Getting LHCb into Production
- eos-xrootd-4.12.7 and eos-4.8.39 fix the XRoot "write to files opened read-only" bug and also the "missing check of targetsizeerror" bug fixed by Steve.
- Julien has deployed these versions on eosctalhcbpps (15/02/2021).
- Michael followed up with Chris Haen after the meeting about final tests and schedule.
- Chris does not have time right now to do the functional test of DAQ-to-CTA with check file safely on tape check. This is not a blocker for migration, we will do these tests in production.
- On the other hand, Chris does want to test CTA to/from all the other sites before migration: "I am currently deploying HTTPS in production in all our sites to be able to talk with CTA but to have the full coverage, we will need the token configuration, to accomodate for sites using Storm."
- Michael, Chris and Julien will meet on Monday 22/02/2021 @ 11.00 to finalise details of the HTTP tests.
- w/c 22 February: HTTP tests. Prepare OTGs.
- 1 March: disable write access to CASTOR LHCb in preparation for migration. Check all queues are flushed and everything is written to tape.
- 8 March: disable CASTOR LHCb. Migrate to CTA.
- 15 March: EOSCTA LHCb in production
- HTTP+TPC tests
- Publish OTGs
Getting PUBLIC into Production
- AMS tests are in progress. xrdcp transfers work, but when transferring with FTS there is an error, "Operation not supported: Destination does not support delegation." Vova is following up.
- NA61/Shine: Meeting on Thu 18/02/2021 @ 14.00 with EOS+FTS+CTA teams to discuss "best practice" for their new DAQ storage endpoint.
- DUNE: no news, we are not pushing as we have enough to do and there is no rush (data taking starts in 2022).
- n_TOF, COMPASS: need to test with spinner space. Julien will configure this this week so Vova can proceed with testing.
- Repack of public_user is continuing with tapes containing LEP data in preparation for migrating the legacy tapepool.
- Draft KB article for users to access files in CTA is in CodiMD.
- Repack of NA62 tapes is complete and has been checked that all NA62 fileclasses are now in the correct tapepool.
- Vova and Michael met with Barbara (17/02/2021). Barbara confirms that their code will clean up files from EOS PUBLIC for their recall workflow (CTA → EOS PUBLIC → process batch in HT CONDOR → delete files from EOS PUBLIC). Therefore spinner space is not required for the NA62 case. (It is still required for n_TOF and COMPASS).
- This means that NA62 tests are complete and migration to CTA can proceed on the original schedule.
- Mon 22 Feb: remove migration routes for NA62
- Check all archive queues are flushed to tape
- Disable NA62 directories in CASTOR
- Migrate NA62 tapepools to CTA
- Mon 1 March: NA62 in production
There are minutes attached to this event.