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

Re: [Condor-users] Can not get RemoteWallClockTime or CommittedTime report any value other than 0.0



Hi Doug:

On Sun, 2011-04-24 at 21:59 -0500, Doug Benjamin wrote:
> 
> SYSTEM_PERIODIC_REMOVE = ((TARGET.RemoteWallClockTime -
> TARGET.CumulativeSuspensionTime ) > $(Job_Time_Limit))
> 

> Job_Time_Limit = ( ( ifThenElse(( TARGET.AccountingGroup =?= "short_job"
> ),2 * 60,ifThenElse(( TARGET.AccountingGroup =?= "medium_job" ),10 *
> 60,15 * 60)) ) )

I just ran a simple test on our 7.6 pool - my periodic remove expression
worked appropriately. I think you might be getting caught by the fact
that periodic expressions are evaluated by the schedd (with an interval
somewhere between PERIODIC_EXPR_INTERVAL and MAX_PERIODIC_EXPR_INTERVAL,
which evaluate to 60 and 1200 seconds by default).

In addition, my guess is that the RemoteWallClockTime is probably
updated by the startd using UPDATE_INTERVAL (which defaults to 5
minutes). Therefore, you probably will not be able to use a periodic
remove expression with a higher resolution that 5 minutes unless you
change UPDATE_INTERVAL on your startds.

FYI - As my test, I submitted with 
PeriodicRemove = RemoteWallClockTime > 60

The log file contains:
001 (1231.000.000) 04/24 22:34:20 Job executing on host:
<10.0.2.15:52206?CCBID=134.48.90.158:9618#41903>
...
009 (1231.000.000) 04/24 22:39:20 Job was aborted by the user.
	The job attribute PeriodicRemove expression 'RemoteWallClockTime > 60'
evaluated to TRUE

Note the 5 minutes between the submission and the removal.

Cheers,
DJH