Closed vlimant closed 1 week ago
Requests are removed from the forcecomplete
upon set_status()
once they reach status new
or done
, here:
set_status
is called among others from inspect
(inspect
-> inspect_submitted
-> set_status
) once a submitted request reaches announced
or normal-archived
. inspect
is normally called automatically twice a day by the McM Request Inspect and Flow
Jenkins job.
However in the latest run it seems that there were quite a few authentication errors. @ggonzr is this line supposed to work with the new SSO?
Yes, it works with the latest SSO. The cookie is requested by the auth-get-sso-cookie
package and its path is sent to the downstream process. However, it seems the executions #829, #827 had issues using the requested cookie. Other runs like #828 and #830 finished properly. I will check this and see if I can reproduce the issue again.
Thanks Geovanny!
This rare issue seems to be happening due to a problem with the script mechanisms to re-validate the session. The following PR solves this situation: #20
one thing that can be done for now is to remove from https://cms-pdmv-prod.web.cern.ch/mcm/restapi/requests/forcecomplete anything that is already in "done" status. is there an API to do this one by one ? otherwise we need a direct edit in the DB.
then we take care of the remaining ones (likely that will appear in the "dead" limbo
lists/update
lets administrators reset the whole list (similar to a manual db update).
what's the format for the PUT in this case? updating that list is a thing tech support can also do,
what's the format for the PUT in this case?
The same format as in the database (I don't know offhand what it looks like).
updating that list is a thing tech support can also do,
Only @ggonzr knows the database password.
hit #1154 on the way
Looks like the list https://cms-pdmv-prod.web.cern.ch/mcm/restapi/requests/forcecomplete does not get updated once a request does get acted upon and goes into "done" status. This issue is to review what is the mechanism under that list and how to implement a cleanup mechanism (if needed) so that the list does not grow forever big.
The desired behaviour is that an entry in https://cms-pdmv-prod.web.cern.ch/mcm/restapi/requests/forcecomplete is set by operating in McM, gets picked up by unified/ops (https://gitlab.cern.ch/CMSProductionReprocessing/WmAgentScripts/-/blob/master/Unified/completor.py?ref_type=heads#L87)