XMPP Service Operators - 2021-03-31


  1. jonas’

    the prober update I did last night broke everything.

  2. Ge0rG

    jonas’: is this question still relevant?

  3. jonas’

    Ge0rG, well… not right now ;D

  4. jonas’

    but if you could check around thet timestamp of my question it’d be great

  5. Ge0rG

    Is there some string to easily grep for in prosody.log?

  6. jonas’

    I don’t recall unfortunately

  7. jonas’

    we can also postpone this until I fixed the prober again :)

  8. Ge0rG

    jonas’: the log file has 37M lines. "Around that timestamp" is 1.3M lines. I'd happily grep for whatever you want, but I'm not going to review it manually.

  9. jonas’

    hahaha

  10. jonas’

    yeah, I’ll try to find something

  11. Ge0rG

    👍

  12. Martin will also enable debug logging

  13. tom

    » <Ge0rG> jonas’: the log file has 37M lines. "Around that timestamp" is 1.3M lines. I'd happily grep for whatever you want, but I'm not going to review it manually. Is that amount of logging useful?

  14. tom

    Speaking of, is it really neccecary for prosody to log by default every time it makes or recieves a successful federated connection?

  15. tom

    Instead of just the errors

  16. tom

    Or significant events

  17. Ge0rG

    tom: often you only know after the fact which event was significant. Most of the logging is junk, but I don't have a mechanism to filter relevant from unrelevant

  18. tom

    Idk

  19. Ge0rG

    exactly.

  20. Ge0rG

    just running debug logs and letting the disks rotate is much cheaper than doing significant work to improve the logging framework

  21. tom

    I've never had an event where the message like: Mar 31 01:29:30 s2sin557956e4ae70 info Incoming s2s connection yax.im->conference.nuegia.net complete was actually useful in troubleshooting something

  22. jonas’

    08:25:52 tom> Speaking of, is it really neccecary for prosody to log by default every time it makes or recieves a successful federated connection? Looking at my logs, I was also wondering about that. `info` on the other hand seems reasonable for that.

  23. tom

    I don't think so. Perhaps messages like these would be better suited for the 'debug' syslog severity

  24. tom

    » Mar 31 01:29:24 s2sin557957e68af0 info Incoming s2s stream yax.im->conference.nuegia.net closed: connection-timeout » these on the other paw ^, useful to know

  25. jonas’

    but then you only see when the connection was closed, but not when it was opened

  26. jonas’

    but then you’d only see when the connection was closed, but not when it was opened

  27. Ge0rG

    I don't care if info or debug, as I'm debug-logging anyway

  28. tom

    I wonder if prosody would accept a patch to change that

  29. jonas’

    you could ask :)

  30. tom

    Maybe some other time

  31. tom

    For now I think i'm just going to enable a debug rolling buffer and only lag warns and errors to disk

  32. tom

    have a yeen https://cdn.nuegia.net/cc0a3999-fff5-43f1-adbf-879cf43d9e9b/db3abbd947288b4c.jpg

  33. Ge0rG

    tom: there is also mod_log_json

  34. tom

    I don't see how that's useful

  35. Ge0rG

    tom: it's structured logging and it can be dumped into kibana or somesuch