> [Again, Glasgow, for example, passes Ops storage tests partly because the
> ops tests don't talk to the same filesystems that ATLAS use. The ops tests,
> therefore, do not actually test
> the functionality of the overwhelmingly most important filesystems for the
> availability of Glasgow as a functional site.]
But that's your choice, isn't it ? At Imperial ops can basically write
almost anywhere.
I don't think this should be micromanaged. If you are confident that
your storage systems are appropriately monitored and you decide to
sent the ops tests around a dummy loop, so be it. So I get around the
queuing times, by having a special queue, but the rules are known and
I comply with them. Ops test run. Obviously I myself know best on
where the tests are lacking - and I don't think there is a one size
fits all solution - and we have a local site monitoring page with a
lot more detail to plug the gaps, but I don't think they should be
discarded as a whole, because a lot of the times, they do detect real
problems.
Cheers,
Daniela
|