You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Never passes locally for me, will make it a few hundred expected lines in before missing one.
Ran into this problem IRL while debugging engine stuff w/ buildkit scheduler debug logs enabled: #7128
Only workaround was to just run a non-nested dev engine, which made the full logs available via docker logs.
I wouldn't personally consider it a bug that we ever drop logs; having some sort of limit in some part of the telemetry pipeline makes perfect sense, however:
I'm not actually sure that's what's happening here, as opposed to logs being missing due to some actual bug
Even if that is what's being hit, it seems like we hit a lot earlier than ideal. The repro above is only writing 2000 lines, which isn't particularly crazy even if they are being written rapidly.
The fact that docker logs handles it okay seems particularly telling. Obviously we have a lot more going on with our telemetry, but the fact that I can't get all the logs even just locally feels wrong.
In that PR I linked to above, I also sometimes saw errors like OpenTelemetry error: grpc received message larger than max (approximately that, based on memory), which does feel like a genuine bug.
However, I am not seeing that anywhere in the repro above, so I don't know what to make of it. Maybe red herring or maybe the error itself gets dropped sometimes in these cases?
Repro test:
Never passes locally for me, will make it a few hundred expected lines in before missing one.
Ran into this problem IRL while debugging engine stuff w/ buildkit scheduler debug logs enabled: #7128
Only workaround was to just run a non-nested dev engine, which made the full logs available via
docker logs
.I wouldn't personally consider it a bug that we ever drop logs; having some sort of limit in some part of the telemetry pipeline makes perfect sense, however:
docker logs
handles it okay seems particularly telling. Obviously we have a lot more going on with our telemetry, but the fact that I can't get all the logs even just locally feels wrong.OpenTelemetry error: grpc received message larger than max
(approximately that, based on memory), which does feel like a genuine bug.cc @vito
The text was updated successfully, but these errors were encountered: