1
1
openmpi/orte/tools/orte-submit/orte-submit.1in

1431 строка
48 KiB
Plaintext

.\" -*- nroff -*-
.\" Copyright (c) 2009-2014 Cisco Systems, Inc. All rights reserved.
.\" Copyright (c) 2008-2009 Sun Microsystems, Inc. All rights reserved.
.\” Copyright (c) 2015 Intel, Inc. All rights reserved.
.\" $COPYRIGHT$
.\"
.\" Man page for ORTE's orte-submit command
.\"
.\" .TH name section center-footer left-footer center-header
.TH ORTE-SUBMIT 1 "#OMPI_DATE#" "#PACKAGE_VERSION#" "#PACKAGE_NAME#"
.\" **************************
.\" Name Section
.\" **************************
.SH NAME
.
orte-submit, ompi-submit \- Execute serial and parallel jobs in Open MPI using a DVM.
.B Note:
\fIompi-submit\fP and \fIorte-submit\fP are synonyms for each
other. Using either of the names will produce the same behavior.
.
.\" **************************
.\" Synopsis Section
.\" **************************
.SH SYNOPSIS
.
.PP
Single Process Multiple Data (SPMD) Model:
.B ompi-submit
[ options ]
.B <program>
[ <args> ]
.P
Multiple Instruction Multiple Data (MIMD) Model:
.B ompi-submit
[ global_options ]
[ local_options1 ]
.B <program1>
[ <args1> ] :
[ local_options2 ]
.B <program2>
[ <args2> ] :
... :
[ local_optionsN ]
.B <programN>
[ <argsN> ]
.P
Note that in both models, invoking \fIompi-submit\fP via an absolute path
name is equivalent to specifying the \fI--prefix\fP option with a
\fI<dir>\fR value equivalent to the directory where \fIompi-submit\fR
resides, minus its last subdirectory. For example:
\fB%\fP /usr/local/bin/ompi-submit ...
is equivalent to
\fB%\fP ompi-submit --prefix /usr/local
.
.\" **************************
.\" Quick Summary Section
.\" **************************
.SH QUICK SUMMARY
.
.B
Use of \fIorte-submit\fP requires that you first start the Distributed Virtual
Machine (DVM) using \fIorte-dvm\fP.
.P
If you are simply looking for how to run an MPI application, you
probably want to use a command line of the following form:
\fB%\fP ompi-submit [ -np X ] [ --hostfile <filename> ] <program>
This will run X copies of \fI<program>\fR in your current run-time
environment (if running under a supported resource manager, Open MPI's
\fIompi-submit\fR will usually automatically use the corresponding resource manager
process starter, as opposed to, for example, \fIrsh\fR or \fIssh\fR,
which require the use of a hostfile, or will default to running all X
copies on the localhost), scheduling (by default) in a round-robin fashion by
CPU slot. See the rest of this page for more details.
.P
Please note that ompi-submit automatically binds processes as of the start of the
v1.8 series. Two binding patterns are used in the absence of any further directives:
.TP 18
.B Bind to core:
when the number of processes is <= 2
.
.
.TP
.B Bind to socket:
when the number of processes is > 2
.
.
.P
If your application uses threads, then you probably want to ensure that you are
either not bound at all (by specifying --bind-to none), or bound to multiple cores
using an appropriate binding level or specific number of processing elements per
application process.
.
.\" **************************
.\" Options Section
.\" **************************
.SH OPTIONS
.
.I ompi-submit
will send the name of the directory where it was invoked on the local
node to each of the remote nodes, and attempt to change to that
directory. See the "Current Working Directory" section below for further
details.
.\"
.\" Start options listing
.\" Indent 10 characters from start of first column to start of second column
.TP 10
.B <program>
The program executable. This is identified as the first non-recognized argument
to ompi-submit.
.
.
.TP
.B <args>
Pass these run-time arguments to every new process. These must always
be the last arguments to \fIompi-submit\fP. If an app context file is used,
\fI<args>\fP will be ignored.
