Unclear error messages and logging #38

Closed
flah00 opened this Issue Sep 25, 2013 · 2 comments

Comments

Projects
None yet
3 participants
@flah00

flah00 commented Sep 25, 2013

https://gist.github.com/flah00/6701958

Running postgres 9.1.9 and wal-e 0.7dev on ubuntu 12.04 lts

I came across that error on a secondary, after doing a pg_dump/pg_restore on my primary, deleting my wal-e logs, generating a new backbackup, and restoring the basebackup to the secondary.

To test the validity of the error, I restarted the secondary. It processed all of the segments again. Every time I restarted, it would start from the first wal log, and fail on the last segment. The segments were increasing, so it wasn't always failing on the same ID, just the last one.

The logs remained unchanged on the secondary for 15 minutes, which was odd, because the primary was receiving changes. I also noticed that the pg_xlog directory on the secondary had new data written to it, even though the log file did not mention the arrival of the new segments.

Eventually, after about 20 minutes I saw logs updating on the secondary... The secondary is keeping up, but the errors I initially found lead me to believe there was a problem.

Are these errors wait-and-see or actionable?

@fdr

This comment has been minimized.

Show comment Hide comment
@fdr

fdr Sep 27, 2013

Member

They're the result of polling for a new segment. Yeah, they are exorbitantly ugly compared to their severity. Care to patch it, mostly to elide input?

In particular, the "HINT" text is right, or could be expanded to include "polling for WAL segments"

Member

fdr commented Sep 27, 2013

They're the result of polling for a new segment. Yeah, they are exorbitantly ugly compared to their severity. Care to patch it, mostly to elide input?

In particular, the "HINT" text is right, or could be expanded to include "polling for WAL segments"

@bshi

This comment has been minimized.

Show comment Hide comment
@bshi

bshi Oct 11, 2013

Member

It looks like an older version at one point avoided the scary stack trace

https://github.com/wal-e/wal-e/blob/3f683075/wal_e/worker/s3_worker.py#L366

Member

bshi commented Oct 11, 2013

It looks like an older version at one point avoided the scary stack trace

https://github.com/wal-e/wal-e/blob/3f683075/wal_e/worker/s3_worker.py#L366

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment