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

[Condor-users] Disabling Preemption for a subset of users...



Hello All,

For sets of machines in my pool that have been purchased by individual
research groups, memebrs of those groups are meant to have preemptive
access to those nodes and to never be preempted themselves (even by
other group members.

I thought I had this all worked out using RANK, SUSPEND, and PREEMPT
in the condor_config.local, but in atleast the newest set of these
systems I'm seeing a case where one group member preempted the jobs of
another (causing him to loose 2 days of progress as their jobs expect
to run to completion).

Here are the runtime values I belive are relevent, there's a bit too
them fo rother reasons, but the && ( (owner == "foo") == False) at the
end should keep these users form ever being suspended or preempted, yes?

[jon@borg-login-1 ~]$ ssh btcs-2400-003 condor_config_val RANK

(( owner == "drhagen" || owner == "nsilver" || owner == "bking" ||
owner == "cuiyy" || owner == "nirmala" || owner == "ispatel" || owner
== "jbiddle" || owner == "yshen" || owner == "filipe" )* 10000000)


[jon@borg-login-1 ~]$ ssh btcs-2400-003 condor_config_val SUSPEND

((( (KeyboardIdle < 60) || ( (CpuBusyTime > 1 * 60) && (CurrentTime -
JobStart) > 90 ) )) || (SlotID == 1 && Slot1_RequiresWholeMachine =?=
True && (Slot2_Activity =?= "Busy" || Slot3_Activity =?= "Busy" ||
Slot4_Activity =?= "Busy" || Slot5_Activity =?= "Busy" ||
Slot6_Activity =?= "Busy" || Slot7_Activity =?= "Busy" ||
Slot8_Activity =?= "Busy" || Slot9_Activity =?= "Busy" ||
Slot10_Activity =?= "Busy" || Slot11_Activity =?= "Busy" ||
Slot12_Activity =?= "Busy" || Slot13_Activity =?= "Busy" ||
Slot14_Activity =?= "Busy" || Slot15_Activity =?= "Busy" ||
Slot16_Activity =?= "Busy" ) )) && ( (owner == "drhagen" || owner ==
"nsilver" || owner == "bking" || owner == "cuiyy" || owner ==
"nirmala" || owner == "ispatel" || owner == "jbiddle" || owner ==
"yshen" || owner == "filipe") == False)

[jon@borg-login-1 ~]$ ssh btcs-2400-003 condor_config_val PREEMPT

((( ((Activity == "Suspended") && ((CurrentTime -
EnteredCurrentActivity) > 1 * (60 * 60))) || (SUSPEND && (WANT_SUSPEND
== False)) )) || (ifThenElse( isUndefined(MemoryRequirementsMB),
ImageSize/1024*0.7, MemoryRequirementsMB ) > (Memory*0.9) ||
ImageSize/1024 > (VirtualMemory*0.9))) && ( (owner == "drhagen" ||
owner == "nsilver" || owner == "bking" || owner == "cuiyy" || owner ==
"nirmala" || owner == "ispatel" || owner == "jbiddle" || owner ==
"yshen" || owner == "filipe") == False)

-Jon