#the-water-cooler
1 messages · Page 82 of 1
Your bike looks quite clean
Are the small ones on the floor chonky blade servers?
Well, they are edgeline 1000s
Kinda blade ish
Not quite
More, edge nodes for rough environments
That have cartridges
We have the Moonshot 1500 too which fits 200 servers in 4.8u
Or was it 4.2
fire extinguisher is also ready, in case atx decides to cook
you are always welcome :)
Yeah it can
random idea i had
And this for the forums?
You would be surprised how many there are, doing all sorts of stuff
Want a gif of a raccoon riding a bike?
I got your back
no
atx got your back
Did someone say internet backbone
Figured out root cause, the hdd i was using was starting to fail
Glad i picked up a spare cuz gut feeling told me there was a chance
cool
my nvme drive in my pc died last week
still sad about it
i specifically said it wasnt haos cuz if it was buggy shit we'd all be complaining daily
got an ssd this time around instead of spinning rust
yep
Also explains why i couldn't find logs of the root cause when it came back up from reboot lol
restore process took a lot longer than the estimated 45 mins by the webui
2.5h
But, i'm back up and running 🙂
Interestingly, the restore process doesn't restore network config
does restore the DNS servers set in the network config, but not the network interface setup
was slightly frustrated by that but it's not that big of an issue tbh
i can see the value of being able to restore a backup to say a secondary system before replacing it, making sure services are running so it's a relatively seamless replacement
it depends on database size and what hardware you are on etc
and the network config used to be 1:1 after a restore
but that would fuck up when i wanted to restore to a test VM
so imo probably best that way
you'll need to do a full host reboot after a restore btw. also we are in #the-water-cooler this is a shitpost den
do you guys think a 90* adapter for the 12vhpwr is a good thing to buy?
eh i see its 3-4 weeks delivery time lol
Has anyone used the Govee Outdoor Permanent light? And if so, have you added your own custom jumpers to go from soffit to soffit? If so to Q2, have you had issues with the last few lights now working?
I only added about 25 feet total of jumper wire, and also didn't use about 8 of the LEDs, and I am having issues with the last 5-7 LEDs lighting up or even being recognized in the Govee app.
I bought a different power adapter that is still 36V, Sam as the original, but has 144W of output, double the original power adapter which is rated at 72W. I was thinking this would give it enough umph to let the whole strand, but it is dying off at the same spot.
Any suggestions of what to try next?
if you want a reliable led strip setup with proper power injects you build them yourself you dont buy rebranded chinese shit at a premium (govee)
I hear what you're saying but it's also naive to think that everyone has That's required to do that or the bandwidth to learn it. Some people (like myself) like to tinker with home assistant and not necessarily build everything from scratch. Not that I don't want to but I simply don't have enough time to learn everything about it.
So my question wasn't designed to be a debate whether you should use pre-built versus homemade strips, there's a question of if I'm missing something about why the rest of the strip isn't lighting like it should.
its kinda caused by the voltage drop over the lenght of the strip
you could try and see if both halfs will run properly in parallel and if so, add some injection in either the middle or the end
I had someone else suggest powering them in parallel but they did not mention anything about any sort of injection.
The ultimate LED strip power injection guide Back to article index One of the most popular topics with addressable LED strip is power injection (or sometimes also wrongly referred to as voltage injection). This is the practice of running extra wires next to your LED strip to add extra power injection points. Especially with 5v […]
its effectively trying to achieve the same, but in a slightly more flexible way
Mostly depends if your case has clearance for a normal cable or not. In general additional adapters are additional potential points of failure, but it's better than mashing the cable into your side panel.
i kinda wonder why those strips not have that stuff in built
Because they prey on naive people who think they’re buying a premium product at a premium price
It was kind of strange to me that it happened with the default set. You are supposed to be able to add two more strands on and I actually eliminated lights and just added some jumper, I figured I was actually going the other direction of if it would work or not...
Check solder joints or clips it’s most likely just a bad connection on data wire
i mean if their product page says that will work, it should
If he cut them to length and attached wires to jump areas with no leds there’s just most likely a bad connection
I mean, of course you are not SUPPOSED to modify them like that, but I did do some research before and saw more than a few people wrap far more length that I did around their house with them.
I’m not suggesting you did anything wrong
I soldered the first time, but the uncontrollable nature of doing it while on a ladder outside led me to believe I just got crappy cold solder connections so I clipped them all and did crimps with the same results.
I only use wled open source app and firmware but we have to go in and change led count. You’ve done that but leds are dead right? It’s possible you bridged power and ground and zapped one. Cut the first led off that isn’t turning on and try again
Usually it just zaps one led if that happens
I'll take a look at the link you sent earlier and see if that leads me anywhere before I put another cut in the strand.
I linked power injection cuz at first glance i thought you were talking about brightness and colors at end of strip not being the same as the beginning
Cute pupper
Can confirm, that indeed is a good pupper
gonna add some HAAS to my HASS so I can HAHA more
And evil HACS
I see your pup and raise you with a homemade egg McMuffin eating pup
can also confirm, that is also a good doggo
got a general ssh question someone might be able to answer - I have Server A which can locally connect to Server B. I'd like to access server B via SSH but through an open port that's open on Server A, but already has sshd running. I think the solve would be an SSH tunnel if I understand them correctly
does that sound right?
yesish
the tunnel would be from you to A so you can connect to B
pretty much like a proxy
well, that's kind of how I'm operating rn - I can SSH into Server A because there's an open port, and that machine can SSH into B. Ideally though I'd like to SSH directly to B through a tunnel I think. Trying to SSH git pull a repo on Server B
B doesn't have a port forward
It's only accessible on the LAN but it's a different port
another option would probably be to forward all traffic from some port on A to ssh on B
but that may be a mess to configure
yeah... my goal is to not need another open port... I think if I configure the tunnel to bind to the open SSH port on A, it should work because the user managing git doesn't exist on A, so I think it would try B?
no
you cant really have two things on one port and have it figure itself out that easily
so I'd need to kill sshd on server A
that would kinda kill ssh access and the tunnel
yeah
you may just setup a dedicated ssh profile which will ssh into A and then ssh into B over A
or the tunnel solution
or forwarding via iptables
i dont think there are more viable options
i think I got something... on my localhost:
ssh -L localport:serverB:serverBSSHPort serverAUser@serverA -p serverAPort
telnet localhost 1234:
SSH-2.0-OpenSSH_9.2p1 Debian-2+deb12u1
telnet was just to demonstrate that it's pointing to the correctp lace
i see
didn't clarify that debian one is Server B
and yeah ssh localhost -p localport is gucci
localport is 1234
appreciate your help 👍
heh, new problem - git cloning a repo from localhost:1234 prompts for a password
i think i'm missing the ssh key
Sort of off topic, but I got notified by my router today of a suspicious connection to my home assistant, at the time my DNS service was completely stopped so I don't anticipate any connection outside my local network. Any thoughts? The conspiracist inside me says Netgear is making these up so I eventually pay for their security services
I'd reverse IP search that
It's probably just a bot
^
My logs are full of bots
If it's a datacenter it's either a bot or someone doing something nefarious
They're trying to poke holes
Had someone hitting me from an Amazon AWS datacenter from multiple IPs
Reported to Amazon and never saw it happen again
I guess I assume they were doing nefarious - they were hitting my authentication page with incorrect password attempts
HA was sending me notifs
if you're not already, fail2ban on home assistant host
I'll look into it, thank you
only open absolutely necessary ports
reverse proxy is a good way to keep things local
er, map local things to outward facing things without exposing too much
I'm always on a vpn but I also have nabucasa
does netgear tell you anything specific about the connection, like the port trying to be accessed? @west trout
yeah
While looking for the log, I read that Netgear has a lot of false positives
For their security attack detection
Damn log already wrote over itself
No
makes me wonder if they actually have one set 🤔
what is a "root server" ?
do they mean their dns root zone server/s ?
or do people nowadays setup a server just for the root user? whats the purpose?
