information is displayed somewhere in the UI. For those devices not >= Matter 1.3 something ‘Matter 1.2 or below’ could be stated.

#new-matter-server message

[grizzled burrow]

That would be awesome! Yes please.

[meager zenith]

not >=
this is criminal 🤣

[deep drift]

This pain is the very reason why this feature would be so helpful. Not every user digs as deep as we do and I’m sure even only few of us really searched the CSA database. Easy accessibility is the key to awareness..

[fringe fjord]

Basically 1.2 is also detectable ... only 1.0/1.1 is not possible to differentiate

[fringe fjord]

I will add this info to the server for the upcoming version ... it will only be shown in the Matter server dashboard for now but we can see whats we do in HA with it then

[deep drift]

This is AWESOME! 🙌🏼

[hollow cloak]

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

[tranquil agate]

The debug log...

[fringe fjord]

yeahh .. please eopen an issue ... this is "kind of known" innmatter,js ... the problem is that id did not found a way to reproduce this and with this debugging and fixing becomes really meeh 🙁 But i need to find a way

So effectively it is some race condition with some devices here and there. Yeah please open issue and I will check

[tranquil agate]

Done Not really into Github Bugs Issues, so please let me know if you need anything else...

[fringe fjord]

Small preview ...

[fringe fjord]

I just released v0.3.2 of the Matterjs server ... changelog see https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#032-2026-01-30
This should give us another try to validate the connection/reconnection handling and should optimize also detecting IP changes better. Depending on if old IPs were collected from a former version it requires a new mdns discovery to get the updated IPs.

[fringe fjord]

@flint compass Please move or recreate issue in https://github.com/matter-js/matterjs-server/issues thanks

[quick leaf]

Thank you for implementing the accessory matter version! 🫶

[left rose]

cc/ping @tranquil agate maybe?

[tranquil agate]

Well, I did make the same mistake and created the issue in the wrong place as well. Mine is recreated in the right place, so I guess this one was intended for Tony indeed

[loud raptor]

On my Beta HA system, I'm runnig the Matter App 8.2.2 (not called an Add-on anymore), with the 'Beta' toggle enabled.
How and where can I find which version of Matter.js i'm running, and how do I update it to v0.3.2 ??

[silk plover]

Just restart the Matter server from the add-on. As long as you have the beta toggled in the Matter Server add, it will update to the latest beta, now 0.3.2 upon restart.

YOu can see the version at the top of the Matter server webui or logs

[loud raptor]

Thanks, found it in the logs as well:
026-01-30 13:26:26.614 INFO WebSocketC~erHandler Starting server: matter-server/0.3.2 (matter.js/0.16.8-alpha.0-20260130-2554cb6e2)

[fringe fjord]

yes I also added it to the log that I can see either way 🙂

[grizzled burrow]

Two things #1 Love the Matter SDK status....very helpful! Thanks. #2) I love the concept of that Matter Graph tool. I installed it, and wow...very neat. Connection to HA didn't work, but did to the Matter server. SUPER helpful information. I hope that either becomes a HACS component, OR baked into HA.

I'm very pleased with the new focus on Matter and the rapid changes! Tickled pink!

[fringe fjord]

... i would propose to wait till next version comes ... https://github.com/matter-js/matterjs-server/pull/166

[grizzled burrow]

Yes it's not perfect..I have 71 Matter devices and not everything is perfect..but it's a great start.

[fringe fjord]

it will never be perfect unless you would aoso have all details from all affected OTBRs

[fringe fjord]

Sooo friends ... ready for the weekend and to give you something to play with after the "just stability version 0.3.2 earlier" we now have also some visual food 🙂 Matterjs server 0.3.3 is out ...
https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#033-2026-01-30

[junior sedge]

hmm, it can't settle down the visualization on my setup, keeps spinning :)

would be nice to have this visualization in HA tho, so it could integrate data from multiple sources (e.g. otbr integration to identify thread border routers and pull in their routing table knowledge, homekit maybe if it exposes some thread network info, possibly even esphome at some point?)

weird, the visualization is showing 4 unknown devices for me, but I can only account for 3 of them. (2 thread border routers, one esphome device)

what's the difference between a solid and a dashed line?

[silk plover]

Awesome work @fringe fjord the one thing the separate option seems to do that I really like is integrate the HA friendly name of the Node into the details, which I guess is easier for them because that one connects separately to Matter server and HA.

[stone bramble]

Hey folks. I am new to matter devices (And really home automation in general) and am trying to understand the "popcorning" effect I have with my smart lighting. I have my Linkind smart bulbs (Matter over wifi) wired to permanent power, but am seeing a few issues and am unsure if it's the devices themselves, my setup, the current state of the matter server, or matter devices in general.

1. Popcorning - When I toggle the light group on (5 lights), they are not synchronized. Devices frequently have > .5s latency differences between all of them
2. Some of my devices seem to disconnect after a few hours and stop responding to commands. Unsure if this is Matter or networking related and why it only impacts a few of them.

Any suggestions on how to debug issues like this? For #1, I was reading a bit and am wondering if it is related to clusters(?) and if the groupcast functionality mentioned in the spec is actually implemented by the matter server and if that is expected to improve the popcorning situation eventually

Basically I am trying to decide if I should ditch the matter bulbs for now and pay up for the Philips Hue instead

[fringe fjord]

yes, future step

[fringe fjord]

dashed line is "to an unknown device", some infos are also in readme

[fringe fjord]

Yes I decided against that, BUT I added something else. YOu can now use the dashboard also to set the "Node label" in basicinformation (usually on the nodes endpoint 0 -> BasicInformation" on top. is a textfield when you expand the "Node label" section. when the label is set then this is used in thiese graphs. Mopre needs to wait till we sorted out how to integrate this into HA

[fringe fjord]

If this is a general question, please head to the https://discord.com/channels/330944238910963714/1346944149303066711 channel. If you use the Beta and so the new Matterjs server in testing then you could set it to debug and post a log and I can take a look whats up there.
Normally when you control 5 lights they are all getting their command separately. yes in parallel sent but then it depends on your wifi network and pptentially link qualities and how ,any repeaters or such need to forward the requests and such ... so all depends on the quality and setup of your network

[silk plover]

Thank you, this works for me.

[stone bramble]

This is on the beta with the new matter server. I will grab a log of the popcorning. Network is Unifi + access points (no wireless meshing) with great signal strength. I'd understand a few ms of difference, but the 1 to 3 seconds differences are surprising to me.

Also, is there a CLI or something that allows manually interacting with the server over the websocket connection?

[fringe fjord]

But debug log please. You can enable debug in the dashboard ui upper right settings icon. That works while it runs and is temporary.

[fringe fjord]

The report has a Cli script to send a command but you want multiple. Use HA to trigger commands.

[orchid swift]

OK I have a question is it forbidden to talk about other stuff other local server that is not HA? I would not like that someone feels offended.

[reef smelt]

Using this beta matter server with something other than HA?

[orchid swift]

no, I bought IKEA matter couldn't wait, I have local sonoff ihost too, I dont have HA hardware, but I'm thinking about it.

does HA have a general caht room?

