#1213 new defect

Incorrect child count in worker threads for GraphOld.pm and HTMLOld.pm

Reported by: volo Owned by: nobody
Priority: normal Milestone: Munin 2.0.0
Component: master Version: 1.4.6
Severity: normal Keywords:
Cc:

Description

I'm passing this bug report and patch along on behalf of the person who identified and fixed it. If you need more details, let me know. This patch fixes a logical error that resulted in the worker thread having a mistaken impression about how many subchildren it had forked (correct answer: zero, previous answer: however many the parent had forked).

I have not tested it against 2.0-rc5 (1.4.6 is our version), but the defect still appears to be there, and the patch should apply with an offset.

Attachments (2)

GraphOld.pm.patch (257 bytes) - added by volo at 2012-04-30T14:50:07+02:00.
HTMLOld.pm.patch (255 bytes) - added by volo at 2012-04-30T14:50:34+02:00.

Download all attachments as: .zip

Change History (5)

Changed at 2012-04-30T14:50:07+02:00 by volo

Changed at 2012-04-30T14:50:34+02:00 by volo

comment:1 Changed at 2012-06-06T10:58:18+02:00 by akheron

I'm running munin 1.4.6 on Ubuntu 12.04, and the cronjob outputs errors like this:

Subject: Cron <munin@dev> if [ -x /usr/bin/munin-cron ]; then /usr/bin/munin-cron; fi

[ERROR] Ran out of children: No child processes
 at /usr/share/perl5/Munin/Master/HTMLOld.pm line 868

Would these patches fix this?

comment:2 Changed at 2013-02-04T09:24:05+01:00 by jenkins

this patch should not be used please remove this...

It makes munin-html fork until server dies due to out of memory.
I have a munin server with 100+ clients...

yes this makes the child count work but it breaks master count.
munin-html seams to spawn one master for each configured client.

removing this patch then munin ends up counting the masters right but the child per host can be lots.

Last edited at 2013-02-04T09:25:36+01:00 by jenkins (previous) (diff)
Note: See TracTickets for help on using tickets.