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

Re: [Condor-users] Max Limit with Condor Queue

On 8/9/06, Al Ramsey <aramsey@xxxxxxxxx> wrote:
Is there a default limit to the number of jobs that can be in a queue?
I couldn't confirm after referring to the Condor manual.

Thank you in advance for any help/advice with this question.

I would guess that the limit on *clusters* is theoretically 2 Billion
(clusterid is referenced in at least one place as an int).

I would expect any schedd attempting to hold that many to die well
before that during negotiation (if not submit time).

I find that If I have more than about 500 clusters available to run
then negotiation pretty much kills my schedd (stops responding, jobs
end up dying). That said this is with 6.6 so 6.8 should be able to
push this number up considerably if the job requirements are similar

AS to how many jobs can be contained if held I've seen several
thousand on one machine quite happily (at the same time as it was
servicing a few hundred shadows).

The answer is therefore - however many your machine is capable of
handling. There is no way you will hit the 2 billion no matter what
your hardware.

Note that if you submit and finish 10 single process clusters every
second (ridiculously unachievable throughput) then you would reach the
max number of lifetime cluster ids in nearly 7 years.

If this unfortunate situation should happen to you simply deleting the
spool/job_queue.log will fix it :)