[slurm-users] sacct got the job record without partition info

[刘文晓]

Dear all，


Now， I found when I submitted a job with error "Requested node configuration is 
not available" on Slurm 19.05.5,  sometimes with "sacct" command i get no 
partition record. like below:




First use command get the job:


Wait several seconds:


In my local lab, I use the accounting_storage/slurmdbd plugin.





After add "DebugFlags=DB_JOB"  in slurmdbd.conf, found that Slurmdbd first 
receive DBD_JOB_COMPLETE messages without partition info to insert job table. 
After several seconds(max 5s), it receives DBD_JOB_START, it updates the job 
record's partition info from plugin's _set_db_inx_thread of 
accounting_storage/slurmdbd.




Is this problem fixed in new version? Or how to fix it?




Thanks