[part of a series of meetings]

Participants

Agenda

[see box to the side]

Status of work for 1.9.5

A (quick?) review of activity needed for the 1.9.5 release

New info service/info provider

Active/passive fixes for SRM client

TODO: remove item: done

PnfsManager based listing in SRM

TODO: remove item: done

Terracotta and SRM

TODO: remove item: done

new http door with https support

TODO: remove item: done

xrootd mover reimplementation

XXX

HSM cleaner for chimera

Will not go in.

Fix ReplicaManager to use new pool msgs

New message might arrive more frequently that old message.

There is a patch in review.

There was some discussion on whether to catch null pointer or fail-fast.

Alex described three problems with current patch from Tigran.

ACL rename checking

Gerd: no changes so far. Try to find some time

Tigran found trunk doesn't work

getRequestTokens failing in a similar way since ...

Tigran: no, not related.

Never written to database and never written into HashMap?.

Tigran: can send srm-devel config. files.

Can also send the stack trace where it fails.

Rolling back to 25 July and still seeing the problem.

Timur would find login into srm-devel useful. Tigran to make sure this happens.

Timur can work on this today.

Error msgs from SRM w/o Space Manager

TODO: remove this item: done

ExtendFileLifeTimeInSpace without access latency specified (What to do)

This has always returned NOT_SUPPORTED.

second observation: removed [...]

Usually extract AL/RP from name-space.

Possible to make space AL/RP.

Can we fail it? No, the spec. requires it to work.

Space reservation.

Spec doesn't bind AL/RP to the space. This is

Should be able to create a space without AL/RP.

AL is nillable, so need this as an external parameters in SpaceManager. RP must be specified when creating a space, so no need for a defaultRP parameter.

What if nearline is set by mistake. Timur: that's OK.

Default AccessLatency and RetentionPolicy in dCache for srmReserveSpace.

Done.

large volume

Large volume of the dcap failures due to the checksum verification on the pool: we would like a close timeout in client be configurable.

Timur:

Cataline reported that more than 15% of transfers are failing for all US CMS. This seems to be due to a 5 minutes for success of close operation.

Switching off the on-write checksum verification is not acceptable for Jon.

Make the timeout configurable on dcap client.

Timur to ask Tigran off-line.

a) whether it can be done, b) when it can be done.

Timur: we need to do this by the end of the week.

Fermi to propose a change and it be incorporated after the 1.9.5 release.

kerberos tunnel in dcap

Inclusion of the kerberos tunnel in the standard dcap rpm

Patch to add a new target to

Completely different kerberos implementation: Heimdale and link statically.

Makes sense to update the Heimdale distribution in SVN (0.8) current version is 1.1.

Switch to Heimdale is OK with Fermi.

Only DESY and Fermi are using Kerberos.

CDF experiment in many countries.

Commit patches and treat in the normal fashion.

Issues from [FIXME: Add link to yesterday's Tier-1 meeting]

Outstanding RT Tickets

[This is an auto-generated item]

​RT 5055: thousands of mounted pnfs .(access) entries

​RT 5070: Disk file in tape restore queue! Bug in 1.9.3-3?

​RT 5072: dcap movers hanging

Review of RB requests

Status of PNFS packaging and release for 64 bit

DTNM

Proposed: same time, next week.