[reef smelt]

No. You should pick a channel and make a post. Using the iHost for HA would probably go to #1284966540617449515

[orchid swift]

HA doesn't have a general chat room?

[reef smelt]

No

[orchid swift]

thank you

[raven harbor]

🤩 that’s amazing

[grizzled burrow]

Thank you!

We definitely need 'map like' zoom in and out controls. My 71 devices are a mess.

I see a number of 'unknown' devices.....two of which I assume are my Apple TV 4Ks. However, I don't know what else would be unknown? Every Matter/Thread device I have should be connected to HA.

and I only have a single Thread fabric

[brave ginkgo]

I want to fully reset my matter system with the beta OTBR and beta Matter server. What is the best way to wipe home assistant clean of the current installs so nothing is remembered. I'm going to reset all devices and repair.

[raven harbor]

The cluster should be introduced in 1.3 so I would assume devices with Matter 1.3 or above would have complete data while old neighbor devices are limited or unknown?

[grizzled burrow]

ah that would make sense

[fringe fjord]

but no Thread network diagnostics exists since matter 1.0 ... ?!

but i guess it is optional so not all devices expose it

[stone bramble]

Well, I checked out the wifi map feature and it turns out that my network is not nearly as "strong" as I thought. The client device RSSI were significantly worse than what is reported by unifi and I found they were attached to some unexpected APs triggering retransmissions. Sorting that out and now the "popcorn" effect is almost completely gone! The only thing is that after forcing the lightbulbs to a specific AP, the wifi map in the matter server UI doesn't seem to want to update even after restarting. Anyway, great feature!

[raven harbor]

makes sense, could be newer chipset SDK to make them more popular

[junior sedge]

honestly, there should be no reason to go through the whole process of resetting and re-pairing? I don't think it'll give you any benefit over just upgrading.

[flint compass]

Can you clarify this comment for me? When you say "Wifi map" do you mean the "Thead map"? And when you refer to the "client device RSSI", what clients are you referring to? I have a Unifi setup, too, and see the popcorn effect in several rooms. Would love more insight on how you improved this!

[stone bramble]

My bulbs are wifi, not thread. By wifi map, I mean the feature everyone here is talking about in the new beta matter server web UI.

My unifi console showed they were all green signal < 50. But on that map, where it is client reported RSSI numbers, I was seeing orange and red in the 80s and 90s. And in the matter server logs with debug on, I saw retransmissions

In unifi, I had to go click on the actual access points to see the retransmissions

I think the root cause for me was a few things

1. Setting transmission power too high on the APs
2. I think the unifi "wifi channel optimizer" was throwing things off and making them connect to unexpected APs. I set all the channels to auto instead.
3. I may go a step further here and lock the bulbs to a specific AP

[flint compass]

Got it and makes sense. Thanks for the reply!

[junior sedge]

yeah, to my understanding it's really a network map, not specifically thread or wifi (shows devices on both network types)

[fringe fjord]

Try a re interview. Does that work and update them?

[fringe fjord]

I guess he means the new wifi device map in the matterjs server dashboard 0.3.3

[fringe fjord]

So issue basically solved? 😉

[fringe fjord]

Two separate maps. One thread and one WiFi

[grizzled burrow]

I use Ruckus WiFi, and the RSSI reported by Ruckus One and my water shutoff valve are basically the same. -69 vs -70

[silk plover]

Unifi cooking the test results to seem better or just inaccurate where as ruckus reports accurately?

[stone bramble]

I think so, at least for me 🙂

It seems like possibly a common-ish issue though, maybe worth mention in the FAQs somewhere. I guess a feature request would be an obvious warning when transmission retries happen / historical data on retransmissions (Maybe this already exists somewhere and I just don't know)

[silk plover]

After updating 10-12 nodes with labels, was curious if you could Put the Node label box in the top most menu and accessible/editable without ever having to drill down to Endpoint 0 | Basicinformation cluster. Maybe the Node id at the top could be clickable and editable there or even put the node Label text input optionto the right of Remove button.

Could the data still be stored in the Enpoint 0 | Basicinformation cluster if we input from that top menu? Not urgent at all just a usability question. Thanks again. Huge progress this week.

[flint compass]

For those of you who were experiencing every-6-hour matter devices going offline, are you still seeing that with matter.js? I am. However, they are shorter lived (thank goodness), but still they occur like clockwork and I'm noticing that up to 40-50 out of 102 of my devices will go offline and come back online. It's a slow build up and then back down, not sudden. At the moment, I'm seeing them around 03:25, 09:25, 15:25, and 21:25. I'm currently collecting logs and plan to open a bug. @fringe fjord, can you think of any event that occurs with the matter server every 6 hours that might cause this?

[fringe fjord]

From other users and when thread that could be devices moving from one Br to another. Also cases I am investigating with @left rose currently. I personally would not know why that schoolyard happen all 6h like clockwork but who knows. So let debug log run and grab a log from such a situation - ensuring to see min 30-60min before such an event and min 60+ mins after it. Then exactly, open issue with log and I can have look how the server behaved

[loud raptor]

On mobile, only Nodes are visible, I don't see the Thread and Wifi buttons.

[deep drift]

You have to use your desktop browser, not the HA app.

[deep drift]

0.3.3 was installed flawlessly and is - at least for me - rock solid again. No errors. Kudos, @fringe fjord what you were able to "fire" at us during the last days! 🚀 👏
The new Matter version display is fantastic, but also a bit disappointing to see that all my Eve devices - except the Eve Thermo 5th gen - and all Ikea devices are still running on Matter 1.3.0. This is especially surprising for Ikea, given their devices just launched. Hopefully, they’ll be updated to Matter 1.4 soon, as the sleepy end devices powered by Ikea’s rechargeable LADDA batteries could really benefit from Long Idle Time (LIT). The visualization of the Thread/Wi-Fi networks is a big step forward, though it’s - again - a little sobering to see the mesh connection quality. Are there any results yet on whether ZBT-2 offers better a signal quality than ZBT-1, aside from a higher bandwidth? Either way, I really enjoy the visualization, and if friendly names were added as discussed, it would make usability even better.

[hollow cloak]

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

[left rose]

Routers will then have to re-announce the devices' services via mDNS, and the matter controller has to catch up in order to learn the new IPs. So if you see these prefix changes happen frequently, that possibly indicates where to search next. @flint compass -- see above message, that obviosly got too long.

[fringe fjord]

I have no clue how all that will display on an app or mobile devices - also visibility. That’s why for now only possible starting certain screen size. See readme. If someone that better knows mobile ui dev want to do a PR I am open accepting this.

[fringe fjord]

That’s now such a thing. I would maybe see this a thing in HA. But because node label is cross fabric it also needs to be taken with caution….. also because you just do it once … but yeah I see your point. Wharves others think? Maybe starting a discussion?

[fringe fjord]

In general … what are reasons for such prefix changes? For me one question is if we can do anything against them? (Beside one router goes really offline for sure)

[deep drift]

I would prefer to use the friendly names that I have already entered for my devices in HA. Why add additional/redundant labels directly in Matter.js? So I guess this would be "a thing in HA", right?

