Tech Meeting Minutes 20110217
Present: Graeme, Mike, Wahid, Andy, David, Sam, Mark, Peter, Stuart
|Table of contents|
- https://gus.fzk.de/ws/ticket_info.php?ticket=66929 - Glasgow, Pheno issues. Seems to be ok now. svr023 is draining to be upgraded with latest proxy renewal packages
- https://gus.fzk.de/ws/ticket_info.php?ticket=66918 - Durahm, ATLAS, in progress. Will be done. Request is not super-urgent, but it is from a major user and does not take that long.
Durham Squid Access
- Access to Glasgow's squid is fine - was already in the ACL list.
- Testing network to/from RAL: outbound link is fine; Inbound link still throttling at ~1Gb/s. Colin Cooper discovered a module under high load and is working on a bypass for that. This would be a temporary fix until the summer (maybe moving onto a noon-resilient link?). svr003 VM was disabled during the test - this has no effect.
- Sam happy to lead a network commissioning effort at Glasgow. This might involve lower level tests (like iperf) with friendly sites. Mark thinks out network path may go down the east coast, increasing the number of hops.
- Advice was for Durham to install CREAM on SL5 and not to upgrade old lcg-CEs.
- To reboot CEs completely safely, jobs in the batch system should be paused (David is pass documentation to Peter
- Running 800 analysis jobs seems to work fine, go up to 1000 now.
- ACTION Raise ATLAS pilot job limit to 1000.
- Wahid still having trouble with ATLAS HC tests and random failures.
- Next meeting: Feb 24, http://indico.cern.ch/conferenceDisplay.py?confId=129019.