Server becomes unresponsive after a while during ffmpeg AttachmentExtractor task

[magguzu]

Just a quick update, I've started keeping a spreadsheet of the instances of this to try to narrow it down. I've made some small DNS changes but no no avail. Right now the only thing that the problematic occurrences seems to have in common are:

* Jellyfin-web or web wrapper app
* Remuxing
* ASS subtitles

I have a few other network changes in mind to try (namely running container in Host mode) but it being just web makes me unconvinced this is the issue.

I wish I could separate jellyfin-web from server to get their logs separately somehow, that might help to narrow it down.
Edit: on second thought the whole server is affected when this happens, so it slow even if I move to Android TV (not sure if I mentioned this) so it definitely seems to be server side. But only initiated when watching from web... Bizarre.

[magguzu]

So this issue seems to have stopped.

Some changes I've made, not sure which, or if a combination of them might have fixed things:

* Changed jellyfin to run in host mode rather than bridged
* Because of the above change, changed my reverse proxy to target http://containerhost.localtld:8096 rather than internal DNS with http://jellyfin:8096
* I removed an entry jellyfin.localtld from internal DNS. This was targeting the host IP (same as containerhost.localtld). Because I had localtld as one of my DHCP's lookup domains, I realized this might have been causing some resolution conflict between http://jellyfin:8096 (container IP) vs http://jellyfin.localtld:8096 (host IP). In the end this shouldn't have mattered since I was running in host mode anyway, but just covering all bases...

None of this really explains why this only happened with one specific kind of show and only over web. I thought the web portion could be explained by how it targets the backend server (my thought was some DNS conflict related to my third bullet) but people on Discord made it seem like it doesn't work this way.

Without understanding the architecture super well, could it be a DNS issue when the attachments that AttachmentExtractor extracted were trying to be retrieved?

For now things seem stable. Fingers crossed!

[TheDreadPirate]

For all weird issues, blame DNS.