Include maintenance node in HTCondor playbook #1057
Conversation
kysrpex
commented
Dec 19, 2023
kysrpex
commented
- Add maintenance node to the HTCondor playbook
- Remove usegalaxy_eu.htcondor role from the maintenance playbook
- Remove vars specific to usegalaxy_eu.htcondor role from maintenance group
Let the maintenance node play the submitter role in the HTCondor playbook. This is achieved by adding the maintenance node to the `htcondor-secondary-submit` group, a subgroup of the `htcondor` group, that the HTCondor playbook targets.
`usegalaxy_eu.htcondor` has now been replaced with `grycap.htcondor` (part of the HTCondor playbook).
… group The role `usegalaxy_eu.htcondor` no longer runs on the maintenance node, thus its role variables are no longer needed.
Do not run `usegalaxy-eu.htcondor_release`, `usegalaxy-eu.fix-stop-ITs` nor `usegalaxy-eu.remove-orphan-condor-jobs` on the maintenance node.
|
@sanjaysrikakulam Add whatever stuff may still be missing. I suggest to run this by calling |
|
I created the htcondor Jenkins project as discussed. It runs and is enabled. |
Cool, thank you! |
| @@ -294,11 +294,11 @@ | |||
| when: htcondor_role_submit | |||
| - grycap.htcondor | |||
| - name: usegalaxy-eu.htcondor_release | |||
| when: htcondor_role_submit | |||
| when: htcondor_role_submit and inventory_hostname != "maintenance.galaxyproject.eu" | |||
There was a problem hiding this comment.
I guess this will be removed again once the nspawn container stops doing that?
(just for my better understanding)
There was a problem hiding this comment.
The problem here is that usegalaxy-eu.htcondor_release is designed to run on max one submitter node (this is of course a bad solution and needs to be addressed later, but we want to have the monitoring working properly too). If we do not prevent usegalaxy-eu.htcondor_release from running on more than one submitter node, then #896 comes back.
So unfortunately no, this will not be removed again once the systemd-nspawn container is decommissioned. It will not be as simple as that :(
There was a problem hiding this comment.
Manually deployed and tested. Discussed with @kysrpex regarding the various issues found during the deployment, and all of them were fixed via different PRs in respective repos.
The dashboard is back online (except for a couple of panels, need to look into it). So, merging this PR.
