ScotGrid Technical Meeting

Wednesday 11 Jun 2014, 10:55 → 12:00 Europe/London

Other Institutes

Gareth Roy (University of Glasgow)

ScotGrid Technical Meeting 11 June 2014

 

-

Ewan

Gang Q

David C

Gareth R

Sam S

Andy W

-

Durham

-

 

No tickets until this morning.

Ewan: Ticket this morning re CE submission error - nagios tests are

now green, though. (Seems to have been transient high load.)

 

Ollie has been away for the last couple of weeks, reasonably quiet.

Still getting to grips with SLURM (Ollie is the expert). Playing with

fairshare, to get it to work as we want.

 

-

Edinburgh

-

2 Tickets:

 

Andy: The longstanding glExec ticket (in the Ops meeting, there was

some confusion as to if ATLAS were going to require it or not). The

glExec can't progress without the working glExec tar ball.

Glue publishing ticket - On CREAM CE there's a glite-info-dynamic

script (dynamic queue info). The problem is that we have to use a

non-root ssh to get the info. Need to work out how to get it

functioning directly.

It did work pre-EMI3, so our theory is that it must be a simple config change.

 

Andy is visiting Glasgow to discuss monitoring solutions for ECDF.

Also working on elastic search with logstash for our log monitoring

(in combination with University of Edinburgh core and with data

analytics studies within ATLAS).

 

Operationally, things have been fine. ATLAS SW&Computing week, gave

update on the project to use ARCHER/HPC resources for ATLAS. Might be

moving towards testing in production.

 

Gareth noted the existence of the HEPSYSMAN talk on centralised

logging. Dave C: two strands - one general discussion about

centralised logging from a security perspective, but also a

presentation about using elastic search for logging of some of their

systems at RAL. (CERN are also looking at this - in the context of

central logging and (monitoring and consolidation) .)

 

-

Glasgow

-

 

Tickets:

 

Dave C: predominantly, we have had no major issues. The one ongoing

issue is the post-brownout issue with disk074's raid sets. (We had to

do some work to get it back, and we lost one of the two raid sets -

this is being resolved (we're draining the other raid set before we

actually do anything to recover the system). )

Ticket will be updated with the list of lost files once drain is

complete (currently chasing up the last few straggling files in odd

states.)

 

Now caught up with publishing, which was straightforward using the new

EMI3 APEL. Because we had the parsed data for the whole year, we just

republished the gap for the missing months. (Now publishing fine from

the CREAM CEs.)

 

(Andy noted that they've been having issues with publishing at ECDF.

Getting a nagios warning for a tiny discrepancy, but now making more

work by trying to republish and sync the APEL2 and APEL3 publishing to

get it perfect.)

 

In Glasgow's case, because we had all the records for all the levels

(CE, batch), we could always reconstruct records for republishing.

 

-

Others

 

-

 

Gang: Condor test pool on the test cluster - ARC-CE is now submitting

to them and passing test. ARC-CE publishing of accounts should now be

working. Planning on testing WNs of all 4 classes on prod cluster.

 

Sam: DPm-drain issues.

 

 

-

AOB

-

 

- Publishing

 

Gareth double checking our info publishing.

 

ATLAS records - double count ECDF and multi count Durham (but okay in Gstat?)

 

Ewan noted that the Gstat numbers are also out for Durham (possibly

twice the real value). The problem seems to be that ARC-CEs might be

being double-counted (each CE publishes full cluster).

 

 

- 4trly Reports.

 

We need narrative parts for the report, talking about the interesting

things that we've done.

(Andy on ARCHER, Durham leading the UK on SLURM, etc)

 

 

- The Pheno user using 700TB of storage on Glasgow shared FS.

(We think he is no longer in Pheno or doing anything, so would like to

clean up the data).

Ewan will check the user's existence.

 

 

(Ewan) - Face to Face?

 

Gareth will ask Dave Britton re this and timing.

 

 

(Dave C, Andy W): Will both be on holiday next week.

 

-

Chat Log

-

 

 

Ewan: (11/06/2014 11:14)

I was considering that for us

if you get anywhere can I see what you do

errrrm....

just a little bit :/

11:00 → 11:15 Tickets

11:15 → 11:25 Research Updates

11:25 → 11:35 AOB