Last modified 5 years ago Last modified on 03/21/17 15:54:09

dCache Tier I meeting March 21, 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


Things are going OK at PIC.

NFS issues

One issue with NFS service, with one use-community.

Tigran said the file-handle changed in the last version, so some clients are trying to use the old file handle.

Remounting the client.

Intensively using NFS; after a few hours, they start seeing IOErrors. WN that is specific for this community, with no errors in the kernel log.

The WN are Scientific Linux 6.

PIC is currently running dCache v2.16.29.

The NFS door doesn't show any problems.

Can also check for dCache returning NFS4_SERVERFAULT error, as this is (likely) due to clients trying to reuse the old/existing file-handle from before the upgrade.

Problem with multiple stage requests

Could be the fail-over within pool-manager?

Looks like the is triggered by the enstore script returning an error.

Marc to investigate further: check whether the script failing is evident in all duplicated staging, and try to reproduce the problem.

SRM flushes

Perhaps use the delayed flush-to-tape to support delaying flushing until after srmPutDone is called.

CRC on restore

Is needed CRC-checksum on restore?

Enable checksum on restore in pool, update script to generate checksum file.


dCache is working fine.

SQL script

Still on Tigran's white-board.

Getting the pool name

pool's name or rather cell name.

staging pool

Question: how to handle several HSM interfaces for writing.

hsm create <type> <instance>

Different URIs: osm://instance-1/... osm://instance-2/

istate enable/disable, stored by priority.

Solution: have HPSS-dedicated pool and force client to stage through this pool.

Zookeeper prefix

The dcache.zookeeper.connection property is also used to specify a zookeeper name prefix.

Xavier reported this is working for him.

Schema upgrade problems

Testing the upgrade from 2.13 to 2.16.

pins_v3 table.

Marc opened a ticket about a very similar-sounding problem: RT 9153

Xavier to open a fresh ticket.

logback connection

dCache tries to connect to the local logback server

default log-back configuration, it tries to connect.

xrootd problem

Issue from a while ago: a French certificate -- could not be validated by dCache.

Should upgrade to 2.13.51 (or newer) to fix the problem.

Support tickets for discussion

[Items are added here automagically]


Same time, next week.