supervisorctl unable to shutdown all processes

[Maarten Siersema]

Hi,

I've noticed that when I perform a "supervisorctl shutdown" event not
all spawned processes from supervisor shutdown correctly. When running
supervisord in debug mode through "supervisord -e debug" I see the
following messages in my log:

2011-09-20 06:59:41,245 INFO stopped: appcfg_service (terminated by
SIGTERM)
2011-09-20 06:59:41,245 DEBG received SIGCLD indicating a child quit

Both nginx and appcfg_service won't close down. All the other
processes (mongod, memcached, the tae app, etc) quit as expected. I'm
currently running supervisor under root.. Must supervisor run under a
specific user? Or are there any other reasons why the processes won't
quit?

regards,
Maarten

[Tobias]

Hi Maarten,

Interesting. Haven't seen this behavior before. You could try to tweak the [program:nginx] and [program:appcfg_service] section in the supervisor config file. To be honest, I haven't tried this, but regarding http://supervisord.org/configuration.html#program-x-section-settings you could try another stopsignal parameter. Please, let me know if it makes any difference.

Cheers,
Tobias

[Maarten Siersema]

Hi Tobias,

I've added s3fs support for the blobstore, this introduced the problem
due to timing issues for starting the supervisor processes. I've now
solved the issue by reverting back to the original configuration
files. I added the s3fs support by setting up rsync to a s3fs mounted
filesystem. This works for now..

regards,
Maarten