[left rose]

The most common case that I know of is when the network falls into partitions, or any other reason when leadership is in question. This can also happen when partitions re-join. So in my experience that is something when multiple border routers are involved. Or when a single border router goes offline for minutes instead of seconds. So the crucial part is probably having a stable platform for the router to run on, and of course the infrastructure link to your local (W)LAN, So generally LAN is probably better than WLAN. Fixed channels on a WLAN is probably better than than auto-negotiation. Netwok settings that avoid fancy client shifting between APs, and finally having TREL enabled BRs. Channel management in general: Take care your 2.4GHz wifi doesn't interfere with your Thread Channel, join your thread networks.

[left rose]

So, more specific to your question: From the matter orchestration level you probaly cannot do much to avoid thread net splits and the like (unless maybe make use of thread 1.4's new observability capabilities, but idk if there is a generally specified API already). From the HA perspective: Release the best possible OTBR add-on. The openthread guys have put a lot of effort into stabilizing connectivity, optimizing routing decisions, imlementing solid SRP/mDNS components and the like over the course of the last 1.5 years or so. But all this stuff has yet to land in consumer products. Also, if you have multiple border routers and some of them are 1.3, still bad things can happen. For example:

[left rose]

In a multi border router setup, you will generally have more than one SRP server (responsible for promoting the devices service announcements to mDNS). So those generally act as a Proxy to the (W)LAN, so devices and the thread network don't have to handle mDNS by themselves. Especially for sleepy devices. Before thread 1.4 the prefered choice was to use mDNSResponder on the platform level to handle mDNS. Now, you have multiple SRP servers for availabilty reasons, and devices are free to chose an SRP server to do the advertising job for them. In fact routers can be promoted and demoted from the SRP server role and others will jump in. This also can create friction in mDNS continuity/availability, especially since the SRP spec requires that only one SRP server be responsible for one device at a time (for several reasons) (*). Neither avahi nor mDNSResponder were built with these sepcific needs in mind. So openthread rebuilt that from scratch. But it also means that mostly all 1.3 BRs today have these quirks.

A 1.4 router can compensate for that to a certain degree, but it also requires this router accidentally be in the SRP server role.

(*) You already see, the freedom of choice is limited here, because in order to implement the specific SRP constraints, SPR servers reject a devices registration request once they know another server is currently in charge for this device. Devices have to implement a coping strategy or give up. The SRP spec suggests a coping strategy that doesn't work well with matter (chose a different service name, but the service name is a negotiated/induced identifier during matter commissioning, so the device is actually not free to change it). That makes literally every matter over thread device violate the spec, because they have to retry in order to come online again, once their lease expires from SRP servers (or they drop their server role and another "clean" one takes over). 🤷‍♂️

[left rose]

(For reference, but also read from the beginning to understand why that is, and also keep in mind that SRP is rooted somewhere before 2017 used by apple before anyone thought of matter: https://www.rfc-editor.org/rfc/rfc9665.html#name-name-conflict-handling . Also the wording only changed lately from server to service which potentially opens a door to fix all of that in the future)

[fringe fjord]

Also the whole charting can be "better" when more infromation in HA are added, but this then would work there, so later step

[fringe fjord]

sounds like we shoukd write doen some "best practices for a stable thread and wifi network" 🙂

[deep drift]

The chart could be arranged/subdivided according to area /room information which would add to its readability.

[silk plover]

Agree having the friendly names as default should be the goal. I have input all 90 plus names by now and as apollon pointed out that is a one time event. Plus does this capabilty stay in Matter server or move to HA? So tabling this to later seems logical.

I would hope we retain the ability to rename/edit that friendly name for the charting/graph, mostly because friendly names can be quite long and abbreviating the HA name can make the chart/graph much more readable, but thats a minor concern overall.

[deep drift]

My friendlies are basically the device name plus a counter. But I am not sure my naming scheme is really sustainable…

Abbreviations like initials could be a way to automate creating short names for the chart

Just thinking aloud…

[rigid bone]

My 2 cents: A couple of weeks ago I was playing around with the otbr cli for mesh diagnostics and by deduction determined that my Aqara D&W sensors show up in a thread child table using one MAC, but show up in the Matter Server with another MAC address. As I only have 2 of these sensors, and there were 2 MAC addresses in the meshdiag thread tables that I could not account for (i.e. the Mater Server had these 2 device's MAC address that were not showing up in the thread tables) I made this deduction. I seem to recall that one of these D&W sensors went offline and I could then no longer see one of these MAC addresses show up in the thread table. I just switched over to Matter.js and I am seeing in the topology that the unknown devices showing with (?) correspond to these 2 differing MAC addresses.

[flint compass]

@left rose I just had my on-time morning matter storm. I checked the route tables in PowerShell before, during and after the storm, and there were no prefix changes. During this storm, I saw about 17 devices go offline out of 102. Thanks to @fringe fjord 's work and the new ability to put friendly names on nodes, I was able to see that the devices that went offline were clustered in two areas of my house that are on the outskirts of the mesh. I don't know if that's the "cause" or just the result of something that caused the weakest links to drop. What makes me feel like the cause is not something external to HA is that rebooting the HA host changes the time that the storms occur. Last time I paid attention, the clock started right at my host restart time. Since installing the new Matter server, the clock has shifted again, but unfortunately, I did not mark down exactly what time I installed it to know if it restarted the clock at exactly that time. But it would line up with the approximate time I believe I installed it.

Also, I don't have an OTBR, just Apple TBRs. However, I briefly installed the ZBT-2, and when it caused my matter network to churn, I ended up uninstalling and removing it. My 6-hour problems started after my next restart. So I don't know if briefly having OTBR installed may have started something that didn't rear itself until after the host restart. I don't know...

As @fringe fjord asked, I started capturing logs about 30 minutes before and I'm still capturing now so that I have an hour after and I will create a github post.

[fringe fjord]

I guess Monday will be my log day 😉

[junior sedge]

i'd expect matter 1.4's improved idle handling would help with some of the ikea devices that report infrequently - the window/door sensor, leak sensor - but the temperature/humidity sensor (and I bet also the light sensor built into the motion sensor) report so frequently that they don't really go idle for long. Kinda surprising tbh, i'm a bit worried that the temperature/humidity sensor will go through batteries really quickly.

also, huh, the zbt-1 can't do 460800 baud? That's very strange. From the specs it should be able to, I have a SLZB-07 with the same silicon labs EFR32MG21 radio chip and CP2102N usb-serial chip, and it does 460800 baud in thread mode.

[silk plover]

Disappointingly, All of my Ikea stuff is reporting as Matter 1.3.0 . So word is still out if Matter 1.4 will help those sleepy end devices.

EDIT: NVM I see that phaze75 stated the same in his post.

