ScotGrid Technical Meeting
Other Institutes
-
Andy Washbrook
David Crooks
Ewan Steele
Gang Qin
Gareth Roy (chair)
Oliver Smith
Sam Skipsey (minutes)
-
- Durham
EMI3 ticket.
Ewan: just the EMI2-3 upgrade. Discovered good things in Ops - Chris
Brew pointed out that APEL service is not needed if you have an
ARC-CE.
The SE needs a physical upgrade, but not a package-upgrade (just the
EMI-release updated).
CEs “progressing”.
Ganglia + Nagios installed, logging stats. Can sort out access when finished.
Gareth noted that at one point “unified monitoring” was mentioned.
Oliver: not progressed with WMS issue, as it seems to have gone wrong
elsewhere in the ARCs. Getting close to production ready (the GridFTP
seems to be flakey without logging).
Gareth noted that the issue may be the ARGUS-auth bug that caused it
to timeout and fail some connections (the workaround was to remove the
caching in ARGUS).
- ECDF
EMI3 ticket.
Andy: Mostly focussing on the EMI ticket at the moment. Deployed new
CE, new ARGUS server, using them as a pivot to upgrade existing CE.
Ready to go live in the near future (pending new hostcert for ARGUS
server).
Anticipate no problems.
Was caught out by the APEL upgrade (EMI3 release is completely
different architecture to EMI2).
Some issues with migration of pool accounts/integration with ECDF
systems team implementation in NIS.
Monitoring is “next step”. Nagios/Ganglia, “advanced monitoring stuff
from Glasgow”.
Some PANDA auto exclusions but only blips.
- Glasgow
Dave C:
LHCb pub. ticket is ongoing (pending dependant tickets on actually how
to sensibly publish values in Glue2 schema.)
EMI3 ticket about to be closed, as last CE was moved to EMI3 as of
last night. Last step is to reintegrate the APEL publishing and turn
it on (tested from CE -> APEL, but not APEL->RAL). Once enabled, just
need to gap publish for April to sort out messy records.
Monitoring; have some ideas for integration, poss via iframes as
lightweight solution.
Need to remain flexible.
Gang: Set up an ARCCE, published in SiteBDII. Connected and tested
with testq v prod PBS (need to test with real grid jobs). Also have a
test Condor pool, which the ARC has been tested against.
Sam: (DPM Collaboration/Dev meeting).
—
Cloudsoft - is anything happening? (No news from them to anyone.)
Scotgrid NeoWebsite.
Dave C has been reworking the ScotGrid website. There is a
beta-release (demoed in the meeting).
Request for comments on content (sites comment on their own
descriptions). Also contribute any local user communities you want
included.
Comments in the next few days pre: presentation of Beta to Dave B.
(Next step is the Documentation, which is accessible internally. This
is being generated using Sphynx, managed via Git.
Andy asked about avoiding duplication of effort, as he’s also working
on a similar, ECDF local thing.
Gareth noted that our model was to use Git to manage distributed
effort on the Documentation (rather than a Wiki, which is very heavy
for this kind of thing).
(ECDF+Durham using Confluence presently, so migration tool needed?)
)
GridPP-Monitoring blog:
Dave C has started the above blog, as a news/update site about what’s
happening in monitoring. If anyone is interested in contributing…
Decommisioning CREAM-CE
Ewan: who do I talk to to get ATLAS to migrate from CEs to new CEs?
Gareth: email ATLAS_UK_CLOUD support (but in general, also mentioning
the question in Ops would not be useless for other sites in any case).
-
David Crooks: (30/04/2014 11:04)
No, Sam's just getting sorted out
Andrew John Washbrook: (11:06 AM)
opz team rulez
Gareth Roy: (11:06 AM)
eord up
word up
:(
http://argus-authz.github.io/doc/EMI-Argus-SysAdminGuide-1.1.0.pdf
Known issues section
not sure if it's still a problem
Andrew John Washbrook: (11:16 AM)
I'll pass that on to Wahid!
ooh logo
Gareth Roy: (11:20 AM)
Dave did a lot of work on it... cool eh?1
Andrew John Washbrook: (11:20 AM)
indeed
Ewan: (11:22 AM)
is this accessible by us?
yeah we're using confluence and documenting durham there
Samuel Cadellin Skipsey: (11:34 AM)
atlas-support-cloud-uk@cern.ch
Probably.
David Crooks: (11:35 AM)
Cloud support: atlas-support-cloud-uk@cern.ch