RAL T1: Andrew, Steve, David
RAL Storage: Owen, Jiri, Jens
0. Welcome to Sabah who's now joining the list. Next meeting
is 9 March.
1. Review of Actions
Can people outside get write access to storage web site Jens Open
-- Andrew says PPD runs a web server serving pages from AFS.
-- For now, send stuff to me (Jens) and I will happily upload it,
or to Owen and he will patch his documentation.
-- Need for CVS in the longer term.
Can Edinburgh provide bugzilla? Phil Open
-- Part of the same question. Not likely to happen soon.
Can we use the helpdesk? Jens Closed
-- Short answer is no. It will cost a lot of money in extra
licenses (for people). It *is* possible to add storage to
the helpdesk though.
Send external URLs to Jens ALL Open
Update project page to point to new storage web site Jens Closed
Figure out why dCache doesn't use DCAP Jiri Closed
-- Done. Documentation to appear on web.
Send report on data transfers to list Steve Open
-- Still open. Some work may need to be redone.
Resurrect SRM client API Jens Open
-- There is an API but it requires fairly intimate knowledge of
gSOAP and the SRM protocol. Writing a good simple API not
likely to happen soon with other workload, particularly a
C one. C++ or Perl might be easier.
Plan for Tier 2 deployments Jens Closed
-- There is a plan, agreed with Jeremy and the T2 coordinators.
Hasn't been implemented yet though, waiting for
all-other-Tier-2s-than-Edinburgh to catch up. Alessandra
is saying she expects to start this week.
Send DPM endpoint to Jens for interop testing Andrew Open
-- This is Andrew Smith. He has been busy with LHCb tools,
nothing much has happened on the DPM front.
Update docs to include lcg-* Jens Open
-- Still open. More general client documentation to appear.
2. Tier 2 rollout status.
See above. Waiting for everyone-but-Edinburgh to catch up...
3. dCache experiences.
Much captured by useful thread on list (quod videt). Andrew (Sansum)
reports dCache is now operational again, seeing 60 MB/s to disk.
Database moved off head node, now seeing 10-20% load on head and
about 40% on database. Speculating that database is spending lots
of time on inefficient queries [/me or needs vacuuming]
Need to agree with CMS whether they will start to write to tape
soon. Or we can restart the tape tests.
4. DPM status.
No news; see above.
5. Client tools.
See above - we have an API but it's not easy to use. Writing
experimental CLI advisoryDelete. Although we do not officially
support client side tools, we have a lot of experience with them
and need to make this available on the web.
Who's going to GGF13? Apparently no-one. We expect it's Arie's
turn (of the Arie/Peter duumvirate) to go. Owen points out that
there are other storage-relevant areas in GGF [meaning GFS-WG].
More importantly, is there an SRM workshop planned? None known,
but best to check with Arie.
Can GFAL be preloaded Jiri Open
-- From quick tests, GFAL does not appear to pass local files
through to kernel open. GFAL does not provide gfal_fopen
either. Thus it might be necessary to preload another
library that maps open to gfal_open. Test plan agreed.
Contact James or J-P about DPM source Jens Open
-- Phil points out we should get the source from the source,
so to speak.
Contact Arie about next SRM workshop Jens Open
-- I'll volunteer to do that.