It turns that there is an incompatibility between the Cray PMI
library and the default configuration for building Open MPI (master).
To work around this, we now disable use of aprun for direct launch
of Open MPI jobs except under specific conditions.
The problem is that there are now (on master) packages getting
initialized that do not work properly across a fork operation.
As part of a constructor in the Cray PMI library, a fork operation
is done to simplify use of shared memory between the
processes in a job on the same node. This ends up thoroughly
messing up the Open MPI initialization process in the case
that dlopen support is enabled. The initialization process gets
about half-way through when the PMIX framework is opened and
components are loaded, which triggers the Cray PMI constructor
and hence the fork operation.
There are two workarounds for this:
1) configure Open MPI for Cray XE/XC systems using aprun with the
--disable-dlopen option
2) set the PMI_NO_FORK environment variable in the shell in which
the aprun command is run.
Without taking these measures, a Open MPI job will just hang at
job startup in the first attempt to "thread-shift" the PMIx
fence_nb operation. Additional hangs occur at shutdown if this
problem is worked around, again due to the insertion of a fork
operation halfway through the Open MPI initialization procedure.
This commit detects if the conditions that bring out the hang
situation are present, and if so, prints out a message and
aborts the job launch.
Note on systems using slurm, the PMI_NO_FORK environment variable
is set as part of the srun job launch, hence this issue is avoided
on those systems.
Signed-off-by: Howard Pritchard <howardp@lanl.gov>
Bring Slurm PMI-1 component online
Bring the s2 component online
Little cleanup - let the various PMIx modules set the process name during init, and then just raise it up to the ORTE level. Required as the different PMI environments all pass the jobid in different ways.
Bring the OMPI pubsub/pmi component online
Get comm_spawn working again
Ensure we always provide a cpuset, even if it is NULL
pmix/cray: adjust cray pmix component for pmix
Make changes so cray pmix can work within the integrated
ompi/pmix framework.
Bring singletons back online. Implement the comm_spawn operation using pmix - not tested yet
Cleanup comm_spawn - procs now starting, error in connect_accept
Complete integration
Remove use of the Cray PMI KVS - which is designed for a lighweight
MPI that exchanges only a minimimal amount of connection info
(about 128 bytes per rank) - within cray/pmix. Use Cray PMI
collective extensions instead.
This is the first of several steps to accelerate launch of
Open MPI on Cray systems using either native aprun or nativized
slurm.
Use a more reliable way to tell if a process is
1) in a Cray PAGG
2) is actually considered an application process on
a compute node (not for example, a process in a PAGG
on a mom node).
The ess pmi module was not handling aprun launched
daemons. All daemons were thinking they were vpid 1.
Also, turns out that on cray systems using MOM nodes
for launched jobs, just detecting whether or not a
process is in a PAGG container is not sufficient.
Crank up the priority of the alps PLM component in the
event that the configure detected the presence of both
slurm and alps.
Have the ESS pmi component open the pmix framework and
select a pmix component.
This commit was SVN r32773.
WHAT: Merge the PMIx branch into the devel repo, creating a new
OPAL “lmix” framework to abstract PMI support for all RTEs.
Replace the ORTE daemon-level collectives with a new PMIx
server and update the ORTE grpcomm framework to support
server-to-server collectives
WHY: We’ve had problems dealing with variations in PMI implementations,
and need to extend the existing PMI definitions to meet exascale
requirements.
WHEN: Mon, Aug 25
WHERE: https://github.com/rhc54/ompi-svn-mirror.git
Several community members have been working on a refactoring of the current PMI support within OMPI. Although the APIs are common, Slurm and Cray implement a different range of capabilities, and package them differently. For example, Cray provides an integrated PMI-1/2 library, while Slurm separates the two and requires the user to specify the one to be used at runtime. In addition, several bugs in the Slurm implementations have caused problems requiring extra coding.
All this has led to a slew of #if’s in the PMI code and bugs when the corner-case logic for one implementation accidentally traps the other. Extending this support to other implementations would have increased this complexity to an unacceptable level.
Accordingly, we have:
* created a new OPAL “pmix” framework to abstract the PMI support, with separate components for Cray, Slurm PMI-1, and Slurm PMI-2 implementations.
* Replaced the current ORTE grpcomm daemon-based collective operation with an integrated PMIx server, and updated the grpcomm APIs to provide more flexible, multi-algorithm support for collective operations. At this time, only the xcast and allgather operations are supported.
* Replaced the current global collective id with a signature based on the names of the participating procs. The allows an unlimited number of collectives to be executed by any group of processes, subject to the requirement that only one collective can be active at a time for a unique combination of procs. Note that a proc can be involved in any number of simultaneous collectives - it is the specific combination of procs that is subject to the constraint
* removed the prior OMPI/OPAL modex code
* added new macros for executing modex send/recv to simplify use of the new APIs. The send macros allow the caller to specify whether or not the BTL supports async modex operations - if so, then the non-blocking “fence” operation is used, if the active PMIx component supports it. Otherwise, the default is a full blocking modex exchange as we currently perform.
* retained the current flag that directs us to use a blocking fence operation, but only to retrieve data upon demand
This commit was SVN r32570.
