squashme add some more faq sections

Signed-off-by: Jeff Squyres <[email protected]>

Author: jsquyres

docs/conf.py

@@ -40,6 +40,7 @@
 
 # Release is a sphinx config variable -- assign it to the computed
 # Open MPI version number.
+series = f"{ompi_data['major']}.{ompi_data['minor']}.x"
 release = f"{ompi_data['major']}.{ompi_data['minor']}.{ompi_data['release']}{ompi_data['greek']}"
 
 
@@ -80,5 +81,7 @@
 rst_prolog = f"""
 .. |mdash|  unicode:: U+02014 .. Em dash
 .. |rarrow| unicode:: U+02192 .. Right arrow
+
 .. |ompi_ver| replace:: {release}
+.. |ompi_series| replace:: {series}
 """

docs/faq/building-open-mpi.rst

@@ -200,6 +200,8 @@ There are two common approaches:
 
 /////////////////////////////////////////////////////////////////////////
 
+.. _faq-building-open-mpi-install-overwrite:
+
 Should I install a new version of Open MPI over an old version?
 ---------------------------------------------------------------
 

docs/faq/index.rst

@@ -24,14 +24,14 @@ that they are worth categorizing in an official way.
    fault-tolerance
    tuning
    ofi
+   troubleshooting
+   ssh
+   slurm
 
 .. TODO:
 
-   troubleshooting
    parallel debugging
-   rsh/ssh
    PBS / Torque
-   Slurm
    SGE
    Large Clusters
 

