search

dmwm / WMCore

Core workflow management components for CMS.
Apache License 2.0
45 stars 106 forks source link

WQE created with wrong priority #7126

Closed vlimant closed 7 years ago

vlimant commented 7 years ago

https://cmsweb.cern.ch/reqmgr/view/details/prozober_ACDC_ACDC_recovery-0-fabozzi_HIRun2015-HIForward-02May2016_758p4__160830_112011_5843

priority is 900k,

has one WQE

https://cmsweb.cern.ch/couchdb/workqueue/_design/WorkQueue/_rewrite/elementsInfo?request=prozober_ACDC_ACDC_recovery-0-fabozzi_HIRun2015-HIForward-02May2016_758p4__160830_112011_5843

priority was 85k. Had to change the priority of the wf to 900k to get the wqe aligned.

ticoann commented 7 years ago

Does this happen only ACDC? Could you tell me the steps to reproduce the problem?

vlimant commented 7 years ago

https://cmsweb.cern.ch/couchdb/workqueue/_design/WorkQueue/_rewrite/elementsInfo?request=jen_a_recovery-6-fabozzi_Run2016B-2-SingleElectron-01Jul2016_8013p1__160812_182758_1667

another one. Not an ACDC. No clear way to reproduce.

vlimant commented 7 years ago

N.B.

jen_a_recovery-7-fabozzi_Run2016B-2-SingleElectron-01Jul2016_8013p1__160812_182816_2206.request.priority = 87000

jen_a_recovery-7-fabozzi_Run2016B-2-SingleElectron-01Jul2016_8013p1__160812_182816_2206.request.schema.RequestPriority = 900000.0

amaltaro commented 7 years ago

How were these workflows created? Can you point which script you are using?

amaltaro commented 7 years ago

Are they assigned within Unified or also standalone scripts? If the latter, please point us to the code.

ticoann commented 7 years ago

Especially please provide which script creates this workflow. (I looked at the code for generating reqmgr1 workflow but I couldn't find specific place yet)

jen_a_recovery-7-fabozzi_Run2016B-2-SingleElectron-01Jul2016_8013p1__160812_182816_2206

vlimant commented 7 years ago

not by unified. Looks like it was assigned by hand, created from makeACDC and recoveryMissingLumis.py @prozober @jenimal

jenimal commented 7 years ago

any workflow with "recovery" in it's name was created by the recoveryMissingLumis.py script and assigned by hand, using the web interface. The acdc's of the recovery were most likely made, and assigned via the web interface as well. Jen

ticoann commented 7 years ago

@jenimal, Jen can you show me or point me the exact procedure? I would like to reproduce on vm or at least on the testbed.

jenimal commented 7 years ago

https://twiki.cern.ch/twiki/bin/view/CMSPublic/RecoveringWorkflows

Enjoy

ticoann commented 7 years ago

FYI

I am trying to create the recovery workflow in testbed. (modify the script to point to testbed)

python WmAgentScripts/recover/recoverMissingLumis.py -q sryu -g DATAOPS -r amaltaro_ReReco_Parents_Integ_Test_Sept2016_Patches_160914_104332_3349

Not related to this problem, but one thing I notice is that following block name comes back with se names. which shouldn't. Blockname: /Cosmics/Commissioning2015-PromptReco-v1/RECO#77f9d258-b5d6-11e4-8732-02163e00d625 Locations: set([u'srm-eoscms.cern.ch', u'eoscmsftp.cern.ch', u'cmsdcatape01.fnal.gov', u'cmsdcadisk01.fnal.gov'])

Although I am not sure where this list comes from. Calling directly dbs gives me PhEDEx node name

https://cmsweb.cern.ch/dbs/prod/global/DBSReader/blockorigin?block_name=/Cosmics/Commissioning2015-PromptReco-v1/RECO%2377f9d258-b5d6-11e4-8732-02163e00d625

[{"create_by": "tier0@vocms15.cern.ch", "creation_date": 1424108239, "open_for_writing": 0, "last_modified_by": "tier0@vocms15.cern.ch", "dataset": "/Cosmics/Commissioning2015-PromptReco-v1/RECO", "block_name": "/Cosmics/Commissioning2015-PromptReco-v1/RECO#77f9d258-b5d6-11e4-8732-02163e00d625", "file_count": 89, "origin_site_name": "T2_CH_CERN", "last_modification_date": 1424108239, "block_size": 347495323824}]

amaltaro commented 7 years ago

Fixed by #7228