[junior sedge]

the zbt-2 does use a radio chip with more ram, so it should be less likely to run out of buffer space when sending messages in heavy traffic or doing relaying.

other than that, the antenna is the main improvement.

oh, the zbt-1 does run at 460800 baud with the openthread rcp firmware. The zbt-2 spec page is misleading/incorrect

maybe the zbt-2 uses a faster serial speed in zigbee mode, but they're the same in thread mode.

[flint compass]

Busy day! The log file is 20MB. 🙂 I'm going to restart my Matter.js server here in a bit and see if that shifts the 6-hour clock to the restart time. That will be another data point. Just want to confirm this.

[deep drift]

Hmm, ZBT-1 supports 460800 Baud. That’s the value it is set to in my OTBR config.

[fringe fjord]

effectively see manufactoring and certification test timelines for 21 devices ... I assume they will update

[grizzled burrow]

One thing I've noticed most days, is that the new Matter server appears to NOT handle multi-device simultaneous commands very well. For example, I have a sunset based trigger that closes 5 Smartwings blinds at once. About 50% of the time one of the blinds doesn't get the close command, even though it's online and responsive. Seems like there needs to be a little "micro" throttling logic.

[raven harbor]

https://github.com/matter-js/matterjs-server/blob/main/packages/dashboard/src/pages/components/header.ts It disappeared due to a global css

        @media (max-width: 768px) {
            .nav-tabs {
                display: none;
            }
        }

I tired to rearrange it without touching other parts

if it looks ok I can open a PR

[wanton mural]

Anyone that can assist me with an error in voice assistant?
Unexpected error during intent recognition
im streaming so u can see

[compact root]

#1284966915055288530 is where you'll want to post a thread for support.

[wanton mural]

did that 2-3 days ago but got no reaction deleteed the post

[compact root]

This channel isn't for support, it's for beta testing the new matter.js server, as the description details. You'll need to post in the appropriate space.

[fringe fjord]

No. Normally there is no limit. Provide a debug log in an issue with details please.

[fringe fjord]

But does it also work on small screens? Is it usable and readable? Especially when you select a node? So from my perspective that was not a bug but wanted especially because of usability of the graph pages. So I only accept a pr to enable the tabs when it really works and makes sense to have it. Getting UIs responsive is a huge effort in any case but such a thing even more.

I am honestly no frontend developer, so I am very limited here. That’s why I decide to offer it in ways that can also test and support. As said open to accept PRs that really make the feature fit and usable also on smaller screens. But when so that it makes sense.

[flint compass]

Issue created: https://github.com/matter-js/matterjs-server/issues/173

Thanks for looking into it!

[silk plover]

I have 2 Onvis plugs (Patio wall Fan 1 and 2 in the pic above) near each other, both function normally in HA and apple home. They are Matter 1.4.1 spec. Wall Fan 1 has never shown any connections on the thread graph. (rebooted the smart plug and Matter server a couple times). Just as a test I unplugged Patio Wall fan 2 for the last 7 hours and it still displays as the router for 5 thread end devices, which obviously is off. What I have not tried is rebooting the primary Apple Home TBR (family/spouse would be less than happy while they are home), which I suspect is feeding Matter server bad info, but thats only an uneducated guess.

[junior sedge]

it's normal for some devices to show no connections, if the devices don't implement the optional network diagnostics cluster.

(a future version of the ui might pull in info from the thread border router to fill in the gaps, but the current version is based only on info provided by the devices themselves via matter)

[silk plover]

OK thanks, but in this case They are identical Onvis S4 smart plugs on the same firmware.

and the one colored red in the pic with 5 end devices linked has been completely unplugged from wall power for 7 hours now.

[junior sedge]

that's a bit odd, the devices should have moved to using different routing, assuming there's another router within range that they could switch to.

you might want to pull the thread diagnostics from the otbr to get a better idea what's going on, in case some of the matter-provided info is wrong or out of date :(

[silk plover]

There are several thread routers nearby and those 5 end devices are functioning normally now and over the last 7 hours. Can only assume they have moved to differnt routing and the graph is not updating.

Not using an OTBR, only apple TBR in my home network currently.

[grizzled burrow]

Weird network name:

[zealous moon]

Utf 8 something?

[grizzled burrow]

Most of my Aqara P2 Motion sensors are floating aimlessly in space with no lines, even with Matter 1.4.1 firmware. Bummer.

[silk plover]

Same here on the P2 motion as well as an Aqara P2 Contact sensor. Assume aqara’s firmware to blame.

[grizzled burrow]

My P2 contact sensors have lines

[silk plover]

Ha, well then I dont know.

[grizzled burrow]

Only my P2 motion are not anchored to a router

[junior sedge]

huh, none of my P2 devices have the thread network diagnostics cluster, so they're all floating.

[silk plover]

OK checked neither the Onvis S4 nor the Aqara P2 (both the motion and contact sensor) have a ThreadNetworkDiagnostics cluster under endpoint 0. So all of those should be floating, if I understand correctly.

[grizzled burrow]

I have S4s too, but they are tethered on my map:

[silk plover]

Well the graph can’t be 100% accurate yet because the S4 that has been unplugged for 9 hours now still displays as a router with 10 connections. While the other S4 that is plugged in and working fine from HA and apple home is floating in space on the graph.

[fringe fjord]

Infos are coming from the devices. We have no access to the BRs. Try to re-interview the node where you suspect old data. Does it update then? I sometimes have the feeling that the data are not always updated via the subscription. But not sure if a polling would not be problematic. I could think about an update button per node or such which then just pulls the details of the node anew or such.

If we really need to poll the data to ensure they are up to date then that might be an interesting challenge

[silk plover]

After 10 hours unplugged the graph still had 8 connections to the unplugged device. Re-interviewing each of those 8 nodes fixed the graph and the unplugged device is now floating alone.

Would agree the data is not always updating on it’s own even after hours.

[fringe fjord]

Ok I will do the following:

• when a device is offline I will show all connections to it of from it to others dotted or such to indicate that might be strange
• a offline device will get an „update peers“ button which updates the infos of all nodes we think the were connected
• a connected device will get also an update button where you can choose to update only this device or this and all referenced peers
  I would not propose to cascade the update further because could lead to an enormous request storm.
  Opinions on that?

[sturdy ivy]

I can confirm this. I have a room with 4 Zemismart blinds. Most times opening/closing all simultaneously works correctly, but probably around 1 in 5 attempts results in one or more not reacting at all.

[sturdy ivy]

Just an update on this. I've tried again this morning in order to gather some logs for you but now everything seems to be working. I don't know if you have changed anything recently but I'm now on version 8.2.2. What is now happening is that in the logs, all 4 Invokes are logged simultaneously, but on some occasions, 3 blinds start moving immediately, with the 4th moving maybe 2-3 seconds later. Good news. 😄

[fringe fjord]

The next version (and 8.2.2 is. Or that relevant - check the server version on dashboard) will get better logging on invokes which includes info when there are retransmissions. So for me it sounds like retransmissions and potentially failing commands.

Also here a debug log would be cool. With next version that’s visible on info log too

[deep drift]

Will retransmissions and potentially failing commands logged on Info level or already on Warning level? I would like to see things that are worth having a look at already on Warning level. The Info level is already very crowded. What do you think?

