Red Hat Bugzilla – Bug 1312576
gluster nagios addons package installed a sysstat crontab that conflicts with existing sysstat crontab
Last modified: 2017-03-25 12:26:55 EDT
The provided gluster-sysstat.crontab file ends up with conflicting metrics collection provided by the /etc/cron.d/sysstat crontab.
gluster-nagios-addons-0.1.16-1.el6rhs.x86_64 : Gluster node management add-ons for Nagios
Repo : installed
Other : Provides-match: /etc/cron.d/gluster-sysstat.crontab
It seems this file should be shipped as an example, and should probably not be run by default.
Too often I see emails from root like the following:
From: firstname.lastname@example.org (Cron Daemon)
Date: Sun, 21 Feb 2016 00:00:01 +0000 (UTC)
Subject: Cron <root@foo> /usr/lib64/sa/sa1 1 1
flock: Resource temporarily unavailable
Tim, can u take a look?
We have this file to enable nagios monitoring for gluster. We can't make it as a sample since our monitoring solution requires data collection in every minute. We have see similar issues in the past as well where this command fails exactly at 00:00:01 +0000 (UTC). Here also the same case, it failed at "Date: Sun, 21 Feb 2016 00:00:01 +0000 (UTC)".
Darshan can u investigate further?
(In reply to Ramesh N from comment #3)
> We have this file to enable nagios monitoring for gluster. We can't make it
> as a sample since our monitoring solution requires data collection in every
> minute. We have see similar issues in the past as well where this command
> fails exactly at 00:00:01 +0000 (UTC). Here also the same case, it failed at
> "Date: Sun, 21 Feb 2016 00:00:01 +0000 (UTC)".
> Darshan can u investigate further?
As you mentioned, we had the problem of an entry not being made into sysstat's saDD file during day's transition (at 00:00:00 hrs). We fixed it by adding another entry in same cron job for making an entry to the file at last minute of the day. with this that issue was solved.
But the issue here seems to be resource un availability. the cron job /etc/cron.d/sysstat tries to collect and update the sysstat data every 10 minutes once and our cron job does the same thing every minute once. and while updating the sysstat data file(var/log/sa/saDD) a lock on the file is held by entity updating it. If multiple jobs try to update simultaneously only one will succeed.
We should check if this issue arises if we have only one cron job running to update sysstat data.
Can you check if this exists with gluster-systat crontab alone?
Is there a supported work-around for this issue?