Actions: dedicated meeting with TAPE providers

Rucio 1.25.4 comes with support for SRM+GridFTP together SRM+HTTP protocol

• only one of these protocols can be configured on RSE
• FTS transfer protocol preference for SRM must be set to https;gsiftp;root
  • no FTS interface to use different SRM preference for individual transfers
  • SRM+GridFTP used only for storage that doesn't support SRM+HTTP at all
• this is sufficient to cover ATLAS use-cases - transfers tape <-> disk
  • motivation - Data Challenges with as little as possible GridFTP (RAL Castor system)
  • CMS plans with tape transfers(?)

New / additional tape bringonline test

• upload ~ 10TB dataset with 1GB files to each tape endpoint
• ask dCache/StoRM administrators to clean these files from disk buffer
  • unfortunately storage administrators can't easily remove individual files and cleanup of whole buffer would certainly affect production
  • use existing old production data(set) with high probability to be on the tape(?)
    • it would be necessary to use production Rucio instance
    • require similar config overwrites (patches for Rucio) used e.g. by ATLAS Functional Tests WebDAV(?)
    • we would have to be more careful, but anyway at some point we have to move SRM+HTTP to production
• add Rucio rule to trigger transfer of NEARLINE file
• don't reuse files, because after test transfer they'll be ONLINE
• once we run out of NEARLINE source files ask again for disk buffer cleanup
  • with current test infrastructure all files will be used ~ in 30 days
  • run less tests or ask for bigger space to reduce cleanup requests(?)
• what would be good test for transfers with SRM+HTTP TAPE destination(?)
  • is transfer to normal disk instead of disk buffer sufficient(?)
  • how to verify that file really reached tape storage(?)

Keep current Fuctional Tests TAPE(?)

• not very useful to test TAPE
• just SRM+HTTP transfer from tape disk buffer
• concern that files are not really deleted from tapes
  • test files will be physically stored on tapes for years
  • currently 200GB/day
• modify to SRM+HTTP tests from disks?
  • e.g. "read timeout" issue is visible also for disks