EXPLAIN
the servers piper runs on
Good old times, when we were able to unleash global thermonuclear war from home. RIP WOPR.
Would you like to play a game?
I really love those breakout things
Damn!
Those ones are the generic volume licence keys
They don't work fully
It's not that amazing
gpt-4-turbo:
gpt-4:
heres the whole recipes just in case anyone actually wants to read them
Both recipes only have 10 herbs and spices 💩
That's the most I could trick out of it too. Unless you count flour.
did you sit on it?
giggity?
Hello. Hope everyone is well!
I hope I am using the correct channel for my message, apologies in advance if I don't.
I am looking for a developer who can help me build a custom integration for home assistant that integrates with an ip gateway by sending a set of hex commands over udp to steer a gateway.
Pls message me here if you are interested or if you have any advice on whom to reach out to.
Thanks,
Magnus
heh, I'm getting somewhere with my drive pool
/dev/sdba
➜ /tmp lsscsi -g | grep disk | wc -l
54
😨
So how many digital racoons is this rated for
160TB of rust, 46TB of SSD, 14TB of spare rust
rust... must be blazing 🚀 fast :^)
nice, what are the specs?
MK001920GXAWR
DDYE1920S5xnNMRI
ARFX1920S5xnNTRI
SSKN1800S5xeN010
MB008000JWWQP
I meant the CPU, RAM etc but nice
All of them!
Dual Intel(R) Xeon(R) CPU E5-2620 v3 @ 2.40GHz
128GB RAM
6G SAS Controller
Duan NVMe boot
10Gx2
1Gx6
Anything more ?
Need some more memory soon though, this is the production system
I have 5 dual 2695 v2 systems with 384GB RAM each
that's an oldie. But nice amount of ram
Good enough for labs
and with power being normally free, not too worried
but prices have been bonkers last few years
so hopefully I'm moving soon
For reference @primal laurel
nice 👍
@clear ferry I converted your message into a file since it's above 15 lines :+1:
bunch of stolen car radios

do you measure how much power that rack uses by any chance? 🤣
About the same amount as an average houses uses in a month
enough to use these
The 48kw plug yes
The production stuff is 800ish watt (DL380Gen9 + 2 full disk shelves + mini itx + 2 switches + edgeline 300
Which equates to about €25 per month
And yes, most modern norwegian houses have 3phase power
My house only has 3phase 230v, but post 90s do 3phase 400v
I think this would cost around 6 bucks a day here
Oh, my math was wrong, it will cost me €2.5 per day
With current prices
with <2020 prices it would have been €0.5 per day
So roughly the cost of a netflix 4k subsciption today
which I have cancelled
🤔 isn't that just whether you measure phase to ground, or phase to phase?
No
The distribution is 230 or 400
Then from the house breaker we then do either L1+N+PE \ L2+N+PE \ L3+N+PE if it is 400v in, and different portions of the house uses different phases, but if it is 230v distribution it would be L1+L2+PE \ L1+L3+PE \ L2+L3+PE
If memory serves right
So my house as 230v in, so I'm limited to 7.4kW EV charging, but if I had 400v I could do 22kW
you mean 7.4kW single phase?
You would only get 400V if you connected your load between two phases. Single phase to neutral would get you 230V
Yes
House is 230 3ph in
and all is 1ph from there
Modern houses is 400v 3ph in
so you can do both 1ph and 3ph from there
just do 3x 7.4kW charging
Not allowed
who will check
EV charger has to be installed by certified electrican, and I for sure won't touch it

Also, since they are all cloud connected, anyone can check later
duh, just disable wifi
They have 4G too 
Pretty much yeah
No switch for the incoming electric though. Meter always has power, there are 3 fuses in the box that feeds it though, same box that does my neighbours
If the neighbors don't see the wire going to their yard, does the theft really happen
What theft 🤣
Exactly
If you're in the EU your phase voltage is always 230V
Phase to N, yes
With massive tolerances
Yes. So not sure what is meant by newer houses have 400V, older have 230V
I wasn't sure what Atx meant either, I have 400V between L1 and L2 for example, but 230V from L1 to N
+- 10% is the law
Well that's what, 50v difference
It may be legal but that at least sounds like quite a lot
No 23V. From 253V to 207V
It's to account in variances between regions/locations, not day to day variances for one spot
All devices are designed to be operable in that region without any issues
Yes
idk about Europe but in the US apartments frequently use a lower voltage for two phase, too, 208v vs the 240v in homes.
What do you mean? This happens mostly do to the fact loading and generation change all the time
It's also a factor of how far you are from a substation
If you're seeing a full 10% swing from day to day in one spot there's probably a problem somewhere
use lead shielding ^^
It's pretty normal here, due to the large amount of solar panels
I am even hitting 253V on very sunny days and then the inverter shuts off
Not sure what the gap with no data is from but voltage here is generally rather stable
And that range is only that wide because that sensor has a fishtank heater on it that turns on/off frequently
It does not very much at our house as well, located in germany.
from last summer
this is what massive solar does to the grid
huge amount of power available, nobody to use it. Power becomes worthless
Sounds like a good thing tbh
it can be very profitable for those on dynamic contracts. I expect prices to go towards minus 1 euro/kWh on some hours in the summer
There's some solar around me, mostly gas power stations though
I counted the roofs here once, it's about 50/50 solar vs no solar
probably higher now
Not many homes have solar here but there's a few fields where they laid out solar around
Basically no wind power where I am, too mountainous
Solar is great. Very cheap, lasts a long time
Our government made it so profitable its basically a no brainer
Cost recovered in 3 years or less if you do well
There's solar rebates here but I'd have to convince the city to cut nearby trees down or my roof is in the shade most of the day
And I don't really want to lose those trees or think they would cut them down anyways
It depends. Usually cutting some branches is enough
I have also seen that in the cold season you suffer much less from shade from trees because the leaves fall, but obviously you also have less sun
In the summer the sun is much higher in the sky, so in general you suffer less from shade
But I have modelling tools to check that for me haha
anyway, this afternoon I have been catching up with cancelling my technical debt as I correct 300 mypy errors over 45 files...
GH copilot is amazing for this stuff btw. So quick
I must say, first I was a bit hesistant
But then I had to write an email regex validator for an input field
and that was the moment I was like, let me give this a shot
and I've been working with it since and its seriously good
I was also hesitant because of my experiences with chatgpt which weren't very good. But GH copilot always works in the context of your project, which I think makes it work so well
hi folks 🙂
I'm writing Master thesis about home automation and I need to collect some stats from the community
here is my questionnaire: https://forms.gle/oiojZFMGNdcBjQ3RA
I will appreciate any responses and feedback
hope for cooperation ❤️
This questionnaire aims to gather a set of popular smart home automation scenarios. The collected data will be utilised in the master thesis titled "NetDaemon Automations for Home Assistant."
If you don't currently use any form of smart home technology, you can respond from the perspective of "If I had a smart home, my answer would be..."
NOTE: ...
Stable, but low. All the UPS's in my house would trip if my mains voltage were ever to go below 114v.
My outlets read 123v on average, with a maximum variance of +/- 2 volts.
114 seems high for a UPS to kick in at
114v is when it starts regulating voltage back up to a stable level (transformer running). 106v is when it trips completely to battery power.
APC default for Smart-UPS units
I'll probably buy solar when I buy a new house
The voltage in this graph would basically leave the transformer and fan humming all day, which technically also shortens the lifespan of the unit.