.
.
.TP
.B -h\fR,\fP --help
Display help for this command
.
.
.TP
.B -q\fR,\fP --quiet
Suppress informative messages from orte-submit during application execution.
.
.
.TP
.B -v\fR,\fP --verbose
Be verbose
.
.
.TP
.B -V\fR,\fP --version
Print version number. If no other arguments are given, this will also
cause orte-submit to exit.
.
.
.
.
.P
Use one of the following options to specify which hosts (nodes) of the DVM to run on.
Specifying hosts outside the DVM will result in an error.
.
.
.TP
.B -H\fR,\fP -host\fR,\fP --host \fR<host1,host2,...,hostN>\fP
List of hosts on which to invoke processes.
.
.
.TP
.B
-hostfile\fR,\fP --hostfile \fR<hostfile>\fP
Provide a hostfile to use.
.\" JJH - Should have man page for how to format a hostfile properly.
.
.
.TP
.B -machinefile\fR,\fP --machinefile \fR<machinefile>\fP
Synonym for \fI-hostfile\fP.
.
.
.
.
.P
The following options specify the number of processes to launch. Note that none
of the options imply a particular binding policy - e.g., requesting N processes
for each socket does not imply that the processes will be bound to the socket.
.
.
.TP
.B -c\fR,\fP -n\fR,\fP --n\fR,\fP -np \fR<#>\fP
Run this many copies of the program on the given nodes. This option
indicates that the specified file is an executable program and not an
application context. If no value is provided for the number of copies to
execute (i.e., neither the "-np" nor its synonyms are provided on the command
line), Open MPI will automatically execute a copy of the program on
each process slot (see below for description of a "process slot"). This
feature, however, can only be used in the SPMD model and will return an
error (without beginning execution of the application) otherwise.
.
.
.TP
.B —map-by ppr:N:<object>
Launch N times the number of objects of the specified type on each node.
.
.
.TP
.B -npersocket\fR,\fP --npersocket <#persocket>
On each node, launch this many processes times the number of processor
sockets on the node.
The \fI-npersocket\fP option also turns on the \fI-bind-to-socket\fP option.
(deprecated in favor of --map-by ppr:n:socket)
.
.
.TP
.B -npernode\fR,\fP --npernode <#pernode>
On each node, launch this many processes.
(deprecated in favor of --map-by ppr:n:node)
.
.
.TP
.B -pernode\fR,\fP --pernode
On each node, launch one process -- equivalent to \fI-npernode\fP 1.
(deprecated in favor of --map-by ppr:1:node)
.
.
.
.
.P
To map processes:
.
.
.TP
.B --map-by <foo>
Map to the specified object, defaults to \fIsocket\fP. Supported options
include slot, hwthread, core, L1cache, L2cache, L3cache, socket, numa,
board, node, sequential, distance, and ppr. Any object can include
modifiers by adding a \fR:\fP and any combination of PE=n (bind n
processing elements to each proc), SPAN (load
balance the processes across the allocation), OVERSUBSCRIBE (allow
more processes on a node than processing elements), and NOOVERSUBSCRIBE.
This includes PPR, where the pattern would be terminated by another colon
to separate it from the modifiers.
.
.TP
.B -bycore\fR,\fP --bycore
Map processes by core (deprecated in favor of --map-by core)
.
.TP
.B -bysocket\fR,\fP --bysocket
Map processes by socket (deprecated in favor of --map-by socket)
.
.TP
.B -nolocal\fR,\fP --nolocal
Do not run any copies of the launched application on the same node as
orte-submit is running. This option will override listing the localhost
with \fB--host\fR or any other host-specifying mechanism.
.
.TP
.B -nooversubscribe\fR,\fP --nooversubscribe
Do not oversubscribe any nodes; error (without starting any processes)
if the requested number of processes would cause oversubscription.