[grizzled burrow]

I see this as well. Haven’t had a failure in a few days now. Was fairly consistent on early betas.

[fringe fjord]

Hm .. are retransmissions a warning? In my eyes not if it was successful. Retransmissions can happen for many reasons and are basically part of the protocol. So it is basically "Ok" for them to happen. They CAN show issues in your network or you flooding the thread network with sending commands to more devices in parallel as they can handle or can happen because of other reasons (e.g. we analyzed nanoleaf light bulb issues when turning off (on was working great with multiple devices, off was a mess). The reason turned out to be on nanoleaf side because the data sent back for "on" were roughly 150byte matter payload. An off generated 1.500byte matter payload per device. majority of the data were nonsense but in fact that was the reason here.
So when retransmissions happen it can be worth to temporary enable debug and look whats up, but in fact it is not automatically "bad". So no "warn" in my eyes.

Matter,js has a notice loglevel above info and below warn ... we can move more things there and so split a bit maybe

For now in the result log for an invoke you will see potentially additional data in that line like

Invoke « @1:xxx•yyyy⇵zzzz 1↔1+3 .....

if that shows up then the "+X" means that 3 retransmissions happened to send 1 message and receive 1 message

And said that ... Have fun with the v0.3.4 of the matterjs server. It mainly includes some enhancements for the thread/wifi graphs and allows to update the data of selected nodes when you think they are outdated. Also enhances more information from routing tables when existing.
It also includes the above mentioned logging enhancement to more easy spot if/when retransmissions happened.
Full changelog: https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#034-2026-02-01

[deep drift]

That sounds reasonable. Thanks. 👍

[silk plover]

Apollon working weekend overtime with a new 0.3.4 beta. Update button, with an option to update the connected nodes looks good from the thread network mesh dashboard.

so for my own understanding for those nodes without a ThreadNetworkDiagnostics Cluster on Endpoint 0 (such as Onvis S4 and aqara P2 in my home) the routing data on that node comes connected neighbors? Which is why if any connections are listed those connected nodes have (reverse) shown next to the connections details.

Nodes without a ThreadNetworkDiagnostics Cluster on Endpoint 0 are also the ones expected to be free floating and not visibly connected to the mesh at times when others nodes, with a ThreadNetworkDiagnostics Cluster, have not responded with a connection to them.

[grizzled burrow]

I was thinking of blogging about the new and wonderful Matter server, but Apollon pushes out too many wonderful features to keep up. 😉

[fringe fjord]

Ok, had a look and commented. from my Pov there was a strange IP change for many devices - i wrote one example but the picture is the same for all affected ones just not the direction. So something was happening on network level in my pov ... https://github.com/matter-js/matterjs-server/issues/173
Maybe someone else might have ideas because i do not see an issue on the matter server side (also because you wrote you had comparable things with the python server!). In fact the matterjs server handled it good, nodes went "offline" (aka subscription timeout), we tried to resubscribe, then reconnect then discover and we found the new IP and was working so connection restored. I could not seeing anything "broken" here...

[fringe fjord]

exactly. it is the same fr my S4 ... and thats why these connections are also dotted (when only exist reversed). Best effort we can do here

[fringe fjord]

lol ... lets say it that way ... I will be out in Utrecht next week tuesday till friday, so in that time I am limited anyway, so use that time g

Aaaaaand ... if anyone is around Utrecht ... On Tuesday 18:00 CET is a HomeAssistant Meetup in Utrecht ... come and lets talk! 🙂

[silk plover]

well earned, enjoy!

[grizzled burrow]

Two suggestions for the maps: Have a 'freeze' button so it doesn't keep spinning around and #2 Zoom in/out buttons like google maps

[sturdy ivy]

Unfortunately I was too quick suggesting success. Just had only 2 blinds close, but 4 logged cleanly.

2026-02-01 16:01:30.854 INFO ClientInteraction Invoke » @1:56•6ed1⇵6740 0x1.WindowCovering:0x102.downOrClose:0x1 with (no payload)
2026-02-01 16:01:30.858 INFO ClientInteraction Invoke » @1:54•6ed0⇵6741 0x1.WindowCovering:0x102.downOrClose:0x1 with (no payload)
2026-02-01 16:01:30.858 INFO ClientInteraction Invoke » @1:52•6ece⇵6742 0x1.WindowCovering:0x102.downOrClose:0x1 with (no payload)
2026-02-01 16:01:30.859 INFO ClientInteraction Invoke » @1:53•6ecf⇵6743 0x1.WindowCovering:0x102.downOrClose:0x1 with (no payload)

Forgive me if this is complete nonsense - I haven’t looked at the server code. However, this feels as if it should be soluble with some jitter applied at the point that the Invoke commands are transmitted. I’m assuming there is a transmit queue somewhere in the system? So would it be feasible to apply a random delay of between, say, 0-50ms at the point that the commands are transmitted. I would have thought that that should be enough to resolve these sorts of command clashes. I don’t know if or when Matter groups might be implemented, but presumably they will inherit the same problem as it is at the Thread level that the network seems to be gumming up.

[grizzled burrow]

I was thinking the exact same thing when it was happening to me. Apply some jitter for commands queued up in quick succession.,

[silk plover]

Has anyone had success setting up a matter binding with recent/latest Matter-js server? Trying to test out binding support for an inovelli beta firmware and get an ACL target error from the matter server when doing so.

[grizzled burrow]

If only all my Hue lights were Thread enabled.....sigh.

[silk plover]

tried Matter Binder helper as well and it shows a little more info, same error if I go switch to switch or or switch to Aqara T2 bulb. No error or warning in the Matter-server logs my untrained eye can see. Stacking betas so unclear where the error source is.

[deep drift]

Yes, I successfully set up a Matter binding between an Eve Thermo 5th gen and an Ikea Timmerflotte temp sensor.

[fringe fjord]

for me it just spins shortly and then stabilize,but maybe too small. open issue please for it. also the zoom buttons

[fringe fjord]

Thats the sending direction ... you should see also a "Invoke «" ... thats the relevant. When we start sending we do not know how good that will be or not

[fringe fjord]

I guess if you ask 10 people on "adding delays to commands" you will get 12 opinions ... I could also say "if your devices can not afford that many parallel controls enhance the automation to delay it" 🙂 But yeahh ... thats maybe a discussion to better do structured. Also: just delayung them a but might not be enought as examples above showed. SO if 4 in parallel do not work ... switch to debug log using dashboard and grab a debug log and lets see what exactly happens

[fringe fjord]

issue with debug log please and I have a look

[fringe fjord]

could also be time now to add the binding helper to the new dashboard 🙂

[grizzled burrow]

#177 opened

[grizzled burrow]

Yes please make it point and click easy to bind devices

[fringe fjord]

I responded in the thread of that guy that presented it in the python server repo ... I would hope he does a PR?

[silk plover]

what do you know ... binding worked switch to switch this time, was my 3rd attempt. will try a couple more to see switch to switch if I can capture and issue. will also check switch to light bulb.

