On 24 May 2012 17:11, Rayson Ho ray...@scalablelogic.com wrote:
But note that I am not against Univa - I believe Univa is improving
Neither am I (except in jest).
William
___
users mailing list
users@gridengine.org
On Thu, 24 May 2012, Rayson Ho wrote:
...
Not trying to launch any GPLv2 vs GPLv3 arguments here, but one of the
reasons why Linux can't be switched to a newer version of the license
is that Linus does not own the copyright.
Or it might have been because he didn't use the optional GPLv2 phrase
On Thu, 24 May 2012, berg...@merctech.com wrote:
...
If possible, it would be extremely helpful to express memory limits as a
percentage of available resources, rather than just as a fixed quantity.
Yes, I do the trick of extracting the swap and RAM sizes for each server
and putting those
Hi list,
I searched through the mailing list and tried using google, but couldn't
found an answer to my question.
The (simplified) situation:
We have a group of nodes with 12 slots.
We submit there parallel jobs requesting 4 or 12 slots.
We defined 2 PEs allocating 4 or 12 slots and associated
On 25 May 2012 10:10, Richard Ems richard@cape-horn-eng.com wrote:
Hi list,
I searched through the mailing list and tried using google, but couldn't
found an answer to my question.
The (simplified) situation:
We have a group of nodes with 12 slots.
We submit there parallel jobs
On 05/25/2012 11:31 AM, Daniel Gruber wrote:
The expected behavior would be that when there is
never an host with 12 slots free, that your cluster
will be filled up with 4 slot jobs, even when they have
lower priorities. The reservation you gave the 12 slot
jobs will be attached at the end
On 25 May 2012 10:31, Daniel Gruber dgru...@univa.com wrote:
The expected behavior would be that when there is
never an host with 12 slots free, that your cluster
will be filled up with 4 slot jobs, even when they have
lower priorities. The reservation you gave the 12 slot
jobs will be
On 05/25/2012 11:51 AM, William Hay wrote:
We have a group of nodes with 12 slots.
We submit there parallel jobs requesting 4 or 12 slots.
We defined 2 PEs allocating 4 or 12 slots and associated to that group
of nodes in one queue.
What do you mean be PE allocating 4 or 12 slots. You have 4
On 05/25/2012 12:04 PM, William Hay wrote:
On 25 May 2012 10:31, Daniel Gruber dgru...@univa.com wrote:
The expected behavior would be that when there is
never an host with 12 slots free, that your cluster
will be filled up with 4 slot jobs, even when they have
lower priorities. The
On 05/25/2012 12:50 PM, William Hay wrote:
But thinking more on it ... perhaps because the 3 jobs J[123]_4 -
started on a 12 slots node N1 - got started at different times, the last
one started could have been started *after* another job J4_12 running on
12 slots on node N2.
So job J3_4 will
Am 25.05.2012 um 12:35 schrieb Richard Ems:
On 05/25/2012 12:27 PM, Daniel Gruber wrote:
Exactly, looks like your runtime estimation for your slot4 jobs
is smaller than for your slot12 jobs. Backfilling must be active
here. Did you submit both jobs in exactly the same way with
s_rt? Try
On Thu, 24 May 2012, Rayson Ho wrote:
...
1) It's not a drop in replacement. If upgrading gridengine on an
existing system, activating your cgroup code will cause an immediate
change in behaviour of jobs, without the user altering their submission
flags. People don't tend to like that sort of
On 25 May 2012 12:27, Mark Dixon m.c.di...@leeds.ac.uk wrote:
It may not be defined by POSIX, but we're free to make up a pragmatic
working definition :)
It could be:
Actual memory usage (typically RAM+swap), as measured by the available
operating-system specific mechanism.
Or OS
Is there an easy way to delete a PE that is no longer referenced by a
queue but which is still matched by a PE wilcard.
[root@admin03 tmp]# qconf -dp qlc-KLB
denied: Pe qlc-KLB is still referenced in job 546941.