This option implicitly sets "max_slots" equal to the "slots" value for
each node.
.
.TP
.B -bynode\fR,\fP --bynode
Launch processes one per node, cycling by node in a round-robin
fashion. This spreads processes evenly among nodes and assigns
MPI_COMM_WORLD ranks in a round-robin, "by node" manner.
.
.
.
.
.P
To order processes' ranks in MPI_COMM_WORLD:
.
.
.TP
.B --rank-by <foo>
Rank in round-robin fashion according to the specified object,
defaults to \fIslot\fP. Supported options
include slot, hwthread, core, L1cache, L2cache, L3cache,
socket, numa, board, and node.
.
.
.
.
.P
For process binding:
.
.TP
.B --bind-to <foo>
Bind processes to the specified object, defaults to \fIcore\fP. Supported options
include slot, hwthread, core, l1cache, l2cache, l3cache, socket, numa, board, and none.
.
.TP
.B -cpus-per-proc\fR,\fP --cpus-per-proc <#perproc>
Bind each process to the specified number of cpus.
(deprecated in favor of --map-by <obj>:PE=n)
.
.TP
.B -cpus-per-rank\fR,\fP --cpus-per-rank <#perrank>
Alias for \fI-cpus-per-proc\fP.
(deprecated in favor of --map-by <obj>:PE=n)
.
.TP
.B -bind-to-core\fR,\fP --bind-to-core
Bind processes to cores (deprecated in favor of --bind-to core)
.
.TP
.B -bind-to-socket\fR,\fP --bind-to-socket
Bind processes to processor sockets (deprecated in favor of --bind-to socket)
.
.TP
.B -bind-to-none\fR,\fP --bind-to-none
Do not bind processes (deprecated in favor of --bind-to none)
.
.TP
.B -report-bindings\fR,\fP --report-bindings
Report any bindings for launched processes.
.
.TP
.B -slot-list\fR,\fP --slot-list <slots>
List of processor IDs to be used for binding MPI processes. The specified bindings will
be applied to all MPI processes. See explanation below for syntax.
.
.
.
.
.P
For rankfiles:
.
.
.TP
.B -rf\fR,\fP --rankfile <rankfile>
Provide a rankfile file.
.
.
.
.
.P
To manage standard I/O:
.
.
.TP
.B -output-filename\fR,\fP --output-filename \fR<filename>\fP
Redirect the stdout, stderr, and stddiag of all processes to a process-unique version of
the specified filename. Any directories in the filename will automatically be created.
Each output file will consist of filename.id, where the id will be the
processes' rank in MPI_COMM_WORLD, left-filled with
zero's for correct ordering in listings.
.
.
.TP
.B -stdin\fR,\fP --stdin <rank>
The MPI_COMM_WORLD rank of the process that is to receive stdin. The
default is to forward stdin to MPI_COMM_WORLD rank 0, but this option
can be used to forward stdin to any process. It is also acceptable to
specify \fInone\fP, indicating that no processes are to receive stdin.
.
.
.TP
.B -tag-output\fR,\fP --tag-output
Tag each line of output to stdout, stderr, and stddiag with \fB[jobid, MCW_rank]<stdxxx>\fP indicating the process jobid
and MPI_COMM_WORLD rank of the process that generated the output, and the channel which generated it.
.
.
.TP
.B -timestamp-output\fR,\fP --timestamp-output
Timestamp each line of output to stdout, stderr, and stddiag.
.
.
.TP
.B -xml\fR,\fP --xml
Provide all output to stdout, stderr, and stddiag in an xml format.
.
.
.TP
.B -xterm\fR,\fP --xterm \fR<ranks>\fP
Display the output from the processes identified by their
MPI_COMM_WORLD ranks in separate xterm windows. The ranks are specified
as a comma-separated list of ranges, with a -1 indicating all. A separate
window will be created for each specified process.
