Project

General

Profile

May 24th

Date:
05/24/2017

Attendees:
Anna M, Brian Y., Felipe A., Marc M., Margherita V.W, Robert I., Steve W., Vladimir P.

Agenda:
  1. v2_1_1 status (All)
  2. Web interface improvements (Steve)
  3. Python 3 migration (Marc & Vladimir)
  4. AOB

Discussion:

1. POMS v2_1_1 Status:

STEVE:
AboutNew features, new look from Dashboard.
It will also show only the experiment(s) the user belongs to.

ANNA:
We still need a preliminary page as an overview,introduction.
Also show campaign dependencies.
Example MCC9 for Proto-Dune. There are 3 stages, DETSIM,RECO AND MERGEANA, the latest two depend on the previous and are automatically submitted.
The campaign dependencies layout depends on the configuration.

STEVE:
Also add submenu for the job menus.

ANNA:
In the jobs pages we see "unknown" ; if we have no info ,the efficiency is calculated wrong. We need to add a task in Redmine about how we calculate efficiencies; if not all the needed info is available, do not put a number.
Summary of MCC9 samples: need to finish, to recover from the disk-full problem.
Anna is doing all this for ProtoDUNE, by herself, they don't have yet a "production team" but two people at CERN are submitting jobs using POMS.
What about v2.1.1 release? It's 2 weeks overdue..

MARC:
2 issues, 14287, basically done; 11963 also almost done.
Trying to figure out the memory 'hammering'. We have to restart ever 4 hours, this is a kludge, not a fix and we need to find a solution..

ANNA:
About Kibana plots: for each job, show memory requested,min and max used and duration time.

VLADIMIR:
16403: about changing tag to version, Steve maybe removed the code that generates that..

ANNA:
Maybe this is not crucial, users don't need to see POMS version.. they don't have the choice like a version sw code.
What about pythin 3 migration?

MARC:
Found few minor issues but it's done. Submitted many campaigns but not found any memory leak.
We need to check our stats: the number of job submitted doesn't show up until reported by Condor.

ANNA:
Any other issues in the Calendar? This should work for next release.
How about POMS client? Felipe is working on it and we should plan to release it soon.
We need to discuss the "inline" guide .Vladimir has this already implemented in devel. Hints shows up and clicking , go to Redmine page; however, we need to make sure
there is full info available in Redmine. Felipe will check what's available and add.

BRIAN:
Comment that was very good that Anna worked with ProtoDUNE on MCC9 and had so much accomplished.

ANNA: ProtoDUNE needs to provide people to learn what to do, there is still no production group in Dune.
About full disk issue: there is no clear policy on how to cleanup. Need to decide how and what to keep stuff in DCache or store on tape.
Another good aspect is that we are doing the exercise to copy data to CERN. Will follow the opposite: running jobs at CERN and transfer them to FNAL in preparation of real data to come.

MARC:
It's important to show that we can have files in both places.

ANNA:
We still need to improve the handling of files used by DETSIM-Reco-MergeAna.
File produced by previous stage should not be kept once next stage is finished.
This would save space on disk.

Action Items:

  1. Put task in redmine to show dependent campaign in Campaign info page
  2. Felipe will check all info in the inline guide are available in redmine.