On 19/10/11 14:31, Andrew Washbrook wrote: > Hi, > > Does anyone know if there is any official recommendation for running cvmfs_fsck on WNs? > I keep hearing it being mentioned but I cannot find any guidelines for appropriate use > - especially for sites now migrating to CVMFS. > > In particular I would like to find out: > - What command line arguments (if any) are used by sites running cvmfs_fsck > - How frequently this should be run At the last GDB, it seemed to be recommended to run it every 24 hours. AIUI, that's what the the RAL Tier-1 do. > - What impact this process has on worker node load (and how long the operation takes) > - What to do if cvmfs_fsck throws errors: is the local cache and subsequently the WN unusable? cvmfs_fsck -p cleans up the cache IIRC. > > If you have any advice it would be much appreciated! My understanding is that the desire to run cvmfs_fsck on a regular basis actually comes from problems with corruption that have now been fixed. The Tier-1 hit them and ran cvmfs_fsck on a regular basis. Manchester also ran into the problem and filed a ticket. Once I'd fixed QMUL's StoRM problems I noticed lots of failed analysis jobs on a subset of nodes - sending the dodgy files to Jakob resulted in a fixed cvmfs a couple of weeks later. I've not seen problems since (Adding regular cvmfs_fsck is still on my todo list). Chris