[ClusterLabs] Log rotation for /var/log/debug-pcmk.log

[Ken Gaillot]

On 01/12/2017 05:33 AM, Jan Pokorný wrote:
> On 11/01/17 18:46 +0000, Andrew Nagy wrote:
>> I hope this is the right place....
>> We have pacemaker running (on at least one) RHEL 7.2 server with a
>> huge (8+ GB) /var/log/debug-pcmk.log file. No one else here knows
>> anything about pacemaker and the person who built/configured the
>> server is unavailable for the next two weeks. There is *no* enough
>> free space on /var/log (bad, I know). How do I fix this with minimal
>> impact to the users or the application (NFS server)? If a service
>> outage is unavoidable, I can work with that too.
> 
> I suppose this, run as root, would work:
> 
>   cp /var/log/debug-pcmk.log <BACKUP-LOCATION>  # optional
>   truncate -s 0 /var/log/debug-pcmk.log
> 
> You have to live with a fact that few lines queued to be logged in
> the interim (there may be overlaps runtime-wise) may be dropped.

The location of the log is most likely specified by PCMK_logfile in
/etc/sysconfig/pacemaker.

If it's really set to debug-level verbosity, that's unnecessary during
normal use. It may have been set that way during an issue for
troubleshooting.

Check in the same file that none of the PCMK_debug, PCMK_trace_*, or
PCMK_blackbox variables are set. Any change in that file will take
effect the next time the cluster is restarted on that node.