Difference between revisions of "Tier1 Operations Report 2016-10-26"

Revision as of 09:13, 26 October 2016

RAL Tier1 Operations Report for 26th October 2016

• There have been some problems over the last fortnight with load on AtlasScratch in Castor. (This did not affect other areas in Castor). The number of Atlas Pilot jobs has been reduced, and then increased, at various times to limit the load on Castor. The weighting of gridftp jobs on atlasScratchDisk within Castor was changed on 10th Oct to increase the job capacity.
• There have also been load problems on the Castor GEN instance. These have been caused by a large number of accesses from jobs running at other sites by a particular ILC user. This has affected other VOs access to Castor (for example ALICE). We are in contact with the user to tackle this.
• There was a problem with the squids on Friday (14th Oct). The problem was in the Atlas Frontier system. However, as the squids are multiple-use this had a knock-on effect on CVMFS clients on the batch worker nodes.

• GDSS648 (LHCbUser - D1T0) failed on Saturday evening (15th Oct). It is showing both disk and network problems. It was finally returned to service on Tuesday 25th Oct.
• GDSS699 (LHCbDst - D1T0) failed on Saturday (22nd Oct). It was returned to service later that day read-only. On Monday (23rd) it was taken down for further investigation - being put back in service again the following day.

• There is a problem seen by LHCb of a low but persistent rate of failure when copying the results of batch jobs to Castor. There is also a further problem that sometimes occurs when these (failed) writes are attempted to storage at other sites.
• The intermittent, low-level, load-related packet loss that has been seen over external connections is still being tracked. The replacement of the UKLight router appears to have reduced this - but we are allowing more time to pass before drawing any conclusions.

• GDSS896 (CMSTape - D0T1) was taken out of service yesterday (25th Oct) to investigate memory errors.

• On Wednesday 5th October the UKLight Router was replaced. This has also enabled a a 40 GBit connection to the RAL border routers (i.e. giving a datapath of up to 40Gbit for the bypass connection.) We have seen a significant increase in traffic over the bypass route since this change - well exceeding the previous 10Gbit limit.
• Some re-balancing of the disk servers in lhcbDst has been carried out following a problem where some became full.
• Some changes were made to increase the number of CMS batch jobs that we run in order to bring the number more into line with the pledge. This was both an increase in the number of worker nodes draining at any time (from 10 to 20) to allow more multi-core jobs (CMS' jobs are purely multicore) as well as an increase in the limit in Condor.

Listing by category:

• Castor:
  • Update to Castor version 2.1.15. Planning to roll out January 2017. (Proposed dates: 10th Jan: Nameserver; 17th Jan: First stager (LHCb); 24th Jan: Stager (Atlas); 26th Jan: Stager (GEN); 31st Jan: Final stager (CMS)).
  • Update SRMs to new version, including updating to SL6. This will be done after the Castor 2.1.15 update.
  • Migration of LHCb data from T10KC to T10KD tapes. The additional 'D' tape drives ave been delivered. We await installation into the tape library by the vendor.
• Fabric
  • Firmware updates on older disk servers.

Key: Atlas HC = Atlas HammerCloud (Queue ANALY_RAL_SL6, Template 729); CMS HC = CMS HammerCloud