Hola Arnau,
> we're debugging the low performance of our WMS and found that there's a
> process (/opt/glite/bin/glite-lb-bkserverd) which is doing nothing but
> consuming resources and increasing the iowait of the host.
>
> If https://twiki.cern.ch/twiki/bin/view/EGEE/GliteWMS is right, a pure
> WMS does not need that process. So we've killed it.
On a pure WMS that process should be harmless. I suspect the problem with
that process resulted from other things that went wrong earlier.
> there's another problem, /etc/cron.d/glite-wms-check-daemons.cron tries
> to restart it always. That cron does not check if the host is LB/WMS or
> LB, or WMS...
That will be fixed in the EMI 1 WMS. You could hack that script...
> And the last problem is the presence of a hanged restart process:
>
> [root@wms02 ~]# ps -ef|grep glite-lb-bkserv
> root 5842 3947 0 11:33 pts/0 00:00:00 grep glite-lb-bkserv
> root 16976 16965 0 Jan24 ? 00:00:00 /bin/sh /opt/glite/etc/init.d/glite-lb-bkserverd restart
>
> and after a new restart:
>
> date:
>
> # date
> Thu Mar 24 11:57:19 CET 2011
>
> ps -ef|grep glite-lb-bkserv
> root 11008 10999 0 11:45 ? 00:00:00 /bin/sh /opt/glite/etc/init.d/glite-lb-bkserverd restart
> root 11012 11008 0 11:45 ? 00:00:00 su - glite -c /opt/glite/bin/glite-lb-bkserverd ??--notif-il-sock=/tmp/glite-lb-notif.sock ??--notif-il-fprefix=/var/tmp/glite-lb-notif ?? -c /home/glite/.certs/hostcert.pem -k /home/glite/.certs/hostkey.pem -i /var/glite/glite-lb-bkserverd.pid --dump-prefix /var/glite/dump --purge-prefix /var/glite/purge ??
> glite 11013 11012 0 11:45 ? 00:00:00 /opt/glite/bin/glite-lb-bkserverd --notif-il-sock=/tmp/glite-lb-notif.sock --notif-il-fprefix=/var/tmp/glite-lb-notif -c /home/glite/.certs/hostcert.pem -k /home/glite/.certs/hostkey.pem -i /var/glite/glite-lb-bkserverd.pid --dump-prefix /var/glite/dump --purge-prefix /var/glite/purge
> root 11491 3947 0 11:47 pts/0 00:00:00 grep glite-lb-bkserv
>
>
> So, could someone confirm that such process is not needed?
Does it still consume CPU or I/O resources?
|