.B Note:
xterm will normally terminate the window upon termination of the process running
within it. However, by adding a "!" to the end of the list of specified ranks,
the proper options will be provided to ensure that xterm keeps the window open
\fIafter\fP the process terminates, thus allowing you to see the process' output.
Each xterm window will subsequently need to be manually closed.
.B Note:
In some environments, xterm may require that the executable be in the user's
path, or be specified in absolute or relative terms. Thus, it may be necessary
to specify a local executable as "./foo" instead of just "foo". If xterm fails to
find the executable, ompi-submit will hang, but still respond correctly to a ctrl-c.
If this happens, please check that the executable is being specified correctly
and try again.
.
.
.
.
.P
To manage files and runtime environment:
.
.
.TP
.B -path\fR,\fP --path \fR<path>\fP
<path> that will be used when attempting to locate the requested
executables. This is used prior to using the local PATH setting.
.
.
.TP
.B --prefix \fR<dir>\fP
Prefix directory that will be used to set the \fIPATH\fR and
\fILD_LIBRARY_PATH\fR on the remote node before invoking Open MPI or
the target process. See the "Remote Execution" section, below.
.
.
.TP
.B --preload-binary
Copy the specified executable(s) to remote machines prior to starting remote processes. The
executables will be copied to the Open MPI session directory and will be deleted upon
completion of the job.
.
.
.TP
.B --preload-files <files>
Preload the comma separated list of files to the current working directory of the remote
machines where processes will be launched prior to starting those processes.
.
.
.TP
.B --preload-files-dest-dir <path>
The destination directory to be used for preload-files, if other than the current working
directory. By default, the absolute and relative paths provided by --preload-files are used.
.
.
.TP
.B -wd \fR<dir>\fP
Synonym for \fI-wdir\fP.
.
.
.TP
.B -wdir \fR<dir>\fP
Change to the directory <dir> before the user's program executes.
See the "Current Working Directory" section for notes on relative paths.
.B Note:
If the \fI-wdir\fP option appears both on the command line and in an
application context, the context will take precedence over the command
line. Thus, if the path to the desired wdir is different
on the backend nodes, then it must be specified as an absolute path that
is correct for the backend node.
.
.
.TP
.B -x \fR<env>\fP
Export the specified environment variables to the remote nodes before
executing the program. Only one environment variable can be specified
per \fI-x\fP option. Existing environment variables can be specified
or new variable names specified with corresponding values. For
example:
\fB%\fP ompi-submit -x DISPLAY -x OFILE=/tmp/out ...
The parser for the \fI-x\fP option is not very sophisticated; it does
not even understand quoted values. Users are advised to set variables
in the environment, and then use \fI-x\fP to export (not define) them.
.
.
.
.
.P
Setting MCA parameters:
.
.
.TP
.B -gmca\fR,\fP --gmca \fR<key> <value>\fP
Pass global MCA parameters that are applicable to all contexts. \fI<key>\fP is
the parameter name; \fI<value>\fP is the parameter value.
.
.
.TP
.B -mca\fR,\fP --mca <key> <value>
Send arguments to various MCA modules. See the "MCA" section, below.
.
.
.
.
.P
For debugging:
.
.
.TP
.B -debug\fR,\fP --debug
Invoke the user-level debugger indicated by the \fIorte_base_user_debugger\fP
MCA parameter.
.
.
.TP
.B -debugger\fR,\fP --debugger
Sequence of debuggers to search for when \fI--debug\fP is used (i.e.
a synonym for \fIorte_base_user_debugger\fP MCA parameter).
.
.
.TP
.B -tv\fR,\fP --tv
Launch processes under the TotalView debugger.
Deprecated backwards compatibility flag. Synonym for \fI--debug\fP.
.
.
.
.
.P
There are also other options:
.
.
.TP
.B --allow-run-as-root
Allow
.I ompi-submit
to run when executed by the root user
.RI ( ompi-submit
defaults to aborting when launched as the root user).
.
.
