On 07/10/2015 07:16 AM, Dave Page wrote:
> Are you able to get a stacktrace from a running instance that's hung?
Yes, but it's not doing anything, it's just polling for the next job.
> Also, does setting the logging level to debug reveal anything useful
> in the filesystem logs? (pgagent -l2 ....)
Nope. And I scanned the postgresql log for SQL errors, and there's nothing.
What appears to be happening is:
1. pgagent picks up job
2. pgagent finds first step in that job
3. pgagent executes first step (successfully, according to the
postgresql log, and according to a fake "touch file" job I added).
4. pgagent abandons the job without updating the status of either the
job or the step.
5. pgagent starts polling for new jobs.
6. Repeat 1-5 until all jobs are in "r" state with an assigned agent,
but not running.
Note that this particular user has now abandoned pgagent and is
switching to centrally managed cron, so my ability to get further
troubleshooting information will be limited.
According to apt-get, this is the latest pgagent package.
--
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com