Presents: Rob, George, Miguel, Giuseppe
RAL:
- CASTOR versions status:
- SRMs on 2.1.16-10.
- 2.1.16-13 on all LHC instances + Gen.
- 2.1.15-20 for the facilities instance, using xrootd and RFIO. Users are aware that RFIO is deprecated.
- 2.1.16-0 on the tape servers.
- SRM issue with double-slashes: worked around for now with a PL/SQL procedure resurrected from the old DB. We have now a ticket in Jira at CERN and as a similar issue (without crashes) was seen with FTS, it will be further investigated.
- Gridftp errors seen end of June: if it happens again, we'll look into the details.
- Q about LTO support: CASTOR will support LTO drives once they will be considered for purchase at CERN instead of the Oracle drives.
- Still observing memory leaks on the transfermanager daemons. Confirmed at CERN. To be further investigated, RAL is available to give access if this helps the investigations.
CERN:
- LHC Technical Stop last week: all instances upgraded to 2.1.16-19. Hopefully last patch release of the 2.1.16 series, unless urgent/critical needs.
- Upgrade to CentOS 7 campaign: well advanced after having sorted out a number of 'exotic' issues. Over 50% of the CASTOR diskservers run on CentOS 7. Next is one headnode per instance.
- The docker container for Ceph is ready but not yet deployed.
- Tape Recommended Access Order (RAO) is going to be implemented in CASTOR. It will be available on 2.1.17-x / CentOS 7. Target (TBC) is the next LHC Technical Stop in mid September.
- Q about recent issues with xrootd transfers from tape servers: not seen any longer, following the latest xrootd releases, which do include a number of specific fixes. We run now xrootd 4.6.1-1 everywhere.
AOB:
Next phone conference: August 22nd.