Page Comparison

XRD-6

Lockless reads

Thomas, Jyothish (STFC,RAL,SC)

Met Rob C. on Monday; VR jobs failing still, generally with [socket timeout]. Seems specific within the job submission. Rob C. also looking into the round robin with local tests. Not much change reported by Rob in job success rate.

XRD-11

Slow deletions

Ian Johnson

1. It would appear that the FTS results are consistent with what we see and also what ATLAS sees in production. LHCb’s results appear to be significantly longer and we don’t understand why. We need to continue investigating this in th eGGUS ticket.

2. We should talk to the FTS team about what they believe are sufficient deletion rates.

3. Ian is trying to repeat the FTS tests, he can then look at the spread of results (as the CERN tests didn’t have error bars) as well as seeing if the parallel deletes were hitting the same gateway.

N/A

Corrupted LHCb file(s)

Thomas, Jyothish (STFC,RAL,SC)

Caused by write conflict between davs and gridftp. A variety of problematic outcomes observed from replication test. Xrootd/davs takes a couple mins to clean up after reporting failure, during which unexpected behaviours can still arise.

• Running over the full list of files in each pool is outstanding