.TP
.B -aborted\fR,\fP --aborted \fR<#>\fP
Set the maximum number of aborted processes to display.
.
.
.TP
.B --app \fR<appfile>\fP
Provide an appfile, ignoring all other command line options.
.
.
.TP
.B -cf\fR,\fP --cartofile \fR<cartofile>\fP
Provide a cartography file.
.
.
.TP
.B --hetero
Indicates that multiple app_contexts are being provided that are a mix of 32/64-bit binaries.
.
.
.TP
.B -ompi-server\fR,\fP --ompi-server <uri or file>
Specify the URI of the Open MPI server (or the ompi-submit to be used as the server)
, the name
of the file (specified as file:filename) that
contains that info, or the PID (specified as pid:#) of the ompi-submit to be used as
the server.
The Open MPI server is used to support multi-application data exchange via
the MPI-2 MPI_Publish_name and MPI_Lookup_name functions.
.
.
.
.
.P
The following options are useful for developers; they are not generally
useful to most ORTE and/or MPI users:
.
.TP
.B -d\fR,\fP --debug-devel
Enable debugging of the OmpiRTE (the run-time layer in Open MPI).
This is not generally useful for most users.
.
.
.
.P
There may be other options listed with \fIompi-submit --help\fP.
.
.
.SS Environment Variables
.
.TP
.B MPIEXEC_TIMEOUT
The maximum number of seconds that
.I ompi-submit
.RI ( mpiexec )
will run. After this many seconds,
.I ompi-submit
will abort the launched job and exit.
.
.
.\" **************************
.\" Description Section
.\" **************************
.SH DESCRIPTION
.
One invocation of \fIompi-submit\fP starts an MPI application running under Open
MPI. If the application is single process multiple data (SPMD), the application
can be specified on the \fIompi-submit\fP command line.
If the application is multiple instruction multiple data (MIMD), comprising of
multiple programs, the set of programs and argument can be specified in one of
two ways: Extended Command Line Arguments, and Application Context.
.PP
An application context describes the MIMD program set including all arguments
in a separate file.
.\" See appcontext(5) for a description of the application context syntax.
This file essentially contains multiple \fIompi-submit\fP command lines, less the
command name itself. The ability to specify different options for different
instantiations of a program is another reason to use an application context.
.PP
Extended command line arguments allow for the description of the application
layout on the command line using colons (\fI:\fP) to separate the specification
of programs and arguments. Some options are globally set across all specified
programs (e.g. --hostfile), while others are specific to a single program
(e.g. -np).
.
.
.
.SS Specifying Host Nodes
.
Host nodes can be identified on the \fIompi-submit\fP command line with the \fI-host\fP
option or in a hostfile.
.
.PP
For example,
.
.TP 4
ompi-submit -H aa,aa,bb ./a.out
launches two processes on node aa and one on bb.
.
.PP
Or, consider the hostfile
.
\fB%\fP cat myhostfile
aa slots=2
bb slots=2
cc slots=2
.
.PP
Since the DVM was started with \fIorte-dvm\fP, \fIorte-submit\fP
will ignore any slots arguments in the hostfile. Values provided
via hostfile to \fIorte-dvm\fP will control the behavior.
.
.PP
.
.TP 4
ompi-submit -hostfile myhostfile ./a.out
will launch two processes on each of the three nodes.
.
.TP 4
ompi-submit -hostfile myhostfile -host aa ./a.out
will launch two processes, both on node aa.
.
.TP 4
ompi-submit -hostfile myhostfile -host dd ./a.out
will find no hosts to run on and abort with an error.
That is, the specified host dd is not in the specified hostfile.
.
.SS Specifying Number of Processes
.
As we have just seen, the number of processes to run can be set using the
hostfile. Other mechanisms exist.
.
.PP
The number of processes launched can be specified as a multiple of the
number of nodes or processor sockets available. For example,
.
.TP 4
ompi-submit -H aa,bb -npersocket 2 ./a.out
launches processes 0-3 on node aa and process 4-7 on node bb,
where aa and bb are both dual-socket nodes.
