[gridengine users] Max load per execution host
w.hay at ucl.ac.uk
Wed Jul 3 07:58:21 UTC 2013
On Wed, 2013-07-03 at 07:01 +0000, Guillermo Marco Puche wrote:
> I've experienced some problems on my SGE cluster.
> Sometimes compute nodes go down if the CPU load is too high. RAM
> consumption is ok. I know you can limit the memory limit per job. I
> would like to know if there's any way to set a max CPU load per
> compute node (execution host) or per job. So I can prevent my nodes
> from crashing.
> Thank you.
You could set the queue up to start suspending jobs if the load gets too
high. This doesn't work too well for parallel (multi-host jobs) but
these don't usually cause this sort of problem so the simplest solution
would be to put them in a different queue that isn't suspended on load.
> Best regards,
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 836 bytes
Desc: This is a digitally signed message part
More information about the users