Sure - there is always an MCA param for everything:
OMPI_MCA_rmaps_base_oversubscribe=1
> On Apr 25, 2017, at 2:10 PM, Eric Chamberland
> wrote:
>
> On 25/04/17 04:36 PM, r...@open-mpi.org wrote:
>> add --oversubscribe to the cmd line
>
> good, it works! :)
>
> Is there an environment varia
If it helps, I believe I added the ability to just use ‘:*’ to indicate “take
them all” so you don’t have to remember the number.
> On Apr 25, 2017, at 2:13 PM, George Bosilca wrote:
>
> Thanks Ralph,
>
> Indeed, if I add :8 I get back the expected behavior. I can cope with this (I
> don't us
Thanks Ralph,
Indeed, if I add :8 I get back the expected behavior. I can cope with this
(I don't usually restrict my runs to a subset of the nodes).
George.
On Tue, Apr 25, 2017 at 4:53 PM, r...@open-mpi.org wrote:
> I suspect it read the file just fine - what you are seeing in the output
On 25/04/17 04:36 PM, r...@open-mpi.org wrote:
add --oversubscribe to the cmd line
good, it works! :)
Is there an environment variable equivalent to --oversubscribe argument?
I can't find this option in near related FAQ entries, should it be added
here? :
https://www.open-mpi.org/faq/?cat
I suspect it read the file just fine - what you are seeing in the output is a
reflection of the community’s design decision that only one slot would be
allocated for each time a node is listed in -host. This is why they added the
:N modifier so you can specify the #slots to use in lieu of writin
Sigh - that _is_ the requested behavior. The -host option defaults to
indicating only one slot should be used on that node.
> On Apr 25, 2017, at 1:32 PM, George Bosilca wrote:
>
> I confirm a similar issue on a more managed environment. I have an hostfile
> that worked for the last few years,
Oh, forgot something important,
since OpenMPI 1.8.x I am using:
export OMPI_MCA_hwloc_base_binding_policy=none
Also, I am exporting this since 1.6.x?:
export OMPI_MCA_mpi_yield_when_idle=1
Eric
On 25/04/17 04:31 PM, Eric Chamberland wrote:
Ok, here it is:
===
first, with -
Okay, so what’s happening is that we are auto-detecting only 4 cores on that
box, and since you didn’t provide any further info, we set the #slots = #cores.
If you want to run more than that, you can either tell us a number of slots to
use (e.g., -host mybox:32) or add --oversubscribe to the cmd
Just to be clear, the hostfile contains the correct info:
dancer00 slots=8
dancer01 slots=8
The output regarding the 2 nodes (dancer00 and dancer01) is clearly wrong.
George.
On Tue, Apr 25, 2017 at 4:32 PM, George Bosilca wrote:
> I confirm a similar issue on a more managed environment.
I confirm a similar issue on a more managed environment. I have an hostfile
that worked for the last few years, and that span across a small cluster
(30 nodes of 8 cores each).
Trying to spawn any number of processes across P nodes fails if the number
of processes is larger than P (despite the fac
Ok, here it is:
===
first, with -n 8:
===
mpirun -mca ras_base_verbose 10 --display-allocation -n 8 echo "Hello"
[zorg:22429] [[INVALID],INVALID] plm:rsh_lookup on agent ssh : rsh path NULL
[zorg:22429] plm:base:set_hnp_name: initial bias 22429 nodename hash
810
Okay - so effectively you have no hostfile, and no allocation. So this is
running just on the one node where mpirun exists?
Add “-mca ras_base_verbose 10 --display-allocation” to your cmd line and let’s
see what it found
> On Apr 25, 2017, at 12:56 PM, Eric Chamberland
> wrote:
>
> Hi,
>
>
Hi,
the host file has been constructed automatically by the
configuration+installation process and seems to contain only comments
and a blank line:
(15:53:50) [zorg]:~> cat /opt/openmpi-3.x_debug/etc/openmpi-default-hostfile
#
# Copyright (c) 2004-2005 The Trustees of Indiana University and I
What is in your hostfile?
> On Apr 25, 2017, at 11:39 AM, Eric Chamberland
> wrote:
>
> Hi,
>
> just testing the 3.x branch... I launch:
>
> mpirun -n 8 echo "hello"
>
> and I get:
>
> --
> There are not enough slots a
Hi,
just testing the 3.x branch... I launch:
mpirun -n 8 echo "hello"
and I get:
--
There are not enough slots available in the system to satisfy the 8 slots
that were requested by the application:
echo
Either request f
15 matches
Mail list logo
