Hi Andreas,
Andreas Haupt wrote:
>>> [root@lcg-ce1 ~]# ps auxww|grep gma
>>> patlas36 5717 0.0 1.9 107792 39444 ? R 16:02 0:00 globus-gma: polling job https://lcg-ce1.ifh.de:20021/23015/1238514168/
>>> root 5722 0.0 0.0 51120 668 pts/0 S+ 16:02 0:00 grep gma
>>> root 15279 0.9 1.9 106720 40008 ? Ss Mar31 15:49 globus-gma: polling jobs
>>> patlas01 25746 0.0 0.0 0 0 ? Z 04:02 0:00 [globus-gma] <defunct>
>>>
>>> This is ok, isn't it? Well:
>> It's not really. Gatekeeper needs to be restarted so that GLOBUS_GMA
>> environment variable gets propagated properly.
>
> Thanks again, I just thought if globus-gma is running everything was
> fine already. You proved that I was wrong ;-)
Right. globus-gma is just an "information provider" for WMS and you need
to activate new code inside jobmanagers to use this information,
otherwise you'll have both old and new code running in parallel which is
a waste of resources.
>>> [root@lcg-ce1 ~]# uptime
>>> 16:05:16 up 231 days, 5:52, 1 user, load average: 3.83, 4.09, 4.67
>
> so the load is better now but not really normal...:
>
> [root@lcg-ce1 ~]# uptime
> 11:18:34 up 232 days, 1:05, 1 user, load average: 1.47, 1.44, 1.49
>
> It used to be below 1 under normal circumstances before.
I'd say that 1.5 is quite a moderate load for a 4-core box. If you think
it's too much you can play with parameters described here:
https://twiki.cern.ch/twiki/bin/view/EGEE/LcgCE
--
Cheers,
Andrey Kiryanov.
|