Sounds like a bad UPS configuration
That's how all line interactive APC UPSs operate...
They'll also start running the transformer if line voltage is too high, and trip to battery if line voltage becomes excessive.
If your stuff is destroying/aging itself because it's at 114v instead of 120v it's a crappy unit
It's only aging because the unit is basically running constantly if line voltage is below 114v, which is intended functionality
And it's better for it to run the transformer than it is for it to run from batteries, because batteries age a hell of a lot faster than the other compontents in the UPS
Sounds like a waste for just powering home computers
Modern PSUs are happy down to like 85V
I mean, no waste here. My line voltage is usually well within the butter zone, so the UPSs sit idle all day
85v specifically is when a UPS will detect a brown-out. Very short blips down to 50v are technically allowed by the grid.
I think mine trigger brownout protection at 90v
It's usually adjustable within a range
But a modern computer PSU will run happily on a constant 85v
Usually the issue we have here is line voltage being too high. I have a client right next to a substation, and the transformer in the UPS in their server rack runs almost constantly because the line voltage likes to hover between 126 and 130v.
Poor UPS is regulating down to 120v all day long
If my line voltage were 85v I'd be raising hell with my power company. lol
And I'd fully expect a UPS to compensate for that
Yea, the UPS should be on well before 85v, but 114v seems high
Transfer voltage is 120v +/- 5% by default on APC UPSs
114v would be pretty low voltage for the grid in my area.
You could reconfigure a UPS to deal with that, but it's not gonna be happy out of the box. lol
Anything is possible when you don't label the axis on your graph...
Also when you're not afraid to make a GPU take a kilowatt
Eh, Nvidia has some sanity in that department. The RTX 4090 is perfectly capable of drawing 1kW, but the performance gain of a few percentage points over 450w isn't worth it, so they capped it.
a kilowatt seems about right
Granted, they could have capped it significantly lower than 450w and still retained 90% of the performance.
The 4090 is capped at 600w out of the box
As long as you have the right power connector
If you only have a 3x 8 pin adapter it'll cap at 450w
Then TechPowerUp has the specs wrong for the 4090
They list it as 450w for the founders edition
By capped I meant what you can configure it to take, not it's default limit
But it's just a power draw slider
But as discussed, there's literally almost no point in adjusting that slider.
Even flashing it with a 1kW BIOS doesn't net you more than a couple percentage points
You also start to get paradoxical effects where lower power targets perform better if your cooling can't keep up
There's 5lbs of copper and fans on my 4090, it better keep itself cool
That is in no way an exaggeration either
I already get inverse scaling on a 3080. lol
I run my power target at 96% because it actually performs better that way
I couldn't get my 4090 to thermal throttle, it always started complaining of an unstable power supply at around 580w draw and refused to go higher
It gets really loud though
It's not that it was thermal throttling, it's that it would run slightly warmer with the higher power target, so it'd drop a boost bin or two at the worst possible moments.
Reducing the power target led to better 1% lows. Go figure 🤷♀️
I upgrade my gaming PC every 4 years or so. Next gen the main differentiator is going to be perf/W for me lol
I'm not spending this much on a GPU again, that's for sure
Only got it for VR anyways and game devs seem to have given up on VR
I bought my 3080 in the middle of the GPU shortage for $999
Also for VR, because a 2070 wasn't cutting it.
I sold an RX 5700x for $700 and got a 3070 during the shortage for $850, later sold the 3070 for $300 and got a 3090ti, found nvidia screwed up displayport on the 3090ti and it wasn't going to work reliably with my headset and got a 4090 instead.
Though at least I also have triple monitor to use the 3080's horsepower with. Driving 5760x1200 is pretty demanding in its own right.
My VR headset screens are 7680x2160
Speaking of DisplayPort, I'm still pissed that Nvidia reduced the DisplayPort count on their consumer cards from 4 to 3 after the RTX 20 series...
I had 3x DisplayPort monitors + 1x DisplayPort headset connected to my 2070 no problem.
Want to do that with a 30 or 40 series? Gotta go Quadro.
Or get an mst hub
Doesn't work, tried several
The Valve Index absolutely refuses to work with an MST hub, as does Nvidia Surround / G-Sync
I have gsync through one
I could technically get output from all three monitors using an MST hub, but they wouldn't create a display group.
Never tried surround, and yea, an index won't work
The Nvidia Control Panel just wouldn't see the monitors connected to the MST hub as avaiable for Nvidia Surround, at all
I have a USB/Displayport KVM switch that acts as a MST hub too that's been solid for me
Weird
Apparently it's expected. Nvidia doesn't certify Surround to work through any kind of hub or daisy chain.
There's HDMI to Displayport adapters too
G-Sync doesn't work with those, though Surround does.
https://www.amazon.com/gp/product/B0BYJP8QQG I have this one, never tried gsync or surround through it
And the Index also won't work with one of those
Index/any other VR headset is going to want a direct displayport connection
Yuuup
The Index doesn't even like being on a DisplayPort switch
So you can't even sacrifice a monitor to plug in the Index in a semi-automated fashion, gotta physically swap cables.
https://www.lindy-international.com/35m-DisplayPort-1-4-Repeater.htm?websale8=ld0101.ld021102&pi=38415 I tried a half-dozen displayport extenders and ended up importing this from some store in Germany because it was the only one I could find that'd work with my VR headset
: Extends 8K UHD DisplayPort signals up to 12m or 4K up to 35m
Cost like $100
I don't need a repeater, I just need my card to have a 4th DisplayPort output likethe 20 series did <_<
I've currently downgraded my right monitor to HDMI, which works, but isn't great.
Isn‘t usb-c -> DP a thing?
Yea, I was just saying the VR problem extends to repeaters and extension cables too
Yes, with an asterisk.
USB-C can do alt dp mode and carry displayport but it needs to be fed from a displayport source.
Only certain USB-C ports support DisplayPort output. The RTX 20 series had one such USB-C port onboard, which is where the 4th native DisplayPort output comes from.
You can get USB-C cards with a Displayport in that'll give you USB + Displayport out over a USB-C cable
Yeah, but you have to connect a DisplayPort output from your graphics card into the USB-C card for that functionality to work
They'll have a DP-in port for that
Yup.
My motherboard has two USB-C ports with DisplayPort support, but they connect to Intel Integrated Graphics, which is useless for my purposes
You can with PRIME on Linux but idk about on a Windows desktop
With a latency hit, and good luck getting VR, Surround, or G-Sync working like that...
but it is possible, you just don't try hard enough
idk about surround but VR and G-Sync won't go through the igpu at all
I guarantee you I can't make a Nvidia Surround group with a display connected to Intel Integrated Graphics
Right, a 4th native Displayport output on the primary GPU is required here
Otherwise you run into issues with simultaneous support for a VR headset AND a triple-monitor setup with Surround and G-Sync
Get an ultrawide and reduce your screen count
Most solutions for connecting three DisplayPort devices to an RTX 30 or 40 series card result in something breaking
Or Nvidia could stop nerfing thair consumer products. Remember when you could run 5 displays off one card? GTX 1080 was their last consumer card to support that. Limit is 4 now...
And 4x DP outputs wasn't a problem for the 20 series. Could have been included on the 30 and 40 series, but they decided to restrict it to workstation cards.
I'm fairly certain every consumer nvidia card has been limited to a max of 4 displays
Nope, Nvidia supported 5 displays for quite a while
Sadly they removed the USB c port
5x portrait monitors in an Nvidia Surround group was a setup some people ran
They definitely didn't on the 900 series and everything I can find online indicates they didn't on 1000 series either
And I know 2000 series and later is all limited to 4
It's a bullshit limit imposed for market segmentation though
Quadros will happily do a dozen screens over a MST hub
Here's an GTX 780 doing 5 displays
That's three titans
Pretty sure 5 displays were supported from the 200 series through the GTX 10 series
That literally lists three titans
Thanks google. lol
Also comes in handy for any AI stuff 😅
Google image swearch result doesn't match the search critera. My bad 😛
I had two 900 series cards, they were hard limited to four monitors per card.
