Self-hosted bot crashes when analyzing recording (fetch to stt failed) #42
Comments
|
Hi, thanks for raising this! Do you see log entries like the following from the STT container after speaking? Do you see a spike in CPU usage after speaking (for example through htop)? What are your hardware specs? Is it virtualized or bare metal? Redis not finding a config is to be expected, but I am surprised by the connection, given that the bot container is instructed to wait for Redis to start first. Can you try pinging Redis by running the following command |
|
Hello again and thanks for a quick response! :D I'll start off with your initial question, and no, no such entries can be seen in the log. Regarding specs; it's a KVM with 1 core and initially 2 GB of memory, but increased to 8 GB see if it was a resource issue, but unfortunately with no change in outcome. The host hardware is a Ryzen 7 3900x with 128 GB 3200 MHz DDR4 ECC. Memory is well managed and nowhere near capacity. Storage consists of 8x1 TB 7200 RPM 64MB Cache drives using raidz2 with lz4 compression enabled. The guest storage is currently qcow2, so all in all I would like to think that it is close to optimal in its configuration when aiming for performance and based on what it is. :D I've reviewed the host performance in the period leading up to and including the crash which I just provoked. There's no iodelay of note and as it's my home server, there's not much load due to me and my machines being its only users. During the test i just performed, the hosts CPU utilization peaked at 4%. Here's a screenshot of the KVMs performance graphs where the tiny spike at the end is when I invited the bot into a voice channel and started talking: The spike and over all increase in resource utilization in the screenshot above is due to me installing and then starting glances on the KVM to see if it would produce any warnings, which unfortunately it did not as seen here after the crash: I can't see anything else in journalctl on the host nor guest that strikes me as relevant or would fall into the error category, but will provide output from the same time period if requested. :) docker container ls after crash: Unfortunately, I forgot to ping redis before provoking the crash, but it replied after so I would think it would have before as well. ghcr.io/wotschofsky/discord-live-translator log: libretranslate/libretranslate:v1.3.11 log: No other information outside of this line is produced when restarting the container, and no errors related to the crash present. redis:alpine log: ghcr.io/wotschofsky/discord-live-translator/stt log: Hope this is of use. :) |
|
Can you please try resolving the internal ip for the stt container like follows:
|
|
Hey! Pardon the late response. Been busy days. Edit of shame: Brain started ticking over and here's the output as requested. :D |
|
Hi, thanks for your patience! Can you try updating the images and see if that fixes it? |
|
Hello again, and no problem. Take the time you need. :) Unfortunately, no dice. Could this be some missing dependency thing? Again, my ignorance may shine through, but the Debian install is just base + ssh server, then I've installed as per procedure in README.md, or at least to the best of my recollection. #weeksof4hournights :D Maybe it would be of use if you could provide me with a complete list of packages installed in your containers where you have it running successfully? At least then we are closer to being sure that we're not chasing ghosts of my making here. Again, thanks for your ongoing support. Our Spanish team mate will absolutely love this thing if we can get it to start talking. :D |
|
Hello again! Sorry, I really don't want to nag, but would it be possible to get some more information about the OS and versions of packages you're able to get this running on? I'm not married to running any configuration, I just really, really want to have this bot working as it would make our teams current issues regarding communication much less of a problem. Thanks again! |
|
Hi, sure: Here are the two configurations that work for me:
In both cases running the latest versions of Docker. I hope this helps! I've been considering moving this app to the Bun runtime. Maybe that would solve the issue or at least give some better errors. |
|
Alrighty, put that plan on hold and I'll try to replicate your environment hopefully by the end of the weekend. Enjoy yours! :) |
wotschofsky
changed the title
|
Ok, so we've lost the spanish guy, so our need for the bot seems to have evaporated. If I find time (as I've surprisingly not been able to to date), I'll keep my promise to see if this is related to Debian 12, but yeah. May take a while. This would be my suggestion if anyone else is struggling, and if you find that it does work - please relay. :) |
Hello there!
I'm attempting to get your creation up and running on a Debian 12.1 install as per instructions in README.md. The install is up to date and the box has plenty of free storage and memory as the problem I'm about to delve into occur. vm.overcommit_memory = 1 has also been set and verified as part of my blind stab at resolving this issue.
I've successfully managed to get the bot onto our server, commands are being recognized and all seems well up to this point. The bot acknowledges our desired language, joins the voice channel when requested and plays its little fanfare. :)
However, a few errors are thrown during startup.
ghcr.io/wotschofsky/discord-live-translator log:
Otherwise the logs look alright besides of redis lusting for a config.
Redis log:
Libretranslate log:
ghcr.io/wotschofsky/discord-live-translator/stt log:
The problems begin as someone starts speaking. One can see in the logs that voice is being recorded, but the following error pops up in the ghcr.io/wotschofsky/discord-live-translator log:
Any thoughts on what might be going on and how to resolve it? I'm a complete newbie to docker, so I figured before I poke at things further; I'd consult the oracles. :D
Best regards,
yours truely.
The text was updated successfully, but these errors were encountered: