Last modified 3 weeks ago Last modified on 07/04/17 15:03:38

dCache Tier I meeting July 4, 2017

[part of a series of meetings]

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


(see box on the other side)

Site reports


Everything running smoothly.

Running 2.16.38 -- starting to test v3.0 -- already deployed in middleware readiness instance.

Testing v3.0 -- in theory 3.2,

PIC plan was 2.16 to 3.1

Plan to upgrade to 3.1 directly and then to 3.2 ASAP.

Not clear when the upgrade will happen -- not in summer, but in September.

ssh problem

The ssh issue fixed, but that was the wrong version. Still investigating.


dCache is running just fine in the past few days.

checksum calculations

Checksum calculations

We could calculate the checksum with a larger buffer, but it would be good to establish that this improves performance. The fio results suggest this change would have little impact.

Xavier reported that GPFS block size is 16 MiB.

Tests were run on an ATLAS pool node.

Mover inspection =

Client IP address, PNFSID involved, door involved, ...

Shared memory makes showing information per-mover hard. It might be possible to add per-transfer-service memory usage.

Client request information also would be useful; e.g., requested buffer sizes.

Support tickets for discussion

[Items are added here automagically]


Same time, next week.