SAM tests:
- On Thursday all CMS sites went into drain due to nobody noticing the HC jobs cease running a few days previously. This was tracked to a config file that CERN had given additional security protections. Fixed within the day.
Production:
- A drop in running cores at the moment, below pledge, but assume this is simply fair share balancing (CMS had higher running cores earlier in the week).
- CMS-Rucio also suffered from the oversized CRL issue seen at RAL last week.
AAA machinery:
- Both recycled machines SVC42 and 41 now in prod; SAM tests are green.
- Other AAA gateways have been turned off for now.
- New XRootD load-testing tool used to test performance.
- Regular crashes observed (42) which didn't impact production it seemed - Jyothish fixed the config to accommodate the core dump size and we observe no crashes since (except one, but no core dump visible).
AAA tests, with the load-test tool pushing on top of normal production activity:
- Memory: Climbed a bit, not as bad as old machines, but worth keeping an eye on
- Throughput: High, almost filling the NIC at one point (25Gb NIC).
- Affect on Echo: None observed (Jyothish can comment)
/unmerged:
- Deletion still on-going, parallelisation may not have scaled well, so we may have to ensure this is running for the next 1-2 months.
DC27 progress: Katy put the numbers for CMS desired rates into her all-VO spreadsheet, ready for other VOs to start adding info too.
FTS4: ATLAS are testing in Production, agreed with Steve Murray to have CMS test also in ~3 weeks assuming no show-stoppers