I had nvidia optimus working between that and an igpu so the igpu output was rendering on my 970, even then it would only activate four displays. I tricked it to get around it by making the last display my final monitor + the monitor I was putting on my igpu and making the framebuffer that got copied to the igpu start half way through the final monitor
I know AMD supported 6 Displays since the HD6000 series
Which is somewhere around GTX 600 iirc. And AMD was slightly earlier on consumer cards
So it thought that my final monitor and the monitor on my igpu were clones but the 'clone' was offset
That was a weird time. AMD technically did triple-monitor spanning before Nvidia, but Nvidia eventually enabled it on much-much older cards than AMD.
I just need 3 x 4K@60Hz and im good
Offset clone us weird, since the limiting thing is usually Crtcs 🤔
So GTX 200 series SLI owners suddenly found themselves able to run triple-monitor games
my first Nvidia Surround setup was two GTX 260's in SLI
Tripple was supported a bit longer
I mean actual 6 CRTCs
Though it required DP MST which was rare back then
It was a buggy setup. Normally the display would pan to the end of the fourth monitor but I found if I used xinput to move my mouse cursor in a certain pattern it broke that
So I had a script that'd bug out the screen scrolling so it'd lock at 0,0
Huh, that's a very interesting thing
Because that's more Crtcs then
The panning isn't done by the card.
Should have been possible to use independent entirely
Though maybe not exposed properly by some older config
Yeah, AMD had their "Eyefinity 6" card with a ton of DisplayPort outputs onboard
I think that thing was intended for video walls, though
The first four screens would pan, the fifth offset clone one was always locked where it should be
KDE had a feature where you could define custom desktops that didn't align with detected monitor geometry, I used that to get the fifth monitor to work normally
I think that was under fire pro brand
That's a Radeon HD 5870 Eyefinity 6 Edition
I had an nvidia card with 8 at work
I just want 4 back 🥲
Get two cards
That would be incredibly expensive...
Buying a second 3080 just to run the VR headset because Nvidia can't be asked to put a 4th DisplayPort on the card is kinda stupid
Especily now that SLI isn't an option
Get a cheap nvidia card and put a non-primary desktop monitor on it
Should be able to do surround across cards
I keep telling y'all, that doesn't work. Can't make a Surround group like that.
There's some for a couple tens of euro
Render will mostly only happen in one either way
You don't need 2 powerful cards
You posted a screenshot of a group spanning 3 cards earlier
A Surround group that spans cards requires SLI
And the 30 series doesn't support SLI
Don't use a toy OS 😂
yes, older cards that supported SLI
When Nvidia dropped SLI support, multi-card spanning went out the window. Only Quadro/workstation cards can do it now.
Technically the 3090 supports NVLink, like a workstation card. NVLink can emulate SLI as a protocol, though.
I used a hacked nvidia driver to SLI two different 970s around 2018, I wonder if anything like that could still be done
Ignoring the issues with getting G-Sync and Surround working, how do you solve for the latency hit of transferring the framebuffer over PCIe?
I've done the "dGPU outputs through the iGPU" thing before, and there's always noticable latency
Which would lead to the 3rd monitor being de-synced from the first two
Yeah, Nvidia Surround merges the monitors into one giant monitor in hardware.
OS sees one giant ultrawide
I know what Surround does
They are async either way
Clock domains of scan out engines aren't synced
Though the latency would be a bit higher
I still don't want one monitor a frame or more behind the others. My HD 6970 did that in Eyefinity, and it suuuuucked
If you want actual synced scan out, there's some proprietary Intel tech though
If you used mismatched outputs on the HD 6000 series (like 2x DP and 1x HDMI) the HDMI monitor would be a frame behind and/or experience constant tearing even with V-Sync enabled. It made turning feel "wobbly"
They use it to drive 8k Displays
Because they didn't fit the pipe of a single scanout engine
There's a program for VFIO to stream a display from a gpu assigned to a VM to a GPU on the host through a shared memory block that from what I've heard is <1 frame latency, idk why the OS wouldn't be able to equal that.
I've also used software-based triple monitor, and the latency on monitors that aren't on the primary card was awful
Because vfio uses a buffer in an address space the host os can access if told how
It can 0 copy
SoftTH actually works pretty well when you can connect all monitors to one card, though
Yea, ivshmem
I just don't want one wonk monitor, which Nvidia Surround seems to avoid
Scanout on a different GPU necessarily has to transfer the framebuffer
Vfio allows to transfer a handle instead
For looking glass it renders on the guest gpu then reads it back into system memory to be rendered on the host gpu
🤷 back in the day my displays were heterogeneous enough for it to not matter
And now a 3840x1600 or 4k display us more than a GPU fan handle either way D:
Using the same api that steam link/nvidia game streaming use to capture the current image
Not sure what you mean. I'm running three identical monitors. Heterogeneity doesn't enter into it 😛
I think he means there used to be a lot less variance in available monitors so there was a better chance everything just worked
heterogeneous == diverse
Lol, yup, I read that as homogeneous
Display tech has gone through some weird periods. Early home computers using composite output, the various RGB output standards that followed, then VGA became the standard for a very long time before DVI / HDMI / DisplayPort hit the market.
I'm surprised DisplayPort hasn't completely dominated, given its capabilities and lack of associated license fees.
mda, ega, cga, vga on just the IBM side
DisplayPort is even cheaper to implement than VGA, which is pretty crazy
What's the bug?
Get thee to a Blockbuster!
or your favorite bittorrent client, your call
Ever see Once Upon a Deadpool? Deadpool kidnaps the kid from The Princess Bride (as an adult man), and reads a PG-13 version of Deadpool to him.
They even recreate the bedroom from The Princess Bride
I knew that movie was a thing but didn't have any desire to see a PG-13 deadpool after having already seen the normal cut
Yeah, I get that. All the cut-away gags back to the bedroom a worth a watch at some point.
I just never really had any desire to watch that one
The hotplug detection is underspecced
Because it allows to virtually disconnected displays that are turned off
And it leads to all kinds of issues
With some disconnecting in various DPMS states
Worst we've seen, some even disconnect on modset because they technically turn off for a moment
DisplayPort monitors not waking from sleep in certain configurations is a really common issue
This sounds fixable, though.
Likely because the OS thinks they are disconnected and doesn't try to wake them
/me implemented some of those bugs in os side handling 😄
Thankfully haven't had that issue with any of my DisplayPort monitors, though I have had that issue with various DisplayuPort-to-HDMI adapters
Having to disconnect and reconnect the one monitor on an adapter got old real fast
Yea, passive/active
And active without external power can be super iffy
It was an active DisplayPort to Dual-Link DVI adapter with external power that was the most flaky actually...
Anything advertising above 1080p@60 active without extra power is either straight lying or unstable
I had one of those
Earlier days of DP, and hardware was just generally worse
Or i bought too cheap because I was still a student
But yeah, haven't had issues since avoiding adapters. I figured it was just an issue with the adapters being cheap pieces of crap.
I'd just check the supported hardware list of my previous employer these days
Don't know if our PM actually ever did add it, but we had different supported resolutions and refresh rates by revision (approximated by manufacturing date) in dev...
One of the cooler things that DisplayPort supports, that nobody uses, is software defined display coordinates.
DisplayPort spec allows you to carve up an attached display into any number of sub-displays, but literally nothing supports it.
