Cream01/Cream02 sync problems. Argus02 lost the NFS mount point to the shared gridmapdir, and hence was mapping users to different uids than Argus01. On the other hand, Cream02, that depends on Argus02, was having the correct mapping, so, with new maps (6 CMS users that we know of) it was returning a "[Missing property local.user.id]" error.
GPFS is under testing. Performance is promising, but we need to reduce it order to have good reliablility.
New Nagios server being created from scratch, also under construction.
Waiting for AMD Magnycour upgrade.
PSI (reports Fabio):
No news about middleware upgrades or installations, configuration stable; Chimera upgrade pending.
We're playing a lot with the SGE accounting DB to analyze usage trends and tune the scheduler; I suggest the DB installation in a SGE site. The SQL Schema is the same so we can share queries. Btw, regrettably SGE CAN'T make a scheduler simulation based on a past workflow+different policy.
I've collected several storage offers about our next HW upgrade and we're now in the decision phase.
We suffer even 5 broken disks in 1 week in our Solaris storage cluster but thanks to SMARTd I've detected each failure, I suggest its installation on Linux and Solaris.
UNIBE (reports Gianfranco):
Stable operations
Upgrade to ARC 11.05 required by ATLAS by end of October (according to the experts, it can be done on the live system!)
Will ARC 11.05 solve the ARC CE in site-BDII publishing issue?