The \fI-npersocket\fP option also turns on the \fI-bind-to-socket\fP option,
which is discussed in a later section.
.
.TP 4
ompi-submit -H aa,bb -npernode 2 ./a.out
launches processes 0-1 on node aa and processes 2-3 on node bb.
.
.TP 4
ompi-submit -H aa,bb -npernode 1 ./a.out
launches one process per host node.
.
.TP 4
ompi-submit -H aa,bb -pernode ./a.out
is the same as \fI-npernode\fP 1.
.
.
.PP
Another alternative is to specify the number of processes with the
\fI-np\fP option. Consider now the hostfile
.
\fB%\fP cat myhostfile
aa slots=4
bb slots=4
cc slots=4
.
.PP
Now,
.
.TP 4
ompi-submit -hostfile myhostfile -np 6 ./a.out
will launch processes 0-3 on node aa and processes 4-5 on node bb. The remaining
slots in the hostfile will not be used since the \fI-np\fP option indicated
that only 6 processes should be launched.
.
.SS Mapping Processes to Nodes: Using Policies
.
The examples above illustrate the default mapping of process processes
to nodes. This mapping can also be controlled with various
\fIompi-submit\fP options that describe mapping policies.
.
.
.PP
Consider the same hostfile as above, again with \fI-np\fP 6:
.
node aa node bb node cc
ompi-submit 0 1 2 3 4 5
ompi-submit --map-by node 0 3 1 4 2 5
ompi-submit -nolocal 0 1 2 3 4 5
.
.PP
The \fI--map-by node\fP option will load balance the processes across
the available nodes, numbering each process in a round-robin fashion.
.
.PP
The \fI-nolocal\fP option prevents any processes from being mapped onto the
local host (in this case node aa). While \fIompi-submit\fP typically consumes
few system resources, \fI-nolocal\fP can be helpful for launching very
large jobs where \fIompi-submit\fP may actually need to use noticeable amounts
of memory and/or processing time.
.
.PP
Just as \fI-np\fP can specify fewer processes than there are slots, it can
also oversubscribe the slots. For example, with the same hostfile:
.
.TP 4
ompi-submit -hostfile myhostfile -np 14 ./a.out
will launch processes 0-3 on node aa, 4-7 on bb, and 8-11 on cc. It will
then add the remaining two processes to whichever nodes it chooses.
.
.PP
One can also specify limits to oversubscription. For example, with the same
hostfile:
.
.TP 4
ompi-submit -hostfile myhostfile -np 14 -nooversubscribe ./a.out
will produce an error since \fI-nooversubscribe\fP prevents oversubscription.
.
.PP
Limits to oversubscription can also be specified in the hostfile itself:
.
% cat myhostfile
aa slots=4 max_slots=4
bb max_slots=4
cc slots=4
.
.PP
The \fImax_slots\fP field specifies such a limit. When it does, the
\fIslots\fP value defaults to the limit. Now:
.
.TP 4
ompi-submit -hostfile myhostfile -np 14 ./a.out
causes the first 12 processes to be launched as before, but the remaining
two processes will be forced onto node cc. The other two nodes are
protected by the hostfile against oversubscription by this job.
.
.PP
Using the \fI--nooversubscribe\fR option can be helpful since Open MPI
currently does not get "max_slots" values from the resource manager.
.
.PP
Of course, \fI-np\fP can also be used with the \fI-H\fP or \fI-host\fP
option. For example,
.
.TP 4
ompi-submit -H aa,bb -np 8 ./a.out
launches 8 processes. Since only two hosts are specified, after the first
two processes are mapped, one to aa and one to bb, the remaining processes
oversubscribe the specified hosts.
.
.PP
And here is a MIMD example:
.
.TP 4
ompi-submit -H aa -np 1 hostname : -H bb,cc -np 2 uptime
will launch process 0 running \fIhostname\fP on node aa and processes 1 and 2
each running \fIuptime\fP on nodes bb and cc, respectively.