docs/faq/slurm.rst

@@ -0,0 +1,174 @@
+Running MPI jobs with Slurm
+===========================
+
+.. JMS How can I create a TOC just for this page here at the top?
+
+/////////////////////////////////////////////////////////////////////////
+
+.. _faq-slurm-support-label:
+
+Does Open MPI support running jobs under Slurm?
+-----------------------------------------------
+
+Yes.
+
+Open MPI supports two modes of launching parallel MPI jobs under
+Slurm:
+
+#. Using ``mpirun``.  Open MPI's ``mpirun`` will detect that it is
+   inside of a Slurm job, and will automatically utilize the SLURM
+   infrastructure for launching and controlling the individual MPI
+   processes.
+
+   Using this method, you get the full power and extensive features of
+   Open MPI's ``mpirun`` command (see the ``mpirun(1)`` man page for
+   details).
+
+#. Using ``srun``.  Assuming that Slurm installed its Open MPI plugin,
+   you can use ``srun`` to "direct launch" Open MPI applications
+   without the use of Open MPI's ``mpirun`` command.
+
+   Using direct launch can be *slightly* faster when launching very,
+   very large MPI processes (i.e., thousands or millions of MPI
+   processes in a single job).  But it has significantly fewer
+   features than Open MPI's ``mpirun``.
+
+   .. note:: In versions of Open MPI prior to |ompi_series|, using
+             ``srun`` for direct launch was faster than using
+             ``mpirun``.  **This is no longer true.**
+
+Unless there is a strong reason to use ``srun`` for direct launch, the
+Open MPI team recommends using ``mpirun`` for launching under Slurm jobs.
+
+/////////////////////////////////////////////////////////////////////////
+
+How do I run use ``mpirun`` to launch jobs under Slurm?
+-------------------------------------------------------
+
+Pretty much exactly as you would if you were not in a Slurm job.
+
+For example, you can launch Open MPI's ``mpirun`` in an interactive
+Slurm allocation (via the ``salloc`` command) or you can submit a
+script to Slurm (via the ``sbatch`` command) that includes an
+invocation of the ``mpirun`` command.
+
+Regardless of how ``mpirun`` is invoked, if it detects that it is
+running in a Slurm job, ``mpirun`` automatically obtains both the list
+of hosts and how many processes to start on each host from Slurm
+directly.  Hence, it is unnecessary to specify the ``--hostfile``,
+``--host``, or ``-np`` options to ``mpirun``.  Open MPI will also use
+Slurm-native mechanisms to launch and kill processes --
+``ssh`` is not required.
+
+For example:
+
+.. code-block:: sh
+   :linenos:
+
+   # Allocate a Slurm job with 4 nodes
+   shell$ salloc -N 4
+   # Now run an Open MPI job on all the nodes allocated by Slurm
+   shell$ mpirun my_mpi_application
+
+This will run the 4 MPI processes on the nodes that were allocated by
+Slurm.
+
+Or, if submitting a script:
+
+.. code-block:: sh
+   :linenos:
+
+   shell$ cat my_script.sh
+   #!/bin/sh
+   mpirun my_mpi_application
+   shell$ sbatch -N 4 my_script.sh
+   srun: jobid 1234 submitted
+   shell$
+
+Similar to the ``salloc`` case, no command line options specifing
+number of MPI processes were necessary, since Open MPI will obtain
+that information directly from Slurm at run time.
+
+/////////////////////////////////////////////////////////////////////////
+
+How do I use ``srun`` to directly launch Open MPI applications?
+---------------------------------------------------------------
+
+.. note:: Per :ref:`this FAQ entry <faq-slurm-support-label>`, the
+          Open MPI team generally recommends using ``mpirun`` for
+          launching MPI jobs.
+
+First, you must ensure that Slurm was built and installed with PMI-2
+support.
+
+.. note:: Please ask your friendly neighborhood Slurm developer to
+          support PMIx.  PMIx is the current generation of run-time
+          support API; PMI-2 is the legacy / antiquated API.  Open MPI
+          *only* supports PMI-2 for SLURM.  :-)
+
+Yes, if you have configured OMPI ``--with-pmi=foo``, where ``foo`` is
+the path to the directory where ``pmi2.h`` is located.
+
+.. error:: JMS Ralph -- what else do we need to say here?
+
+Open MPI applications can then be launched directly via the ``srun``
+command.  For example:
+
+.. code-block:: sh
+   :linenos:
+
+   shell$ srun -N 4 my_mpi_application
+
+Or you can use ``sbatch`` with a script:
+
+.. code-block:: sh
+   :linenos:
+
+   shell$ cat my_script.sh
+   #!/bin/sh
+   srun my_mpi_application
+   shell$ sbatch -N 4 my_script.sh
+   srun: jobid 1235 submitted
+   shell$
+
+Similar using ``mpirun`` inside of an ``sbatch`` batch script, no
+``srun`` command line options specifing number of processes were
+necessary, because ``sbatch`` set all the relevant Slurm-level
+parameters about number of processes, cores, partition, etc.
+
+/////////////////////////////////////////////////////////////////////////
+
+I use Slurm on a cluster with the OpenFabrics or UCX network stacks.  Do I need to do anything special?
+-------------------------------------------------------------------------------------------------------
+
+Yes.
+
+You need to ensure that Slurm sets up the locked memory
+limits properly.
+
+.. error:: JMS Need to point to general web pages about setting locked
+           memory limits.
+
+           They used to be at
+           ``category=openfabrics#ib-locked-pages`` and
+           ``category=openfabrics#ib-locked-pages-more``.
+
+           This should probably be in a general networking section --
+           not specific to verbs/openib.
+
+/////////////////////////////////////////////////////////////////////////
+
+My job fails / performs poorly when using mpirun under Slurm 20.11
+------------------------------------------------------------------
+
+There were some changes in Slurm behavior that were introduced in
+Slurm 20.11.0 and subsequently reverted out in Slurm 20.11.3.
+
+SchedMD (the makers of Slurm) strongly suggest that all Open MPI users
+avoid using Slurm versions 20.11.0 through 20.11.2.
+
+Indeed, you will likely run into problems using just about any version
+of Open MPI these problematic Slurm releases.
+
+.. important:: Please either downgrade to an older version or upgrade
+               to a newer version of Slurm.