Reuti <[email protected]> writes:

> On the exechost? I don't do it at all on a per job basis. In case your users 
> fight for the disk space you can implement a consumable for the disk space in 
> combination with a load sensor:
>
> http://gridengine.org/pipermail/users/2012-February/002914.html

For what it's worth, the flex-grid load sensor reports on /tmp space.

> (there are some other points in the thread too, like mounting a limited loop 
> device on $TMPDIR)

Using the supplementary gid to enforce a quota (assuming your filesystem
an be quota'd) seems most obvious and straightforward -- that's what
identifies a job POSIX-wise.  (The talk about filesystems in that thread
is odd.)  Is there a good reason not to?

In case anyone wants to try, and it's not clear:  with multi-node jobs,
the job-wise complex you need is debited on the master node, and you'd
need to set a trivially small quota on the slaves.  That's assuming the
relevant i/o is on the master node.  The supplementary gid is available
in the job spool -- I forget the file name.

-- 
Community Grid Engine:  http://arc.liv.ac.uk/SGE/
_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users

Reply via email to