At CIT /etc/condor/config.d/99-request-disk
and 99-memory
add a submit requirement for users to specify request_disk
and a default memory request, respectively.
Again, we have the caveat that the behavior is different on IGWN grid access points: there is no disk requirement and I don't see the transform for memory but it apparently defaults to 1024.
This MR implements #5.
Duncan Macleod (d8fe7dc3) at 02 Feb 14:02
Merge branch 'requestdisk-transform' into 'main'
... and 4 more commits
This MR implements #5.
Looks fine, will approve with a suggestion to explicitly note in the docs that this should not be applied on CEs where we exclude the pilot user(s) (this config is applied to IGWN pool APs, though).
@james-clark, can you please review these changes?
One of the most frequent issues users encounter in a Condor pool, and one of the most difficult for them to understand and debug, is held jobs. One important incremental improvement to the UX is breaking the multi-part SYSTEM_PERIODIC_HOLD expression into sub-expressions via SYSTEM_PERIODIC_HOLD_NAMESso that Condor can tell the user (and/or admin) which component of the expression "fired" and needs to be addressed.
One of the most frequent issues users encounter in a Condor pool, and one of the most difficult for them to understand and debug, is held jobs. One important incremental improvement to the UX is breaking the multi-part SYSTEM_PERIODIC_HOLD expression into sub-expressions via SYSTEM_PERIODIC_HOLD_NAMESso that Condor can tell the user (and/or admin) which component of the expression "fired" and needs to be addressed.
This MR implement #9.
This configuration was imported from ldas-grid.ligo.caltech.edu
on Fri Nov 24 08:15:28 PST 2023.
Duncan Macleod (2994d2c7) at 18 Jan 14:58
Merge branch 'system-periodic-hold' into 'main'
... and 1 more commit
We should centralise the required configuration setting for DAGMAN_USE_DIRECT_SUBMIT = False
.
This MR implements #11.
Duncan Macleod (54269a9c) at 17 Jan 12:05
Merge branch 'dagman-direct-submit' into 'main'
... and 1 more commit
All access points (formerly submit machines) will need to run the credd
and credmon_oauth
daemons to support user tokens in jobs, so we should pick the best implementation we have and centralise that.
All access points (formerly submit machines) will need to run the credd
and credmon_oauth
daemons to support user tokens in jobs, so we should pick the best implementation we have and centralise that.
This MR closes #7 by adding a new vault configuration file.
Duncan Macleod (75184eb1) at 17 Jan 11:01
Merge branch 'vault-config' into 'main'
... and 1 more commit
Once that is the 'norm', we can configure that.
This MR implement #9.
This configuration was imported from ldas-grid.ligo.caltech.edu
on Fri Nov 24 08:15:28 PST 2023.
Not certain but IIRC this was related to the EL7->EL8 transition