Katy is at the Rucio workshop.

Last week there were some token/ARC-CE tests timing out during the final test in the suite. There was no log file so harder to follow up. The problem went away in a couple of days and hasn't reappeared. 

More job failures than normal, caused by User Analysis. I wrote to one individual who was pulling large (sometimes more than 100GB) inputs per job which were frequently failing, but did not hear back. For now the level of failures has dropped to normal low levels. Overall CPU efficiency remains good. 

Tape writes have been busy, with 2.25PB arriving on Antares in the last 30 days - a lot of data coming from CERN but also the usual rate coming from Echo. This is higher than agreed total rate given the repacking needs. Do we need to pull back a bit or is this fine? Please let me know if we need to reduce. N.B. Failure rate is negligible and everything appears to be working perfectly from CMS side.