lots of /tmp/apport_core_* files laying around

[Neale Pickett]

On many production retracers, there are a lot of old (>24h) /tmp/apport_core_* files gobbling up space. On the machine I'm currently looking at, there are 45 such files, as old as 5 days, using around 22GB.

I don't know what logs you guys are going to want but please ask for anything that'd help and I'll pull it for you :)
Meanwhile, here's a run against all the production retracers.

[Brian Murray (brian-murray)]

I was pinged about this again today:

08:44 < jacekn> bdmurray: hello. prod e-t-retracer-app/20 has over 21G of files like apport_core_y7zomh in /tmp. Oldest one is from Sep 20th and newest are less than 5h old. Can some of them be wiped? Do you know why they are not cleared?

[Brian Murray (brian-murray)]

There is a cronjob on the retracers called "cleanup-tmp-crash-files.cron" that could be modified to clean these up until the reason for them staying around is sorted out.

[Brian Murray (brian-murray)]

I suspected this was related to the retracers being restarted by the "retracers-cache-restart" cron job, and apport exiting prematurely and not cleaning up the apport_core files in /tmp/. It is apport-retrace that creates these specific files and not the retracer code. Looking at retracer log files for today we don't see many instances of the retracers being restarted, so I pinged pjdc regarding the quantity of apport_core files in /tmp.

He said:

13:24 < pjdc> bdmurray: checked 10.42.32.18. only one file matching
              the pattern

So this particular situation seems resolved, although we should still guard against it.

[Brian Murray (brian-murray)]

revno: 91
committer: Brian Murray <email address hidden>
branch nick: daisy-retracer
timestamp: Mon 2015-09-28 14:39:52 -0700
message:
  templates/cleanup-tmp-crash-files.cron.tmpl: cleanup crash / core files as root