Version 1 (modified by paul, 6 years ago) (diff)


dCache Tier I meeting November 11, 2014

[part of a series of meetings]

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


(see box on the other side)

Site reports


Filter by storage class when doing rep ls.

Things are going well, most of the time. Some issues with ATLAS that we could not a

dCacheDomain 90% CPU usage, out of memory.

Problem didn't reoccur. Heap dump wasn't generated due to directory

NFS v4 support issues.

Increasing max movers solves the problem, but still don't understand.

For now their benchmark tools are completing fine.

They are using ROOT, right? ROOT opens many files concurrently then you reach the max movers for that pool. If TChain has multiple files then it opens all of them before starting to process them.

Inside dCache we don't clean-up for transfers where the client goes away fix with release today.

Testing migration to 2.10. Read migration guide.

How do I find out if a pool was banned? A hint is that all traffic


Pool can now execute a health script. What's the purpose.

Probably want to upgrade to 2.11 immediately -- mostly because of the HSM bulk.


LHCb SRM connection got cancelled.

Some rules that kill idle connections after 3 hours.


Make sure we log client or server terminating the connection

"Connection closed"


Contact ATLAS plugin Ilya.

Support tickets for discussion

[Items are added here automagically]


Proposed: same time, next week.