Job 546941 requests qlc* which refers both to the PE I want to delete
and various
On Fri, 25 May 2012, William Hay wrote:
Out of curiosity, why do people still want the SSH integration code?
In my case sshd is convenient to wrap if you want to do something to
each task in a job before dropping privileges.
Fair enough.
Out of curiosity why do people still want gridengine
On 25 May 2012 14:00, Fritz Ferstl ffer...@univa.com wrote:
Hi William,
remove it from the queue's pe_list. Then wait for the job(s) to finish and
you can delete it.
I deleted the only queue that referenced it. There were no jobs in
the PE or the queue
that had referenced it. The job it
On 25 May 2012 14:45, Fritz Ferstl ffer...@univa.com wrote:
Hhhhmmm ... would need to look into the code whether it even would
complain about that for a queued job. It's clear for a running job and if
that job was running before and got requeued then it would be easily
explicable, I guess.
On May 25, 2012, at 16:46 , William Hay wrote:
I think I'll leave it for now as the PE doesn't cause any harm I was just
looking to eliminate config cruft.
We had the same problem about half a year ago; our cluster is quite
heterogeneous, and it has a lot of isolated IB switches, so we rely
On Fri, May 25, 2012 at 8:30 AM, William Hay w@ucl.ac.uk wrote:
Or OS dependent, on linux whatever cgroups allows us to control most easily
and seems like a good thing to control.
May be we can ask the Linux kernel guys to give us a better interface??
Out of curiosity, why do people
Ralph: How common will we see jobs that request dynamic allocations? I have
never seen Hadoop presentations talking about them in any BigData conferences.
Just also want to mention that Moab is not open-source, and I don't think we
will see much information about the integration from Moab.
Hi list again,
something seems not be working as I *expect* it to be.
Several jobs are waiting and are asking for a reservation, waiting for
12 slots to become free and are at the top of the priority list.
Several other jobs with less priority and asking only for 4 slots get
backfilled, but
Okay, this going to be a stupid question coming from someone who's been
on this list for years, but here goes...
I've just upgraded the RAM on a few cluster nodes to 32 GB (instead of
16 GB). A few users could benefit from this, so I'd like to be able to
specify h_vmem or mem_free or s_vmem for
Hi list,
what happens if there are more reservations being requested by waiting
jobs asked for than the value set in max_reservations?
Do reservations on running jobs still count for the scheduler
reservation plans?
Is there any script to analyze the output of the schedule file?
Thanks,
If I wanted to be sure that a particular parallel job ran only
on a single client node (as opposed to being spread across
multiple nodes in our cluster), is there a good way to do that
under gridengine? To be clear, I don't care which client is
used, but I want the parallel job to run on a
On 25 May 2012 19:35, Richard Ems richard@cape-horn-eng.com wrote:
Hi list,
what happens if there are more reservations being requested by waiting
jobs asked for than the value set in max_reservations?
Do reservations on running jobs still count for the scheduler
reservation plans?
The
BTW, we dual-boot the RHEL kernel the UEK (haven't tried UEK2 yet).
The advantage of this setup is that we could always go back to the
RHEL kernel with just a reboot. Also, we just need to setup OGS/GE (or
any other software packages) once and can test it with both kernels.
Rayson
On Thu, May
On 05/25/2012 11:34 AM, Prentice Bisbal wrote:
Okay, this going to be a stupid question coming from someone who's been
on this list for years, but here goes...
I've just upgraded the RAM on a few cluster nodes to 32 GB (instead of
16 GB). A few users could benefit from this, so I'd like to be
Hi all,
In GE6.2u5 , I defined an ACL with sub ACL entries such:
nameoctopus-users
typeACL
fshare 0
oticket 0
entries @octopus-master
---
nameoctopus-master
typeACL
fshare 0
oticket 0
entries sanki
--
user_lists parameter of octopus.q set to octopus-users . In this case
28 matches
Mail list logo