Prediction in the past: is it normal?

[bzizou]

bzizou@dahu-oar3:~$ oarsub -I -l /nodes=4,walltime=00:30:00 -t devel --project test
# INFO:  Moldable instance:  1  Estimated nb resources:  128  Walltime:  1800
# INFO:  Moldable instance:  1  Estimated nb resources:  128  Walltime:  1800
# INFO:  Moldable instance:  1  Estimated nb resources:  128  Walltime:  1800
OAR_JOB_ID=7250
Interactive mode: waiting...
[2024-06-12 11:30:58] Start prediction: 2024-06-12 11:09:04 (R=128,W=1860,J=I,Q=default)

The start prediction is is the past... is it normal?

[adfaure]

Hi, I could not reproduce this bug with a simple test...

Does it happen often? Is the database clock in sync with the frontend? How long did it take to reach this prediction? Is the queue empty?

Anything that could lead us to reproduce it into a simple test could be helpful.

Thank you for reporting, very intriguing.

[bzizou]

Hi, I could not reproduce this bug with a simple test...

Does it happen often?

Could not reproduce it

Is the database clock in sync with the frontend?

postgres@dahu-oar3:~$ date ; psql oar -c "select now();"
Wed Jul  3 08:05:04 AM CEST 2024
              now              
-------------------------------
 2024-07-03 08:05:04.286853+02

How long did it take to reach this prediction? Is the queue empty?

It took a few seconds. Checking into the database, the submit_time is 1718184635 which gives "11:30:35" and the prediction came at "11:30:58"
Pretty sure that the queue was empty, and there was 2 nodes available and 2 other to be started by Greta (standby)

Anything that could lead us to reproduce it into a simple test could be helpful.

Play with greta and nodes in the standby mode I think...

Thank you for reporting, very intriguing.