[left rose]

I would guess in thread jitter is less about (not) overloading the devices, and more about not overloading the thread channel. Jitter is also common practice in the thread stack itself in certain situations. But yeah, the logs will certainly reveal what's good and necessary.

[fringe fjord]

whats a "swutch 2 switch" binding? WHich sense does that have? Which devices support that?

[silk plover]

Matter Binding fails #179 opened

[silk plover]

Just testing for inovelli on/off dimmer firmware beta. They allow their wall light switch to bind/control another of their wall light switches On/off and dim. not sure of the practical uses, other than you always want 1 conveniently located wall switch to control the loads (bulbs or plugs) of 2 separate circuits in your house. Don't use this potential functionality myself.

[fringe fjord]

ahahh ... puhh ... yeaahh ... a typical "because it works" thing 🙂

[fringe fjord]

Ahhh …. And if Utrecht is no option please be remembered that Mo 9.2. 19:00 CET we have the next matter community sync on Google meet. So if you have feedback (good or bad) or want to discuss ideas or just hear about the next steps or such. Anyone invited. Just join.

[shell citrus]

I’ll be at the meet-up in Utrecht on Tuesday. Let’s say hi 👋🏻

[rough moth]

So i am assuming that when asking about matter this might be a good channel.

Here is the thing i added the ikea diriga hub to matter server and i got entities and devices however not all of them namely the new ones. alpstuga, myggspray and kajplats

Anyone have the same issue?

[fringe fjord]

responded ... I will improve errormessage but effectively ... RESOURCE_EXHAUSTED ... https://github.com/matter-js/matterjs-server/issues/179#issuecomment-3831867422

BTW .. The pythonserver MIGHT hided such issoes from you because I am unsure which code stand it exactly uses, so in python serevr it could be that when writing so many values only the first 5 were really added and the rest was also erroring because "too many" but the python server never returned this as an error because the first was ok. We are a bit more exact on that and so in case nothing got written, so also nothing "half right".

[manic lichen]

This is not the right channel for general Matter questions. It is just for the new Matter server testing. Use #1284966617670881350 with the "Matter" tag instead.

[rough moth]

oh

nothing to type like the same in here?

[manic lichen]

Create a thread in the channel I linked and use the "Matter" tag. Maybe someone can help you with your devices.

[rough moth]

ah nvm then

[silk plover]

Thank you, with that response and the help of Matter binding helper I looked into the ACL's on the Aqara T2 node giving an error, there are several undeleted ACL stored there (those bindings were deleted with the python matter server months ago), which is likely the cause of the error. Any Idea how to delete those? No rush, and could be addressed in the future if Matter binding helper is onboarded the matter server dashboard.

[fringe fjord]

deleting ACL is basically writing the array without them

And the current logic to add an ACL is basically appending one to the value that already exists. I could improve now but that would hinder the dev from that helper because then he get more conflicts ... so lets see if we get a PR or if I need to add it manually somehow

Is he also here in discord? Anyone can mention him or such?

[silk plover]

Like I said no rush and would rather you all get the time to do it once and right. Not sure about him on discord, I also posted to the Matter Binding helper github asking about integrating with you.

[fringe fjord]

A "real" end-user compatible" binding is funny thing and before that I guess we look better into groups .... but before that some more things ... lets see. First is to stabilize it and fix all pressing issues

[modern olive]

@fringe fjord After the last couple updates, the node I had that was still offline remains so - and now a few others have gone offline as well and are not showing any signs of reconnecting.

One thing I noticed that I don't understand, is when I look at the new (and very nice! 🙂 ) WiFi graph, I see a bunch of disconnected clusters. I have exactly 2 APs serving these devices, each with a single BSSID, and yet there are many different AP BSSIDs in that graph, and so far I couldn't pinpoint any of the actual ones. Is this normal?

[fringe fjord]

try to power cycle the nodes? Maybe they need to be restarted once for us to get current IP or such? What do they show in HA when you check IPs?

For the clusterds ... good question. These should be the BSSIDs from the infosof the devices ...soooo ... good question. if you like check the networkcommissioning and wifi network diagnostics cluster details of the nodes

[modern olive]

Last time I checked they all had the same IPs, and they shouldn't have any reason for changing IPs. I'll check again but it's so hard to match a Matter node to a device in HA. Any tips for that other than clicking each one?

Also, where can I see on the Matter Server which IP it thinks the device has?

Just found one on the server lots:
⁨```
2026-02-01 21:02:49.968 DEBUG PeerSet @1:62 Starting reconnection polling for udp://192.168.49.47:5540 (interval 10m)

Same node on HA shows as ⁨`192.168.48.120`⁩ and responds to pings on this address. As far as I can tell, the device never had the IP shown in the Matter Server logs.

Actually, it looks like both IPs are assigned to the same MAC address. 🤔

And that MAC address doesn't match the device in HA.

[fringe fjord]

Two different network segments?? What’s up there? Are you sure the mdns works cleanly there?

[modern olive]

It's a ⁨/23⁩subnet.

So, it's a single segment / subnet: ⁨⁨192.168.48.0/23⁩⁩, which gives me ⁨⁨192.168.48.1-192.168.49.254⁩⁩ usable IPs, with ⁨⁨192.168.48.0⁩⁩ for network and ⁨⁨192.168.49.255⁩⁩ for broadcast.

[modern olive]

Ok I just triple-checked all this information. What HA has is correct. The device is on ⁨⁨192.168.48.120⁩⁩, and is responding. I don't know why the Matter Server thinks it's on ⁨⁨192.168.49.47⁩⁩.

There is another device (same manufacturer and model, different MAC address) on that IP.

The MAC address that matches that IP is ⁨Node 98⁩, which is also offline. The IP for that node is also correct in HA.

How can I check which IP the Matter Server thinks a certain node has?

[flint compass]

Excellent and thank you so much for looking into this. I'll continue any conversation in the github post because now, it seems, this issue is not related to matter.js and don't want to flood this forum! Thanks!

[fringe fjord]

Basically Via the line yu referenced above. so in this case node 0x62 (98) has its "last known IP" as 192.168.49.47 ... and usually waits for the device to be rediscovered via MDNS and queries it also. Also interesting i find why there is no IPv6 address ... so maybe provide the log because it is avdifference what happens if he get an answer from that IP but the cert credentials do not match vs not getting an answer.
So ideally restart the server, set loglevel to debug before and then post the full log into an github isue when the device stay offline after that restart

[lusty linden]

Hi! I'm wondering if I'm facing something similar. I have 12 bulbs, and one dropped off. When I restart the add-on, it tries to connect to the ipv6 for 11 of them. The 'broken' bulb is a Govee GU10, mac address ends with 87:89:5c, and ip neigh finds it at 192.168.0.39 and fdb7:eb19:183a:0:e4e:a0ff:fe87:895c, and both addresses respond to ping. Here's a long log, with the last run in 'debug' mode.

