Re: [Queue-developers] Error running jobs

[W. G. K. <wer...@ya...>]

This is correct behavior.

The load average is divided by vmaxexec minus number of running jobs or
one, whichever is greater, and then multiplied by p-factor.

In this case, the load is divided by two.

This helps to give unused machines a slight leg-up in the competition.
(p-factor should be set to discourage slow machines from getting jobs.)

Sam Liddicott wrote:

> Now I fixed the syntax errors in the profiles, I get:
>
> wakeup.c getrldavg(): po.ananova.net returned load 5.72e+10.
> Queue "now" at host "po.ananova.net" has rldavg of 5.72e+10.
> The host "po.ananova.net" is not able to serve queue "now".
> Failed to submit job in queue "now" to host "po.ananova.net".
>
> an exceedingly high load.
>
> /tmp/queue/queued.debug:
> po.ananova.net[281]: queued queued.c check_query():
>         The "now" queue: q_status1:
>         "now: enabled: maxexec=2 loadsched=25 loadstop=50 nice=0 cpu
> 71582788 min
> "
> po.ananova.net[281]: queued queued.c check_query():
>         The "now" queue: q_oldstat = -1.
> po.ananova.net[281]: queued queued.c check_query():
>         calculating load... avg = 0.160.
> po.ananova.net[281]: queued queued.c check_query():
>         calculated load = 0.387.
> po.ananova.net[281]: queued queued.c check_query():
>         Queue "now": load average query response:  0.39 (0x3ec60000).
> po.ananova.net[281]: queued queued.c check_query():
>         Load average 0.160156,  vmaxexec 2,  nexec 0,  pfactor 1.
> po.ananova.net[281]: queued queued.c check_query():
>         select()ing on sockets: 6 and 7...
>
> Which all looks about right.
> So how did queue so badly misread the load?
>
> Sam
>
> _______________________________________________
> Queue-developers mailing list Que...@li...
> To unsubscribe, subscribe, or set options:
> http://lists.sourceforge.net/lists/listinfo/queue-developers