I'll even see 4k monitors and ultrawides that support splitting the screen into quadrants, or in-half, but they do it using multiple cables rather than the built-in features of DisplayPort
Pretty sure the ones that do it in one cable do it via MST
MST is its own thing, seperate from software defined displays
Do you have a link for that?
Never heard of it, and a quick text search for "software" in the DP wiki article shows nothing
MST basically is software defined displays
No
MST is multiple transport streams
MST is basically two cables routed through one
Yea, it adds a second channel in an existing stream
MST is Maastricht Aachen Airport. You know that I‘m right
Mountain Standard Time
I'm not sure it's an explicit feature. My understanding was it was inherent to the technology, and how display surfaces can be arbitrarily defined from the available total screen area.
MST is usually used when there are multiple distinct display devices in-play (either using a hub or through dasiy-chaining)
I've not seen a monitor that uses MST to turn itself into 2 monitors, though I have seen monitors with two inputs that can simulate being two monitors.
Why would one monitor use MST to turn itself into two?
MST doesn't increase the available bandwidth
splitting the monitor up is a pretty common feature on 4K monitors and ultrawides, actually
Yea, from two disparate sources
Displaying 4x 1080p streams, all at full resolution, on a 4k display
You wouldn't do that from a single computer though
So you can have one monitor appear as two, obviously...
why
That definition being possible sounds weird
are you sure that's not about scanout Planes on the GPU? They aren't really realted to DP though
So you could have two 1080p displays instead of a single 3840x1080 display?
With one cable, instead of two
For OS reasons.
It's a dumb feature
For the same reason you'd run two monitors instead of an ultrawide?
Window management often works better that way
Usually exists because it's 2 panels either way
Either to support older standards via 2 inputs, or just weird tech
There are so many better solutions for window management than making your monitor appear as two screens using MST
Being able to switch between "two 16:9 displays" mode, and "one 32:9 display" mode, is handy
When I get an ultrawide in here, I'll absolutely be connecting it twice so I have the option of making it simulate dual monitor
I get the point of using two cables to an ultrawide and using surround, one cable is limited by bandwidth and starts using DSC
At least at 144Hz+ refresh rates
But one cable to an ultrawide using MST sounds like the worst of both worlds, lol
In the example of splitting up an ultrawide, bandwidth limitations aren't a concern. The cable is either carrying one 3840x1080 stream, or two 1920x1080 streams. Basically the same bandwidth either way.
The difference is the OS sees one monitor or two, and I didn't need to run two cables
But I have yet to see a monitor that can do that using MST. They all require 2 cables 🤷♀️
There's ultrawides out there like the 5120x2160 ones that exceed a single DP cable's bandwidth so a second cable is the only way to get an uncompressed image on them
3840x1080 will fit on one cable
Latest DP spec handles those resolutions fine now
Unless you run a very specific config, they will tear between the two images though
It's a dumb solutions to run
A 4090 only has DP 1.4.
No other solution if you want to run a game that requires exclusive fullscreen mode, and don't want it to take over the entire monitor.
Being able to spoof that you have two display devices comes in handy
If you run it as two displays, it's fine
It's really just running one display on two scanouts that's dumb
I couldn't even get nvidia surround to activate on my ultrawide when I tried, whenever I activated it it got the resolution wrong and nothing I did would get it to go to the right resolution
Well yeah, LinusTechTips did a video about an old monitor that worked like that
And it did have tearing
I can't imagine gsync would work like that either
I think it was a really early 4k monitor that required multiple inputs + a computer that supported spanning, and they couldn't fix the tearing at the seam.
Eh, I've seen it, it's dumb though
That was the monitor they tested with that like 500ft thunderbolt cable I think
Three ultrawides spanned is kinda unusable
Yea. I know that Intel has tech to avoid that
But 🤣 Intel GPUs and use cases that care
I'm pretty sure I was hitting EDID stupidity
But the entire point of wanting to emulate multiple smaller displays with one large one is to not span things
If I wanted to span things, I'd leave it as one large display. lol
Yea, it's a hardware issue
They need to sync the scanout. At least the vblank
And that's not exposed to even the driver
I wouldn't hold your breath for a display with MST built-in
Too niche
You could get an MST hub and put it right before the two inputs on the screen though
Like I said originally, I want completely arbitrary software-defined displays, not MST. MST is a poor substitute with limited utility for this use case.
Yeah, that'll let you switch to the the split-screen mode with one cable, which is already handy in its own right
There's a powertoy for Windows that'll let you redefine/define custom window snapping boundaries
FancyZones
Couldn't find anything pointing to the feature you reference
Not the same. I want to be able to define a set of screen coordinates and say "that's a monitor"
KDE can do that
And it gets its own resolution, color depth, and refresh rate
I'm not sure KDE can do all that...
Neither can any hardware
The refresh rate and color depth are up to your hardware, all the software is going to do is define a custom region for a monitor
But you can use it for things like getting multiple task bars on one monitor/making windows maximize to only half the screen
Actually, you could do this on a Commodore Amiga back in the 80's. lol
It was limited to changing display modes at every scanline, so you were limited to multiple horizontal monitors per-monitor, but it was still pretty damn cool
Sounds 50x more useless on an Amiga
Nah, it was really useful when you had older applications that needed low-res screen modes on an Amiga running a hires desktop
You could pull a 640x480 monitor over a 1280x960 monitor seamlessly
No mode switching, it was just changing right at the scanline between the two modes
With features like that how did they ever lose
If you had a PC card installed, you could even pull DOS (or if you had a fast enough one, the Windows desktop) down (running at its own resolution, refresh rate, and color depth) over your Amiga workbench
I imagine that wouldn't work at all on any monitor that actively monitored the sync
Are you sure that didn't run at the same rates towards the display and just converted different buffers locally?
And it was all happening in hardware, not with windows or a window manager
That could only ever work on really dumb CRTs
Yee, the Amiga did all kinds of per-scanline fuckery with its custom video chipset
The screen geometry would be way off though
The feature only works on Amiga CRTs with the Commodore RGB connector. You'd lose the feature if you installed a PC video card and used a VGA monitor.
I'm fairly sure it was still just RGBHV
True enough, it's similar enough to VGA that it'll work on VGA monitors that A) support slow scanrates,and B) support sync-on-green
Oh, it's RGsB
That's speaks for source fuckery
I'd have to look up the specific standard, but I'd be very surprised if any supported changing sink side timings mid frame
It supports changing things mid-scanline, so...
Changing resolution half way through a frame like that would surely cause weird problems like the screen bouncing though
And yet, it does it. Here's it running DPaint in lowres and the Workbench in hires:
Looks like it's only changing the horizontal resolution to me
Which is a lot simpler than actually changing the full mode half-way through
Probably just pixel doubling though
Both the horizontal and vertical resolutions are altered
For the screen that's still just one image
Commodore referred to this tech as "ViewPort" if you want to look into it
not really. It's not super interesting. Unless you have a direct link supporting your claim that it did something that shouldn't be possible, I'll assume you are wrong
I see nothing about this that indicates it's changing the mode being output to the screen in one frame
Seems more likely it's using pixel doubling, all the resolutions are multiples of the step below
Pixel doubling wasnt a thing on the Amiga, it would just run a lower resolution...
There was literally no need to pixel double a resolution on an analog CRT
What you're describing isn't a thing.
You can't just change the frequency of your sync signals half-way through drawing a frame
At the best that'd give you a really shaky image
BBC B did something similar from memory
Flipped the mode part way through - Elite did that
Then how do you get images like this, with 320x256 progressive-scan on top and a 640x512 interlaced on bottom?