[lusty linden]

Hmm I see that it indeed doesn't advertise over mdns

This is Settings → System → Network → Zeroconf Browser

Ah well, I'll try resetting that bulb I guess

[lusty linden]

Yeah, I should be seeing 96F9389A1855BBC1-0000000000000021._matter._tcp.local. I blame the bulb! Thanks for rubberducking

[raven harbor]

@fringe fjord Hi there, if I want to add support for custom clusters, do I need to contribute to the new Matter Server before making PR to HA Core?

[signal drift]

anyone konws why eve's custom cluster has name for each attributes but heiman's has not?

[tranquil agate]

Does it make a difference for the Matter Server if it uses a LAN or WLAN? Especially with regard MoT and MoW? I've not really been aware of this before, but it appears that de Matter Server does not follow the global network adapter settings for integrations. (I know it's an add-on (app🤭), but is it possible to make it follow the global settings?)

[fringe fjord]

if your question is about @1:21 then yeahh ... we do not get any MDNS response, hence we try to reach the known IP but we can not do more

[lusty linden]

I must’ve missed that part in the logs where it tries to reach the link local address

[fringe fjord]

we know the following custom attributes in the matter server from heiman and these should be named. oif not please report a bug.
https://github.com/matter-js/matterjs-server/blob/main/packages/custom-clusters/src/clusters/heiman.ts

Additional ones are not named like in eve

[fringe fjord]

Yes we need to add them to the matterjs server first ... see link in my last message for examples how they are defined in the matterjs server, BUT we miss one step and this is generating Python classes for that so that HA understands them. So even if you do a PR for the matterjs server now this would not enable their use in HA. We need to regenerate the python classes. Thats on the TODO list but not yet done

[fringe fjord]

Your restart resetted the loglevel to info, there the IPs are currently not logged

[lusty linden]

My bad. Thanks!

And definitely the bulb’s fault, right?

[fringe fjord]

You can set the interface to use in the addon settings. What you mean wit MoW and MoT?

[fringe fjord]

right now I do not see anything else in the logs

[silk plover]

MoW - Matter over Wi-Fi and MoT - matter over thread

[junior sedge]

on a wifi bulb it would be the bulb's fault (or a wifi configuration issue); on a thread bulb is could be something going wrong with SRP on the thread border router.

[lusty linden]

Thanks, it’s over WiFi. It’s twelve of the same bulbs and one isn’t working, so I’ll just reset it, and replace it if that fails. Don’t buy Govee, everyone.

[silk plover]

I have 3 Govee outdoor lights mounted to the house. Have never been able to keep them connected to Wi-Fi, much less MoW.

[fringe fjord]

lovely ... who invented that? 🙂 All good.
In fact irrelevant - beside the typical "do you want to have such a controller running over a potential instable wifi connection" and such. So that would just bring another potential "all devoces reconnect becasue the controller moved it's IP" cases

[fringe fjord]

mostly just powerdown-püowerup should be enough, full repairing would be very strange

[lusty linden]

Yeah I’ve tried that, so re-pairing it is

[tranquil agate]

Hmm, should I set it using the "Extra Matter Server arguments"? It's a bit hidden then in the settings.

And sorry about my abbreviation for Matter over Thread and Matter over WiFi...

[fringe fjord]

If you need a special interface ensure it is available also in the docker/addon and then provide there. Server will error on start with an MDNS initialization error if interface is not available

[tranquil agate]

Ah, I see, thanks!

It is more or less born out of "trying as many different solutions to rule out anything that makes my Matter over Thread installation fail". Therefore I changed my setup to only use WLAN. But I'm glad to see that it's better to just use LAN for my Home Assistant.

(Also because Home Assistant always enables the LAN interface (using PoE), so I had to manually disable it every reboot)

[modern olive]

Thanks, I'll grab the logs. You don't see IPv6 because I only grabbed the line with the IPv4. 🙂 IPv6 was a few lines down but it's a similar issue as far as I can tell.

I'm interested in how and where the Matter Server keeps this mapping and what it uses as a key. Is this kept on a database, or temp file? It's curious because it randomly happens after a Matter Server restart but the devices haven't changed IPs.

Also, what does it look for in mDNS?

[modern olive]

Here's all logs for that node after a restart. The IPv6 address matches the MAC address of the other device. It looks like the Matter Server has some of those devices flipped around.
⁨```
2026-02-02 18:49:12.336 DEBUG PeerSet Pair with @1:24 at udp://192.168.48.120:5540
2026-02-02 18:50:29.388 DEBUG PeerSet Pair with @1:24 at udp://[fd00:ca55:1ea1:48:9aba:5fff:fe84:286d]:5540
2026-02-02 18:49:59.179 INFO PeerSet Resume failed, remove all sessions for @1:24 before Mon Feb 02 2026 18:48:25 GMT+0000 (Greenwich Mean Time)
2026-02-02 18:49:59.179 INFO PairedNode @1:24 Error waiting for device rediscovery, retrying [pair-retransmission-limit-reached] Failed to connect on any discovered server
at Function.iterateServerAddresses (/opt/matterjs/node_modules/@matter/protocol/src/peer/ControllerDiscovery.ts:237:23)
at async <anonymous> (/opt/matterjs/node_modules/@matter/protocol/src/peer/PeerSet.ts:642:32)
2026-02-02 18:49:59.180 INFO PairedNode @1:24 Reconnecting in 30s ...

[kind sundial]

I understand it’s beta, Seems a bit high on ‘idle’, or normally like this on stable non-js?

[silk plover]

https://www.matteralpha.com/how-to/quick-guide-to-visualize-your-thread-network-with-home-assistant

[fringe fjord]

Thread? Multiple border routers? How many? Or WiFi? There is a lot of „depends“. We are currently trying to optimize exactly the ip Change situations and ensure that once ip change happened and device was connected also this new ip is stored. So in fact details are interesting. Potentially set loglevel to debug and restart the server and wait till all get sorted out. Then try another restart and see if devices had old ip. If yes post both debug logs as issue.

[fringe fjord]

„Idle“ is very relative when it comes to matter because in fact it is never idle … so what does the log say? What is happening? … ahh you mean ram … what does 4.5% mean? What’s 100%? How many devices? After first start or later start? Do again many questions

[kind sundial]

Yes, that’s why I put it in ‘quotes’, as I’m not sure of the correct terminology. Not sure why you’re being passive aggressive, was just asking the average memory consumption. I’m new to this platform, I have it installed and running with no devices at the moment.

Sorry if I offended you in any way

[fringe fjord]

Whhooo … sorry if my answer arrived wrong. My intent was really to get information … else I cannot answer your question. I do not feel offended and my answer should also be not sound like that. But ask yourself: could you answer that with the information you provided? 😉

[kind sundial]

It’s fine, this is all I needed to know - “'Idle' is very relative when it comes to matter because in fact it is never idle”

[fringe fjord]

I was first thinking you mean CPU. That’s why I edited then.

[modern olive]

Ok will do. I don't have any Thread devices, it's al WiFi. When you say "wait till all gets sorted out", what do you mean exactly? These devices just never come back online at all.

[haughty stump]

I see some folks posting screenshots of their matter server and "thread network view" and it shows friendly namees (device name). I only see MFG and Model of devices, it never brings over any friendly names..

