troycomi / reportseff

Tabular seff

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

List duplicated jobs

ggwena opened this issue · comments

Requeued jobs with identical jobId are not shown, only last one is.

Duplicated jobids are shown by sacct with '--duplicates'.

I use : --extra-args=-D to get additional jobs not included by default

Although sacct output contains requeued jobs, they are not rendered in output.

Expected behaviour

Get the list of all restarted jobs AND the last COmpleted one.

How to reproduce

reportseff --debug --extra-args=-D -s COMPLETED,REQUEUED,FAILED 231374

Output example:

$ reportseff --debug --extra-args=-D -s COMPLETED,REQUEUED 235929
|1|01:04:49|235929|235929||1|5000Mc|REQUEUED|02:00:00|01:04:42
|1|01:04:49|235929.batch|235929.batch|3960K|1|5000Mc|FAILED||00:00.106
|1|01:04:47|235929.0|235929.0|13633728K|1|5000Mc|FAILED||01:04:41
|1|01:14:33|235929|235929||1|5000Mc|REQUEUED|02:00:00|01:14:16
|1|01:14:33|235929.batch|235929.batch|4140K|1|5000Mc|FAILED||00:00.089
|1|01:14:32|235929.1|235929.1|13801976K|1|5000Mc|FAILED||01:14:16
|1|01:42:02|235929|235929||1|5000Mc|REQUEUED|02:00:00|01:41:44
|1|01:42:02|235929.batch|235929.batch|4092K|1|5000Mc|FAILED||00:00.118
|1|01:41:59|235929.2|235929.2|13771372K|1|5000Mc|FAILED||01:41:44
|1|00:35:53|235929|235929||1|5000Mc|REQUEUED|02:00:00|35:45.828
|1|00:35:53|235929.batch|235929.batch|4092K|1|5000Mc|FAILED||00:00.085
|1|00:35:51|235929.3|235929.3|14782120K|1|5000Mc|FAILED||35:45.742
|1|00:24:20|235929|235929||1|5000Mc|COMPLETED|02:00:00|24:12.966
|1|00:24:20|235929.batch|235929.batch|4092K|1|5000Mc|COMPLETED||00:00.086
|1|00:24:18|235929.4|235929.4|14591148K|1|5000Mc|COMPLETED||24:12.879

   JobID    State       Elapsed  TimeEff   CPUEff   MemEff 
  235929  COMPLETED    01:04:49   20.3%    99.5%    288.7% 


$ reportseff --debug --extra-args=-D -s REQUEUED 235929
...
   JobID   State       Elapsed  TimeEff   CPUEff   MemEff 
  235929  REQUEUED    01:04:49   29.9%    99.6%    288.7% 

Thanks for the tool, btw!

I haven't come across requeued jobs yet, thanks for the report!

Currently, the job ids uniquely define a job. I need to think how to support this.

Do you know if the 235929.4 indicates the job was requeued 4 times?

Yes, i think so.