You may be able to increase the horizontal sample rate but that's all you'd be able to change
that image doesn't show what you seem to think it does
Commodore has always described this as the mode being changed at the scanline at the boundary between the two screens
Each display mode specifies (among other things) a pixel resolution and a monitor scan rate. Though the Amiga has the unique ability to change pixel resolutions on the fly, it is not possible to change the speed of a monitor beam in mid-frame.
Therefore, if you set up a display of two or more ViewPorts in different display modes requiring different scan rates, at least one of the ViewPorts will be displayed with the wrong scan rate.
Also, if it were always running a higher res and pixel doubling, a lot of software would balk due to a lack of memory. The lower resolution screen modes were required to access higher color depths.
You modeset to whatever looks closest as superset and do some client side processing
that's why you run a specific graphics/scanout hardware
Horizontal pixel doubling can be done as easily as lowering the sample rate on the horizontal data and vertical pixel doubling is just replaying a line twice
emulating that isn't super hard
Actually doing that over the wire is an entirely different thing
This clearly states that while you can do it in software the screen can't change modes so one will have messed up geometry
you can easily emulate that on modern hardware with quite a bit more freedom via planes
But you still modset the display to a single specific mode.
You absolutely cannot change your vertical sample rate half-way through displaying a frame/field.
The screen would lose tracking and start bouncing up and down
ever seen a modern TV think the 240p image of a SNES is a 480i image?
is there anything that works like kubernetes that won't take me months to learn? i want to have a pi be the backup for my primary server. right now i'm going in and starting the containers manually
This took WAY too long but i now have a template for scarping gasbuddy for regular gas prices (at stations of your choice) via Node-RED
kubernetes isn't even super useful for that usecase 😮
oh? interesting, i thought that would be a primary use case. pi only runs zwavejs and backup adguard right now
Yeah, I'd use it for clustering (with inherent failover) rather than as an explicit active/passive type configuration
it's not really suited for a 1 to 1 failover
you need a control plane (which can be HA, but that's slightly more involved) and then it can failover workers easily
ah, i suppose i can't run that on the server
if you run the control plane on the server and it fails, the worker won't suddenly pick up tasks
i haven't really looked into this at all
the other way works. If you set them both up as control plane it might 🤔 but I haven't done much with HA control planes
I finally wound up moving everything off VMWare and onto Hyper-V. I gotta admit, Microsoft does a MUCH better job with container management than VMWare does...
doesn't have to be tied to HA
I automate a bunch of k8s stuff at work. k8s is weird
HA in this case is high availability.
both of these are linux based machines
i.e. seamless failover
Wouldn't seamless be FT rather than HA?
lol, oh ok. misunderstood
Is it just about spinning everything up easily?
In VMWare terms.
Enabling "Fault Tolerance" on a VM or container means it'll survive anything as long as one FT-compliant host remains running. The VM or container is run, in lockstep, on at least two physical hosts at all times. You could walk up to the rack and yank a blade out, and the VM or container would keep running like nothing had happened.
Enabling "High Avaiability" just means the VM or container will automatically restart on another host if the original host dies. There will still be service interruption.
mostly i want automatic failover a couple containers
fwiw, I run my homeassitant in k8s, but and dynamically move zigbee2mqtt and mosquitto. But since I don't replicate the storage for homeassistant, that's bound to a node either way
interesting
You do have to think about any hardware you mide need double (zigbee coordinator?) and which storage to replicate as well
but it's not super hard to script
right, that's the downside. if the pi fails then i have no zwave
I'd go with a network-attached zigbee coordinator in this kind of configuration
to the worker kubelets it's noticeable, to the user it isn't 🤷
I'm not sure how a Zigbee mesh would work with multiple dongles (one per physical host) being used in rotation
yea, I have to buy one at some point
but I have two USB coordinators for now and wanted to play with that
it does. at least on C2351 (iirc. on that model number)
i'm thinking mostly of things like maintenance or, specifically, troubleshooting a bad HDD on my server. i've had all my media storage down for a week. haven't had time to figure out which one yet. i'd like to just reboot a live disk or something without worrying about my hvac setup
I know you can "clone" the dongles so they both have the same ID, which is used during migrations between dongles... you might be abel to abuse that
it's a single nfs mounted config dir though. And only ever one container running at a time
Yea, a lot of cheap-o adapters do that.
I didn't do anything to make them more clones than just flashing both with the debug image that's required to z2m them either way
Built-in composite/component inputs on TVs have sucked for a while, everything is 30Hz 480i to them
Interesting. I thought only a single coordinator was allowed in a mesh
Are all the devices migrating when you switch which dongle you use?
it's only a single at a time. But none of my devices cares swapping between the 2
yea. Behaves the same as restarting the pod on one hardware
heh, my entire mesh had fits migrating between dongles. I had to reset and re-pair pretty much everything.
I've been considering a new dongle. My door sensors keep dropping off and I think it's my conbee
might depend on the specific dongle. I think that debug mode pushes more into software than others would? 🤔
that's my understanding of zwave, pretty much limited to one device because node info is stored local to the dongle. but i'm far from know all the ins and outs
you could look into docker-swarm to failover btw. though I have no idea if that actually covers your usecase
i just know it's the "docker way" of clusters
I've been having a much better time with Zigbee since switching to a Sonoff Zigbee 3.0 (ZBDongle-E) USB dongle.
Almost all of my Zigbee devices work perfectly, it's just my door sensors and very infrequently a couple light switches
I was previously using a HUSBZB-1, and it would drop random sensors from the mesh all the time.
I think more of my door sensors are dropped off now than still online
my map is just a little stupid though
my understanding is that the conbee is basically a bad word in #zigbee-archived, lol
I definitely wouldn't buy one today but they were somewhat more reasonable when I first got mine
I need to move all my lights from the hue to the z2m at some point. But I'm laaaazy
Yeah, I need to set aside a night to do all of my Hue bulbs at once...
My Hue bulbs have been fine on my conbee
Doing them piecmeal doesn't appear to be an option. Even if I delete them from the Hue hub, it keeps trying to reconnect, which causes the bulbs I've paired directly with Home Assistant to go unresponsive, and blink several times, before reconnecting to Home Assistant. They repeat this song and dance every few minutes.
do you use the zigbee integration in HA?
Unplug the Hue hub, they all stop blinking and stay connected to Home Assistant no problem
I use ZHA, yea. My thermostat didn't work with Z2M.
Yeah, using ZHA
I'm a little perturbed that a un-associated Hue hub can still screw with my bulbs, though. I'm assuming the hub sends out some kind of magic packet that their bulbs listen for, even when paired to some other Zigbee coordinator/router
I think they use bluetooth to scan for nearby bulbs
They're waaay too far away from the hub for Bluetooth to reach
do you know how far btle reaches?
Hue doe ssome non-standard stuff
Also, some of these hue bulbs are so old they don't have bluetooth
yes, hue does nonstandard RACCOONS
Yea but it's not somehow hijacking another coordinator's zigbee network to talk to devices only reachable through it
because you're all way too much into #zigbee-archived
Strictly speaking, the bulb is getting kicked into pairing mode, where it blinks until setup is completed. But it gives up and reconnects to my Zigbee dongle where it belongs after a few seconds.
