[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[HTCondor-users] priority factors published by condor_userprio does not change with configuration files -- condor version 8.0.6



Dear Condor Expert:

Recently we found a potential bug in condor version 8.0.6. This is firstly found at Glasgow Tier2 and could be repeated at RAL Tier1.

When we change the priority factors of groups in the fairshare configuration file on the central-manager, condor accepts this change but condor_userprio does not show this change. To see the change we need to delete the user account in that group and resubmit jobs, or use condor_userprio -setfactor to change it directly.

    For example, At the very beginning, I set
GROUP_PRIO_FACTOR_group_ATLAS           = 100000.00
GROUP_PRIO_FACTOR_group_NONLHC          = 100000.00

    condor_userprio shows the factors :

group_ATLAS:   100000.00
group_NONLHC: 100000.00

    This is as expected, then I changed the settings to:
GROUP_PRIO_FACTOR_group_ATLAS           = 1.00
GROUP_PRIO_FACTOR_group_NONLHC          = 1.00

    condor_userprio still shows the prevous factors:

group_ATLAS:   100000.00
group_NONLHC: 100000.00

but the new users in these two groups would have correct priority factors looks like:

-- scotg001 is in group_ATLAS, and scotg002 is in group_NONLHC

group_ATLAS.scotg001: 1.00
group_NONLHC.scotg002: 1.00

    Then I changed the settings to:
GROUP_PRIO_FACTOR_group_ATLAS = 100.00 -- from 1 to 100
GROUP_PRIO_FACTOR_group_NONLHC          = 1.00

    restarting everything and resubmit jobs, it will show:
group_ATLAS:   100000.00
group_NONLHC: 100000.00
group_ATLAS.scotg001: 1.00 ---- not changed from 1 to 100
group_NonLHC.scotg002: 1.00

conodor_userprio shows not change. I removed the account group_ATLAS.scotg001

-- condor_userprio --delete group_ATLAS.scotg001@xxxxxxxxxxxxxxx

    and then resubmit jobs, this time I see it's correctly changed:

group_ATLAS:   100000.00
group_NONLHC: 100000.00
group_ATLAS.scotg001: 100.00              -- changed from 1 to 100
group_NonLHC.scotg002: 1.00

If this is true, it will mean that we can't rely on condor_userprio to get the instant proper information which is quite sad.

Also condor_userprio -setfactor can only be used to apply on user level, do we have similar tools to apply on group level?

    Cheers,Gang