Last modified 5 years ago Last modified on 02/11/14 17:25:33

dCache Tier I meeting February 11, 2014

[part of a series of meetings]

Present, Patrick, Paul), IN2P3(), Sara(), Triumf(), BNL(), NDGF(), PIC(), KIT(Xavier), Fermi(), CERN()


(see box on the other side)

Site reports


Looking fine.

We have received Some complaints from VOs transfers failing due to timeout or poor throughput. Unclear what is the cause, but we have excluded dCache as the cause.

Have managed to start 2.8 by starting PnfsManager before nfs-server. This is to test IPv6 functionality.

RT 7993

No. of incoming requests > sum of all allow tape recalls on HSM-attached pools.

KIT has two HSM-attached pools, each allowed to do 2,500 active requests. CMS requesting more than 5,000 files at the same time triggers this problem.

Discuss this during the team meeting.



  1. Report from lcg-utils replicate files from CERN to


Progress with disk-tape separation; we do see real tape back-end -- shared with the production instance. Date: finalise migration -- February 24 how to stage files.

Problem files pinning -- Natalia will send the information

RT -- about chimera dump

Once a week dump for consistency.

Want to do

production, physics group: local and remote,

scratch space, results

Single number per interested directory.

Handful of directories.

Projects office; in addition to managers. Both Natalia and Tony Wildish are in charge of this activity.

AP/ Paul to trigger mail about time of meeting.

Two instances: tape & disk.



Support tickets for discussion

[Items are added here automagically]


Same time, next week.