Be careful interpreting Queued_time and/or calculations of the difference
between Submit_time and Start_time.

Many factors contribute to time spent waiting in the queue... some of which 
have more to do with how 'busy' the sysadmin and/or attentive the user is than 
how busy the machine is.

Some examples...

- Job dependencies (that can't EVER be satisfied)
- Job at MaxRequeueHold limit
- jobs submitted per assoc vs. jobs running per assoc and relationship to 
job/partition QOS
- Jobs submitted to spec_nodes that aren't 'available' (ie. down, maint or resv)
- ... the list goes on and on ...

Trevor

> On May 8, 2016, at 3:06 AM, Chris Samuel <[email protected]> wrote:
> 
> 
> On Saturday, 7 May 2016 7:20:04 PM AEST Ryan Novosielski wrote:
> 
>> I realize that I probably wasn't clear enough if you didn't already know
>> exactly what I was talking about; "Queued time" in this context and in
>> the e-mail status message refers to "time spent in the queue" before the
>> job was able to run.
> 
> Ahh, sorry, I misread you as wanting the time the job entered the queue.
> 
> I doubt Slurm would keep that queued time as it can be simply calculated from 
> the submit and start times, which I suspect is what XDMoD does to derive its 
> "wait time" for jobs:
> 
> http://xdmod.sourceforge.net/
> 
> All the best,
> Chris
> -- 
> Christopher Samuel        Senior Systems Administrator
> VLSCI - Victorian Life Sciences Computation Initiative
> Email: [email protected] Phone: +61 (0)3 903 55545
> http://www.vlsci.org.au/      http://twitter.com/vlsci

Reply via email to