Hi Jan,
> I'm looking for some brave soul that could explain to my limited
> sysadmin mind what could possibly be the reason for the cronjob
> specified in $SUBJECT on a WMS node. It is apparently designed to
> attempt to start some WMS daemons, regardless of whether the "gLite"
> service is supposed to be running or not. This means that
> `/etc/init.d/gLite stop` on a WMS node is not enough to stop the service
> and perform its maintenance, as some daemons might "suddenly" start again.
Indeed, the stop command should set a flag that is checked by the cron job.
Please open a Savannah bug for that.
> I didn't realize that one has to read through various wiki pages to find
> a document about how to stop a WMS. Am I the only one who finds such an
> approach to a problem "let's make sure it doesn't die" a bit weird?
Various WMS services have suffered occasional crashes that can be cured
to a large extent by having such a cron job. Otherwise the admin would
have to be involved every time, and users would suffer much more from
those crashes. Sure, bugs should be fixed, but in the meantime the
cron job provides a handy workaround.
|