RAL Tier1 Experiments Liaison Meeting
Access Grid
RAL R89
-
-
13:44
Experiment Operational Issues
-
1
Site Operations
-
2
ATLAS Operations ReportSpeakers: Brij Kishor Jashal (Rutherford Appleton Laboratory), Jyoti Prakash Biswal (Rutherford Appleton Laboratory)
-
3
CMS Operations ReportSpeaker: Katy Ellis (Science and Technology Facilities Council STFC (GB))
Katy was on leave last week. The network incident last Tuesday had little effect on CMS compared to other VOs, as CMS did not drop all jobs, and in fact picked up most of the farm as other VOs drained.
Some SAM test failures for AAA since the network incident. Affects the older machines (gw10/11) but not the newer machine (svc20). On Monday SAM tests were green but gw10/11 did not appear to be serving data. Katy did a restart of services on one machine (gw10). Both machines started serving data (so likely the restarts had no effect). Since last night the service has become more busy (50-100MB/s served, but many thousands of connections). gw10/11 are struggling again with spikes up to 10k connections, bad read amplification and intermittently failing SAM tests. Jyothish put both machines in a sandbox this morning.
Request for the 2026 tape pledge to be provided early has been granted. I have not yet added the additional capacity as I am working with CMS DM to add some throttling to what Rucio can submit to FTS. This is to protect RAL from too big bursts so that we can continue and complete the repacking to the new hardware tech.
Transfers look good, we hit 13GB/s to Echo from CERN (1hr binning) last week on the 8th.
Job performance generally good, however the slow job read times has not gone away. Looking more carefully, this is hugely dominated by Merge jobs even though the amount of data read by merge jobs has remained consistent and average below 10GB, see images attached. Are merge jobs reading remotely? Seems to be affecting only RAL of the CMS T1s, and having a very negative effect on job efficiency (of course we do not run many of these jobs, so minor effect on overall efficiency).
-
4
LHCb Operations ReportSpeaker: Alexander Rogovskiy (Rutherford Appleton Laboratory)
Quiet week. Only a few minor issues:
- Lack of MC productions is still present in LHCb
- Job pressure is unstable as a result, and drains happen frequently
- There were two a small spikes (~200 transfers) of download failures from ECHO to RAL WNs last week
- Due to xrootd container upgrade on the local gateways
- Lack of MC productions is still present in LHCb
-
5
ALICE Operations ReportSpeaker: Alexander Rogovskiy (Rutherford Appleton Laboratory)
-
6
LSST Operations ReportSpeakers: Mathew Sims, Timothy Noble (Science and Technology Facilities Council STFC (GB))
-
14:30
Tier-1 Projects
-
7
Anatares Upgrade
New EOS nodes
Repack ProgressSpeakers: George Patargias, Thomas Byrne -
8
XRootD DevelopmentSpeakers: Alexander Rogovskiy (Rutherford Appleton Laboratory), Jyothish Thomas (STFC)
-
9
Utilizing GPUsSpeakers: Jyoti Prakash Biswal (Rutherford Appleton Laboratory), Thomas Birkett
-
10
SSD Storage Evaluation
-
11
Echo deployment
-
15:30
AOB
-
12
Summary of Operational Status and IssuesSpeakers: Brian Davies (Science and Technology Facilities Council STFC (GB)), Darren Moore
-
13
Any other BusinessSpeakers: Brian Davies (Science and Technology Facilities Council STFC (GB)), Darren Moore
-
13:44