Closed amaltaro closed 7 years ago
I didn't manage to reproduce this issue in my VM. Just in case, this is the content of one of the GQE:
{"_id":"029bec0a615bdfa28e2a252e34e8cc71","_rev":"1-de6deb4cc51f0dab404f9d253e5760da","updatetime":1489020225.3563029766,"WMCore.WorkQueue.DataStructs.WorkQueueElement.WorkQueueElement":{"ParentQueueId":"pdmvserv_HIN-pPb502Winter16GS-00003_00003_v2_MB_170309_001506_524","NumberOfEvents":478000,"StartPolicy":"MonteCarlo","CreationTime":0,"NumOfFilesAdded":0,"Priority":85000,"ParentData":{},"PileupData":{},"RejectedInputs":[],"ParentFlag":false,"TaskName":"Production","Status":"Available","Inputs":{},"NumberOfLumis":5000.0,"Jobs":1000.0,"ParentQueueUrl":null,"ChildQueueUrl":null,"PercentSuccess":0,"PercentComplete":0,"WMBSUrl":null,"OpenForNewData":false,"EndPolicy":{"policyName":"SingleShot"},"FilesProcessed":0,"SiteWhitelist":[],"ACDC":{},"NumberOfFiles":0,"TeamName":"production","NoPileupUpdate":false,"RequestName":"pdmvserv_HIN-pPb502Winter16GS-00003_00003_v2_MB_170309_001506_524","SubscriptionId":null,"NoInputUpdate":false,"blowupFactor":1.0,"TimestampFoundNewData":0,"ProcessedInputs":[],"SiteBlacklist":[],"Mask":{"LastRun":1,"FirstRun":1,"inclusivemask":true,"runAndLumis":{},"LastEvent":17208000,"FirstEvent":16730001,"LastLumi":180000,"FirstLumi":175001},"Dbs":null,"EventsWritten":0},"timestamp":1489020225.3563029766,"thunker_encoded_json":true,"type":"WMCore.WorkQueue.DataStructs.WorkQueueElement.WorkQueueElement"}
I was going to say it's related to the fact that this TaskChain starts with MC from scratch AND has the TrustSitelists
flag enabled. However, the previous workflow reported was not assigned with any of those Trust flags enabled.
I'm looking at this issue and we should have it fixed for the next cmsweb production upgrade (next week).
As reported here: https://its.cern.ch/jira/browse/CMSCOMPPR-699
the workflow has a proper SiteWhitelist (and it was properly assigned) and even so the GQE have an empty SiteWhitelist list.
For the record, this is the dict posted during assignment
global_workqueue logs show the following:
as can be seen, it tries to split work again twice. Checking...