It seems like the Hue hub is broadcasting something that kicks Hue bulbs into pairing mode, regardless of their current connection.
better answer me how to get browser NOT turn to google if I give them non-standard TLDs (my domains)
prefix http://
whout that ^^ and without ports ^^ and without ending slash
all of these suck on mobile
https://bugs.chromium.org/p/chromium/issues/detail?id=30636 Chrome has that marked as WontFix
exactly
onsome browsers you can press the down arrow, and the suggestion box will have an entry with the magnyfing class for search and one with the globe to browse to the woname
The Hue dimmer remote can also kick bulbs into pairing mode, no matter what they're already connected to (hold the I and O buttons on the remote, at the same time, near a bulb. It'll start blinking)
So they've got some special thing the bulbs listen for...
what device and browser do you use? It just works™ for me
The first comment seems to imply all you need is the trailing / and not the full http://
oh wait no. I have a normal TLD
yes adding "/" solves that, still sucks
I have normal TLDs too, but I don't want to google my internal ones
I know some browser (probably firefox) has an option to set the mode
And either use a rough metric, or fall back to DNS and connect if it resolves, otherwise search
I hate them, all
Hah
I have the same annoyance with Firefox. It often auto-converts URLs to https://, and I have to explicitly change it to http:// and suffix the TLD with :80 to force it to use HTTP
I use thorium btw, seems more sane than any others to me but still sucks
Manage search engines...
Add a new search engine and use these as the 3 values:
a. Default
b. %s
c. http://%s/
Make default
LoL just use https
Just make a custom search engine for your hostname that redirects to your hostname. :P
I have to test that
I do, wherever possible. Cloudflare domain and LetsEncrypt all over the place. Some things simply don't use HTTPS.
It won't do the tld but it'd do a host
The web interface for Tasmota is one that's real annoying to get to if your browser keeps trying to swap to HTTPS
You can make it so entering a word in the search bar triggers it as a custom search engine, though now that I think about it you have to hit space afterwards anyways
thank you @prisma briar though for that workaround which breaks other stuff
I have some set up at work, I can do 'ip <ip>' to look up an IP address in our IPMI
I have one to look up customers too
how?
Personally, I've always kept both the URL bar and Search bar enabled. Both bars behave a lot more predictably when they don't have to try and guess what you're trying to do based on context.
thats why you have 4 monitors right?
3 monitors, and no...
I've been using Firefox since like 2005. It always had both bars enabled, by default, until recently.
Just kept with it 🤷♀️
If I type "google " in my address bar (including the space) it turns into the second photo and then takes me to the third when I hit enter
It's real annoying to type a search into the combination address/search bar and the browser randomly decides to open a bookmark or history item instead of searching.
yea, omnibox can be rather surprising
So I keep the search box enabled so I can explicitly tell it to search
And I have basically the same thing set up going to a PHP page I have at work that fills out a post form on our ipmi tool to do a search for an IP
Had a 3D print fail, lol
a 5 year old belt on the printer snapped and the x axis got stuck in place
It was clearly trying to escape
It tried taking the silicone insulation from the hotend with it, had to cut the stick of filament from the top to free it
It printed a mobile body and made a break for it
I put a 4 color mixing hotend on my ender 3 and was doing a test print with it, first few layers went down fine but I woke up to that.
It was supposed to be a few multicolor rainbows
This is why some printers will take a moment to hit the limit switches at specific Z heights
The location of the switch is known from the homing process, so if it doesn't trigger when the stepper is at the expected location, there's a problem of some kind
It's usually used to stop a print if there are skipped steps, but a snapped belt would get detected too. lol
This was the first time I've had a belt fail after probably thousands of prints
Yeah, like I said, it's usually used as a way to stop a print if an axis skips steps
Handy to have enabled for long overnight prints so you don't come back to a giant pile of spaghetti or a massive layer shift
vscode ➜ /workspaces/pvcast (dev) $ mypy pvcast
Success: no issues found in 27 source files
well, don't need help but ..something I never saw before ...and I find it a little weird
context: I have a ZigBee switch named "Fan" which, well, runs a fan. At the end of the season, I store the fan and unplug the plug. Since it generates something on logs all the time, I delete the switch from Z2M completely.
However, I don't want to re-do automations every season, so I just disable the parts of the automations that contain that switch, instead of deleting them.
Since couple of days, it felt weird, and I investigated what's going on with my HA. Apparently this disabled section was making the every relevant automation to "fail to setup"
so enabled:false doesn't work like commenting out a section
@boreal cave fedora is fine too. I just find the aur so useful, usually stuff is just there. They even had a version of forticlient for work
I loved the AUR, but rolling release got annoying
it was the wrong version, but don't let that get in the way of a good story, lol.
Just didn’t have time to tinker anymore
My laptop and workstation are set to only do security updates
I haven't run into a problem with rolling release yet, thankfully. I am not developing though. This thing is mostly a fancy firefox derivative display, mpv frontend, weztern frontend, obsidian frontend
Had time to update last weekend
Apparently I don’t upgrade much because I was two versions behind lol
Went from 37>39 and nothing broke
Had a backup laptop, it went from 35>36>37>39 no issues either
I work in DevOps so i value stability more these days
I'm kind of invested in this setup now. I wrote bootstrap scripts for yadm so I install it and run one script and nearly everything is right.
I put the stuff that has to be stable in vms on the proxmox cluster
Haha I get it, I had soooo many custom scripts and hooks for pacman on arch
Took a while to migrate to Fedora, ngl
are you using a dotfile manager in case things go awry?
I love yadm. It's made things so simple.
can you define devops?
are you the guy that makes the software go brr, or are you the guy that makes the software go brr on the distributed infrastructure or are you maintaining the distributed infrastructure?
I use ansible, lots of people find it clumbersome, but I work with it all day with terraform so I’m used to it for deployments
For home I just wrap it with scripts
as long as its working, it's a win.
I’m the guy doing the infrastructure haha
Deployments for k8 clusters
everytime I try to move to kubernetes from docker it screws me over somehow, lol
seems like the future though, even though it's stabbing me in the foot a lot
truenas just moved minio into their apps section on truenas scale, which meant it broke immediately on this latest update, lol
on the bright side, in a corporate setting that is a job creator;)
I have home stuff totally segregated from dev work, I don’t even run k8 at home just because the tools I use for home has been stable for yeaaarrrss
Don’t fix what ain’t broke
I run ESXi/vCenter to host VM’s
Had an enterprise license years ago from a past employer
I still quite enjoy tinkering. I'm semi retired (doing some IT work for the company I founded, upgrading from the podunk setup I was using from the original 4 guys up something more appropriate for the 70+ employees we have now). I hired a guy with the right credentials and use a few hourly people to fill the gaps and try to keep my work hours down to 20 a week (usually fail) but I do ahve time for tinkering.
My dev env actually lives inside esxi lol
I hate vcenter so much....
I got tricked into using that on our few dell servers
it's such a pain in the butt to do anything with it because we don't have a massive san backending the whole thing and all the high availability stuff in pla
we ended up going proxmox at our two satellite offices just because the experience was so much nicer int he context of running vms for stuff like domain controllers and other non intensive things.
head office has esxi and vcenter and I am counting down the days until I can rip it out.
When I was a sysadmin taking care of it, I scripted everything in powercli so everything gets pulled to a backup server, then into backblaze
we ended up using datto for backup
I think I go 5-6 months before manually checking up on it
It’s either total neglect or I trust it 🤔
But of both
Bit
my problem was more upgrading vcenter and esxi is a bit of a nightmare via their systems that take forever and their virtual tpm stuff, while seemingly very secure, contemplates having a control cluster to handle it for the main cluster when my total server infrastructure across 3 offices is smaller than the envisioned control cluster, lol
Did someone say vmware
I was in the middle of taking a dump on vmware....
Ohhhh god yes, upgrades totally suck balls
_cries in VCAP-DCV _
Just do VCF, automated SDDC

I've have very little issues of upgrading VMware for 16 years, I can probably remember 5 issues or so
Across a few hundred customers
If I was supporting 1k+ employees I totally get vmware though, because I would have a beefy, insanely fast san with redundant iscsi pathways and a seperate tiny cluster for tpm/mangement.
