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

[HTCondor-users] condor_userprio -allusers and condor_userprio differ in output and unable to change -setceil



Hello HTCondor experts,

I'm looking for some help with the condor_userprio settings. We've implemented condor_userprio -setfloor and -setceil commands with a cron job with the basic commands: condor_userprio -setfloor $user@xxxxxxxxx 5; condor_userprio -setceil $user@xxxxxxxxx 500; etc.

I went to change these settings and had different behaviour than expected. All commands are run on the central manager with the main negotiator for the pool. Below is how I had gone about things. Maybe someone else here can point out what I've missed or did incorrectly?

##################

1. I went to change the user ceiling to 700 based on the commands listed aboce and if I use condor_userprio -allusers, I see the update values that makes sense

root@stbc-019:~# condor_userprio -allusers
Last Priority Update:Â 6/10 10:50
 Effective Priority Wghted Total Usage Time Since Submitter User Name Priority Factor In Use (wghted-hrs) Last Usage Ceiling ---------------------- ------------ --------- ------ ------------ ---------- --------- XXXX@xxxxxxxxx 500.00 1000.00 0 0.00 19884+08:5 700.00

2. in practice, users are still being caped at the original 500 job ceiling. I ran condor_userprio and found a different set of boundaries:

root@stbc-019:~# condor_userprio
Last Priority Update:Â 6/10 10:49
 Effective Priority Wghted Total Usage Time Since Submitter User Name Priority Factor In Use (wghted-hrs) Last Usage Ceiling ------------------ ------------ --------- ------ ------------ ---------- --------- maryh@xxxxxxxxx 16477.03 1000.00 1 4419.41 <now> 500.00 templon@xxxxxxxxx 477432.09 1000.00 470 127870.44 <now> 500.00 ------------------ ------------ --------- ------ ------------ ---------- ---------
Number of users: 3ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ 631ÂÂÂ 147251.58 0+23:59

3. So now, I was thinking, "well, I need to reset the user ceiling limit then". I tried a condor_userprio -resetall|resetusage|delete. These commands completed with what looked like correct messages. For example,

root@stbc-019:~# condor_userprio -resetusage templon@xxxxxxxxx
The accumulated usage of templon@xxxxxxxxx was reset

But the user cap for jobs remains the same (i.e., condor_userprio and condor_userprio -allusers show different output but are implemented differently?!)

One of my colleagues found this in the source code:

 // knob to disable negotiator modular direct query, in case this causes problems (the results *are* a bit different)
 if ( ! param_boolean("USERPRIO_USE_NEGOTIATOR_MODULAR_QUERY", false)) {
ÂÂÂÂÂÂÂÂÂ negotiatorCanDoDirect = false;
 }

But this is not a documented option and/or also not set when looking at a condor_config_val -dump.

Any pointers or advice are most welcome!

Thanks,

Mary