Frederic Schaer wrote:
> Er... ? Weird.
> Then the globus_config function might not have been called when the node
> was configured (but then, why is it working ? ;) )
FYI, even if the logrotate entry is present, it may _still_ be ignored:
there is a bug in logrotate that makes it bomb out as soon as there is
a duplicate entry, i.e. two entries dealing with the same logfile.
But in that case root normally gets an error mail...
> I'll ask the one who installed the node what he did when he's back
> (since we don't use yaim for everything...)
>
> Thanks !
>
> P.S : FYI, looking at how to disable the gatekeeper syslog logging (all
> the GRAM stuff found in the gatekeeper *and* messages logs), I finally
> found that [globus|edg]-gatekeeper (at least the version in use) uses
> the [hardcoded] daemon.notice facility (and sometimes "daemon.err" to
> say "ping sucessfull" - funny)... but I don't know if something else
> uses this daemon.* facility ...
>
> That's for those who dislike seeing 30 times same log entries in 30
> different files ;)
>
> F - CleanSweep - S
>
> Maarten Litmaath wrote:
>
>> Frederic Schaer wrote:
>>
>>> Hello Mr Rollout,
>>>
>>> I'm wondering if there's something that's advised for setting up a
>>> logrotate for globus-gatekeeper ?
>>
>>
>>
>> Er? With YAIM you get this:
>>
>> --------------------------------------------------------------------
>> # cat /etc/logrotate.d/globus-gatekeeper
>> /var/log/globus-gatekeeper.log {
>> nocompress
>> copy
>> rotate 1
>> prerotate
>> killall -s USR1 -e /opt/edg/sbin/edg-gatekeeper
>> endscript
>> postrotate
>> find /var/log/globus-gatekeeper.log.20????????????.*[0-9] -mtime +7
>> -exec gzip {} \;
>> endscript
>> }
>> --------------------------------------------------------------------
>>
>>> I noticed on one of our CEs that the logfile was becomming very large
>>> (> 500 MB), and I've seen that there has been some kind of logrotate
>>> in previous LCG/EDG releases for the gatekeeper - but nothing's there
>>> anymore.
>>>
>>> So I played with globus, and it seems that the following is working
>>> OK because logrotate runs the postrotate script before compressing.
>>> Anyway, it only *seems* it's working - I've found nowhere
>>> documentation about globus-gatekeepere and logrotate. Also, the
>>> following creates empty globus "self logrotate" files...
>>>
>>> [root@cclcgceli05 logrotate.d]$ cat /etc/logrotate.d/edg-gatekeeper
>>> /var/log/globus-gatekeeper.log {
>>> compress
>>> rotate 31
>>> missingok
>>> ifempty
>>> postrotate
>>> /usr/bin/killall -s USR1 edg-gatekeeper
>>> endscript
>>> }
>>>
>>> It seems it's also possible to use the logrotate "copytruncate"
>>> option (without the killall postrotate), but this could cause data
>>> loss, which is not really good :(
>>>
>>> Any advice ?
>>> Thanks,
>>>
>>> Frederic Schaer
>>
>>
>>
>>
|