[silk plover]

You have to manually open each node from the matter server gui and hand type/edit those in. click into endpoint 0, then click into basicinformation cluster to find the ‘node label’ text field.

[loud raptor]

Why does only the Shelly plug show a connection to an AP? All other (Tapo) plugs don't show a connection.

[haughty stump]

Is there any plans to populate this or link it from HA data? seems like a pita to have to do that

[silk plover]

Not soon, lots of variables, but devs understandably want to get a stable Matter-js server out first. Then make some decisions on what lives in a separate Matter server app versus what goes into HA itself that will impact how/if automatic friendly name importing happens. What we have now is better than what I was told we would get when I asked your question a month ago or so before the dashboard Graph was onboarded.

Difficult not to be impressed with how far Apollon has brought this project since even that short while ago.

[junior sedge]

if the device doesn't provide the (optional) network diagnostic information, then the matter server can't show anything. I guess the tapo devices don't implement it.

[grizzled burrow]

I did the one time pain for 72 devices.

[fringe fjord]

Then that means there is an mdns issue? Is mdns working and all flowing and get routed? Does it come back when you repower the device?

[fringe fjord]

Check if the have a WiFi network diagnostic cluster on the root endpoint? Would be my guess

[fringe fjord]

Is anyone here that also has the issues with timing out subscriptions with some ikea devices? If yes please use the dashboard and go to endpoint 0, ICD management cluster. Show me the featureMap and mode attributes. I think unfounded out what’s up with that behavior.

[silk plover]

coincidently, had one go offline at some point last night. has not been an issue, at least when I have looked, up till now.

[fringe fjord]

Ahh that’s looking good. I think i need to add „more likely if also paired with a second ecosystem like Samsung or Apple 😉 just one ecosystem likely all will work.

[silk plover]

i'm also paired to apple home.

My issue might just have been a sudden battery death, from an older pack AAA I had laying around during install.

[terse loom]

Hello everyone! I'm a smart home enthusiast from China. I haven't set up the Home Assistant platform at home yet—can I create it using Windows 11?

@silk plover

@fringe fjord@grizzled burrow@junior sedge

[silk plover]

Wrong channel for this question, start here: https://discord.com/channels/330944238910963714/1402660587980324954

[terse loom]

I'm sorry.I'm sorry.

[manic lichen]

And please read the server rules.

[terse loom]

Sorry, I'll go translate the server rules right away. Sorry, admin.

[fringe fjord]

Happy battery replacing.

[loud raptor]

The Shelly plug has 'WiFiNetworkDiagnostics', the Tapo doesn't

[fringe fjord]

Then no option to get the bssid. Idea why we should do?

[lusty linden]

Closure: re-pairing fixed it. Don't buy Govee bulbs.

[modern olive]

mDNS works as far as I can tell. Most devices are fine, I can SSH to hosts using their ⁨.local⁩ name on my network, I can browse mDNS and see lots of Matter stuff. Last time I had this issue, power cycling the devices would eventually work (some took 3 or 4 power cycles).

One piece of potentially interesting information, all the devices that are wrongly offline are Tapo Smart WiFi Power Strips (P304M).

[modern olive]

Actually that's wrong 👆. One of them is a Tapo Smart Wi-Fi Plug (P110M).

I have a few more of those though, and they're fine.

[fringe fjord]

In fact you can check. If you know the node id of them … you see the matter operational announces like Xxxxxxx-YYYYYY … the Y is the hex node id filled with 0s and HA seems to be the 8C… so are the missing ones there?

[modern olive]

No, they're not.

After 2 or 3 power cycles, the plug I mentioned last popped up in ⁨_matterc._udp⁩, and then eventually in ⁨_matter._tcp⁩ at which point it went back online.

One of them has briefly popped up on ⁨_matter._tcp⁩ but then was removed again. During that time, I tried to look at its properties on the GUI but it showed nothing when I expanded it.

Now it has appeared again, this time with info. Still offline though.

[fringe fjord]

Then it is really a device issue.

[modern olive]

Logs for it after restarting the Matter Server.

[fringe fjord]

So the matter server had the right ip but the device was not responding to matter messages

[modern olive]

Looks like that

[fringe fjord]

Are these tapo devices updated to latest firmware? I remember that especially first firmwares were horrible… so updates might help. Pot needed via their app

[modern olive]

I'm going to re-check but they should be. The stock version didn't even report energy readings via Matter, it was only on/off.

[modern olive]

Yep, all up-to-date.

They're fine in the app as well, so they're up and running.

[fringe fjord]

Yeah matter is different protocol on the devices 🙁

So removed it’s matter record and IPs. And not responding to matter UDP.

[modern olive]

I'll try to capture packets on the router, see if anything interesting shows up.

It looks like the only packets with the device as source are mDNS announcements.

On the Matter screen in the Tapo app, it shows 2 Bound Services: Apple Keychain and Home Assistant. Would Apple Keychain be the other stuff that pops up in mDNS? Is there a way to get rid of it altogether?

[junior sedge]

apple keychain is a matter fabric added temporarily by the iphone as part of device commissioning. It's safe to remove (you can do that via the home assistant UI)

it would result in an additional matter mdns record being advertised

[raven grotto]

Great tip, was wondering about that but never remembered to ask

[modern olive]

Thanks! Where do I remove it?

[junior sedge]

Matter device page, Device Info panel, , "Manage fabrics"

[modern olive]

Oh that's for each individual device?

[raven grotto]

Just hopefully one quick question so I don't pollute this channel, but if a device is paired to both HA and Apple Home, it is still safe to remove the Apple Keychain entry? Would the device still appear on my phone under Settings -> General -> Matter Devices?

[junior sedge]

I'm not sure. I think once it has been added to apple home you probably only need the apple home fabric?

[raven grotto]

One way to find out 🙂

[fringe fjord]

Apple Home App -> Apple Home
Apple keychain -> Settings - General - Matter devices

Alright guys, I just release 0.3.5 of the Matter server (https://github.com/matter-js/matterjs-server/blob/main/CHANGELOG.md#035-2026-02-04) ... contains some more optimizations.

[deep drift]

I just lost two Timmerflottes in my mesh. I guess they somehow crashed, neither Matter.js nor OTBR restarts could recover them. So I took the batteries out and they came both back. So far so good. Unfortunately, both their Matter bindings to Eve Thermos did not recover. Both Eve Thermos still show only the temp of their own sensors instead of the external temp of the Timmerflottes. Debug is on.

[fringe fjord]

for bindings: try repower the thermostat.
Fpr bindings the controlelr/matterserver is 100000% out ... bindings work between devices, so likely the session is dead and need to be redone .. such would be a bug report fto te eve guys tbh - so you paired Ikea temp sensors to eve thermostats?

Also: how long you waoted for temp to come back?

[deep drift]

I just tried to reinterview the Thermos. did not help

[deep drift]

It's now 17 minutes

[fringe fjord]

no clue how their polling schedule and such is ...give it up to 2h i would assume, but please report back

[deep drift]

I will now take their batteries out. or should I wait for them to recover?