Open issues:
------------
- problems with Lustre and AFS
- lustre: 8.3.2012 -> ?
- lustre: 13.3.2012 -> hardware failure
- lustre: 29.3.2012 -> hardware failure?
noticed at 10am, ticket to NAF, no reponse
tickets: no anwers: 1 ticket to help-desk
quick answer within 15 minutes: 1 ticket (afternoon)
it took the whole day to sovle the problem, this is not
acceptable for a large scale, Germany wide analysis facility
answering tickets and informing user was not well handled,
neither for the naf-helpdesk@desy.de nor for naf@desy.de
- AFS: 21.3.2012 -> overloaded AFS server: 2 volumes moved, rebootet (?)
user identified/user noticed slow down
repsonse from helpdesk after 20 hours
- AFS: 22.3.2012 -> overloaded AFS server: mor voumes moved (?)
same user -> optimised/changed workflow (less jobs,
input/output from scratch)
we can not prevent this, but should be able to put the finger
quickly on the problem -> AFS server monitoring for VO admins
please inform ATLAS support so that we can see what has been
done and investigate/improve
- AFS: 3.4.2012 -> overloaded volumes
o g.atlas.sw.1711: has two RO replicase, how can this happen?
Waiting for answer from naf@desy.de
o g.atlas.sw.1720: missing RO replicase, added
o reponse on ticket: >1d (not acceptable)
ATLAS Support: Period from March 1st to April 3rd (Wolfgang Ehrenfeld):
-----------------------------------------------------------------------
- accounts: ~10
o problems with checking /atlas/de group -> does not look like it
was solved
o accounts for Terascale Introductory school -> not again in this
way ATLAS members can have an ATLAS NAF account, but not non
ATLAS members from ATLAS institutes
o accounts for small ATLAS MC tutorial after Terascale MC school
(no problems)
- ganga from CVMFS (32 bit okay, 64 bit needs special config)
- qstat problem (5.4.2012): high load on SGE server?
CVMFS feedback:
---------------
- ATLAS setup script (asetup) updated and option cvmfs added
- tested with compilation and running on the batch with SGE
(compilation was reduced by factor 3 or higher, use of multiple
cores can now reduce overall time even further)
- side effect: more ATLAS software available, e.g. ganga, dq2, ...
maybe use these also as default
- propose to inform all ATLAS users that cvmfs is optional and should
be tested, after one month (and no problems) make it default
- we would like to request enabling the nighlies repository:
atlas-nightlies.cern.ch
Resource requests:
------------------
- we are tight on local disk space in Hamburg, what are the plans for
SONAS? We have started data taking again.
- we are tigth on dCache space at Zeuthen (LOCALGROUPDISK) and are
preparing a request
