Last modified 11 years ago Last modified on 07/20/10 17:08:46

dCache Tier I meeting July 20, 2010

[part of a series of meetings]

Present,Tanja,Paul), NDGF(Gerd), PIC(Gerard)


(see box on the other side)

Site reports


Gerd reported that NDGF suffered some hardware problems last week with the database servers. These problems were fixed by replacing some faulty memory and NDGF is OK again, now.

Last week, NDGF tried to upgrade to 1.9.9-1. They ran into a few problems, so are planning to upgrade again this Thursday.


PIC upgraded to 1.9.5-21 a few hours ago and have just opened the FTS queues. So far, everything is running fine.

Gerard also reported on the plan for deploying a recent version of dCache as a separate instance at PIC. This would involve deploying 1.9.9-1 on virtual hardware as a development/testing instance of dCache.

Gerard asked if there were any issues with running dCache on virtual hardware? Paul mentioned that people have observed anomalous performance issues with some older versions of PostGreSQL (the database was unexpectedly slow); but, beyond this, dCache should work without problems on virtual platform.

The 1.9.9 dCache instance would be for disk-only projects: some Tier-3 resources and maybe the ATLAS Tier-2 storage. The new hardware would arrive by September 1st; but, Gerard would like to have something up and running before then, using the virtual hardware.

The plan is to start testing 1.9.9 in the next few weeks. Once this is running OK, then Gerard will start testing the migration process. For the most part, the migration will simply involve copying across data, populating the namespace in the process; however, later (Autumn or later on) there may be a need to migrate a portion of the namespace from PNFS to Chimera.

Gerard asked if it's possible to do this? Paul: yes .. just create an empty directory in Chimera and use pnfsDump to migrate the desired subtree of PNFS into that directory.

Support tickets for discussion

[Items are added here automagically]


Same time, next week.