SANs and FC are boring. RoCE is where it's at
RPi
Disgusting
I have no idea what you guys talk, just my 5c
ok insert stupid expensive way to handle virtual machine disks that is very marginally better until you get big enough to enjoy economies of scale and assume I said that..
I mean "my 5 volts" kinda works here too, when you're talking about a Pi
I was more trying to say the bigger your setup gets the more vmware's stuff makes sense and at my business size it's just pain for no reason.
You forgot triple site federated replication
Our VMWare team are just dealing with the fun of upgrading right now.
And then there'll be the fun of "how do you co-ordinate a mass migration from Intel to AMD hardware which can't be done hot".
I have to say, I took some delight in dropping the giant box of DACs on their desk and about 9 metres of labels I'd printed for them 😄
I don't have a vmware team. I have me, one guy who knows the microsoft ecosystem stuff expertly and limited on everything else and a few consultants I can bring in for special occasions, lol. I wish I was big enough to have a vmware team.
I switched to cloud, so I don’t even get to touch fancy hardware anymore
I should have gone proxmox everywhere in my specific situation.
Miss me some blinkin lights
we have aws stuff too, that's an expensive bill
the aws stuff isn't my problem though
I work mostly with our network, and HPC workloads. All of mine is bare metal, which I much prefer.
I clicky the button in Dell OME and an hour or so later the servers have an OS on them, there my involvement largely ends.
I liked it, until I got to click deploy and play with my dog for a bit
we used to do that, virtualization just adds so much flexibility it kind of feels necessary, but it's not a high performance load, lol. It's stupid business applications like domain controllers.
I gotta go to bed though. it was nice talking to everyone.
Take care, great talking to you too
It can be done with HCX I believe, but I've never tried
Can you help me? I'm looking at a few TRVs/radiator thermostats to use with Home Assistant. I'd prefer them not to be too big, since one of them is right next to a door opening. I've looked at Tado, which looks great, but you have to pay 200 DKK a year for some features. I wonder if I could create a flow in Home Assistant that would do the same thing without paying. Otherwise, there's Aqara, which also looks great, but I'm not sure if I'm missing out on any functionality. It's also cheaper.
The #hardware-archived channel is the place to ask - and answer the pinned message there
Can I get one to annoy my neighbours?
Imagine the powergrid if someone just used the brightness slider
HAHA yes
it's Bobby from King of the Hill
gimme back mah purse!
The cooling on that must be crazy
I'm amazed at how bright and vibrant it is from the pictures, and that it isn't completely washed out by the sun
easy at night, but it also seems to look good during the day
I don't know you
Hier kommt dein Ohrwurm des Tages! Die sechs von der Müllabfuhr kümmern sich wirklich um jeden Dreck: Tonne für Tonne, Stück für Stück.
👍 Hinterlasst einen Like, wenn euch das Video gefallen hat!
▶️ Mehr Videos von der Maus: https://www.youtube.com/diemaus
🐭 Die Seite mit de...
🦝
this was last year, but yeah I took it offline. I put it back online in the meantime though so if they reuse the assignments from last year again then I might get another email soon
I would have completely ignored that e-mail, lol
I remember looking it up, this seems to happen quite a lot
One of our vendors at work requires us to take a yearly aptitude test to get a support discount, I've found the answers online before.
Also fuck that vendor, can't wait until we drop them
How?
They give us a discount for proving we're competent enough to not be filing stupid tickets
The discount is substantial, too. A few hundred thousand.
One of my listed job requirements is passing that stupid test.
How much you pay for support shouldn't depend on your mental capabilities
That's essentially discrimination
Feels very wrong
Maybe not in the consumer world but the business world is different
WIth the number of installs we have they'd need to hire additional employees to support us if we didn't know what we were doing.
What country is this?
Doesn't matter, they offer this globally.
No, it doesn't. There's no regulations in any country about the level of support provided between two private companies on multi-million dollar contracts.
This isn't a humanitarian or rights or discrimination issue
It's pretty simple, if we know what we're doing we create less burden on them so they charge us less.
You're definitely not in EU
You're definitely not in the business world
I'm not in the EU but we operate in the EU.
But, yea, this isn't a discrimination issue and I can guarantee you that there's no laws in the EU that regulate things like that, beyond normal contract law.
Yep knew it
I have no idea why you think that would be illegal or regulated. Calling it discrimination is absurd, too.
And if all you can say to explain it is 'well you must not live in the EU' I'm going to assume you're talking out of your ass.
I'm not saying the test itself is illegal, I'm saying that charging based on the result is at the minimum unethical and in the EU illegal
Stupidity isn't a protected class 😂
Unless you can provide a source for what you're saying you're only making yourself look naive.
Microsoft very much does charge based on classifications that depend on a number of certifications in a company
It's a standard step of review when entering into a contract with another company to review how much the deal is going to cost you and adjust the deal accordingly.
My issue with that one vendor is that they require yearly renewal
Gets old after a few years
The idiot tax is a real thing
Also the alternative is distributing costs resulting in us paying for some other company's support even though we have our stuff together and they may not. It's a support contract, not health insurance, it's not protected.
Any hourly billed support is ^^
Yeah, I've never heard of an aptitude test from a vendor
Should a mechanic only charge you for parts because it's not fair for them to charge for something they know how to do but you don't?
Unless you are an MSP, and want to retain certification level
We have two vendors who give us a discount because we're knowledgeable about their products, but only one requires a yearly aptitude test.
Haven't seen aptitude test but discounts when #of people has some cerification
Yes
It's not a basic aptitude test like matching up symbols, it's more like describing a hundred or so situations of problems with their software and asking us what the solution is to make sure we're competent with it.
Yeah, no, never seen that
I've only ever seen it from this one vendor
But even if it's not common I can't imagine any reason it would be legally disallowed
It does sound like something EU trade regulation would frown upon

Annoyingly I can't find anything public on the program, all I have are internal docs we got during our last contract renewal.
I would, I just can't find anything about it even existing here
It's not exactly common over here either
Generally they care more about certs
Or generated ticket volume
Certs usually only come into play with partner level, like premium, gold etc
never seen it matter in other contexts
We're partners with this particular vendor
and then again mostly xSPs
To the point we've had a few employees move back and forth between us and the vendor over the years
My old manager used to party with their CEO
Certs I understand
I think my old manager is also barred from going to conventions and public events on the company dollar because of that time she walked around drunkenly offering blowjobs too
She got moved to another department with a lot fewer people working in it, lol.
Certs aren't anything different than an aptitude test 🤷
Ofcourse they are. One tests knowledge, the other tests conginitive abilities
Knowledge without the ability to apply it is useless
Nobody cares if you know something, they care if you can do something.
Also cognitive abilities are a valid consideration in a contract, they're not a protected class
Then what's the value of a cert?
They do more than just prove you know a thing
I've never taken a cert test that was only 'Where is this located?' 'What does this do?'
Mine always had some questions like, 'given this situation, what is happening?' and such that required thinking
Even then if you knew everything in the world and could do nothing with you still wouldn't be able to keep a job anywhere.
I have 5 or 6 active certs, but like 4 of those are ones I got forever ago that don't expire.
I only have 2 non expiring ones
I got a few MS certs on server administration and ldap administration and an A+ when I was in highschool that don't expire
The trend seems to have changed to certs expiring every few years now though
I know the A+ changed, I was like three weeks off of the cutoff for it expiring when I took my test. My highschool paid for me to take it or I probably wouldn't have it at all.
I remember that test being kind of stupid because it had a lot of outdated questions on things like DOS printer interfaces and serial terminal support, in 2009.
All the plus exams are utter bullshit imho
Greetings people
i have exactly 1 cert, some cpanel cert that keeps my employer in "partner" status
Would you guys trust the raid capabilities of such external USB enclosures?
