Unexplained EMFILE

[Colin Watson]

txpkgupload on ppa.launchpad.net hit EMFILE at 16:02 UTC today. Nagios only noticed at 18:13, and the service self-recovered at 18:20. There doesn't seem to have been unreasonable upload activity at that point, so it seems likely that there's some path that's leaking file descriptors, or perhaps a lack of timeouts somewhere. Of course it's good that we self-recovered, but it would be better to understand why we got into trouble in the first place.

[Colin Watson (cjwatson)]

https://pastebin.canonical.com/p/YgZmWRNwzJ/ (internal link) suggests some badly broken timeout logic - there are connections there that have been open for weeks, as far as I can tell from logs.