.
.SS Mapping, Ranking, and Binding: Oh My!
.
Open MPI employs a three-phase procedure for assigning process locations and
ranks:
.
.TP 10
\fBmapping\fP
Assigns a default location to each process
.
.TP 10
\fBranking\fP
Assigns an MPI_COMM_WORLD rank value to each process
.
.TP 10
\fBbinding\fP
Constrains each process to run on specific processors
.
.PP
The \fImapping\fP step is used to assign a default location to each process
based on the mapper being employed. Mapping by slot, node, and sequentially results
in the assignment of the processes to the node level. In contrast, mapping by object, allows
the mapper to assign the process to an actual object on each node.
.
.PP
\fBNote:\fP the location assigned to the process is independent of where it will be bound - the
assignment is used solely as input to the binding algorithm.
.
.PP
The mapping of process processes to nodes can be defined not just
with general policies but also, if necessary, using arbitrary mappings
that cannot be described by a simple policy. One can use the "sequential
mapper," which reads the hostfile line by line, assigning processes
to nodes in whatever order the hostfile specifies. Use the
\fI-mca rmaps seq\fP option. For example, using the same hostfile
as before:
.
.PP
ompi-submit -hostfile myhostfile -mca rmaps seq ./a.out
.
.PP
will launch three processes, one on each of nodes aa, bb, and cc, respectively.
The slot counts don't matter; one process is launched per line on
whatever node is listed on the line.
.
.PP
Another way to specify arbitrary mappings is with a rankfile, which
gives you detailed control over process binding as well. Rankfiles
are discussed below.
.
.PP
The second phase focuses on the \fIranking\fP of the process within
the job's MPI_COMM_WORLD. Open MPI
separates this from the mapping procedure to allow more flexibility in the
relative placement of MPI processes. This is best illustrated by considering the
following two cases where we used the —map-by ppr:2:socket option:
.
.PP
node aa node bb
rank-by core 0 1 ! 2 3 4 5 ! 6 7
rank-by socket 0 2 ! 1 3 4 6 ! 5 7
rank-by socket:span 0 4 ! 1 5 2 6 ! 3 7
.
.PP
Ranking by core and by slot provide the identical result - a simple
progression of MPI_COMM_WORLD ranks across each node. Ranking by
socket does a round-robin ranking within each node until all processes
have been assigned an MCW rank, and then progresses to the next
node. Adding the \fIspan\fP modifier to the ranking directive causes
the ranking algorithm to treat the entire allocation as a single
entity - thus, the MCW ranks are assigned across all sockets before
circling back around to the beginning.
.
.PP
The \fIbinding\fP phase actually binds each process to a given set of processors. This can
improve performance if the operating system is placing processes
suboptimally. For example, it might oversubscribe some multi-core
processor sockets, leaving other sockets idle; this can lead
processes to contend unnecessarily for common resources. Or, it
might spread processes out too widely; this can be suboptimal if
application performance is sensitive to interprocess communication
costs. Binding can also keep the operating system from migrating
processes excessively, regardless of how optimally those processes
were placed to begin with.
.
.PP
The processors to be used for binding can be identified in terms of
topological groupings - e.g., binding to an l3cache will bind each
process to all processors within the scope of a single L3 cache within
their assigned location. Thus, if a process is assigned by the mapper
to a certain socket, then a \fI—bind-to l3cache\fP directive will
cause the process to be bound to the processors that share a single L3
cache within that socket.
.
.PP
To help balance loads, the binding directive uses a round-robin method when binding to
levels lower than used in the mapper. For example, consider the case where a job is
mapped to the socket level, and then bound to core. Each socket will have multiple cores,
so if multiple processes are mapped to a given socket, the binding algorithm will assign
each process located to a socket to a unique core in a round-robin manner.
.
.PP
Alternatively, processes mapped by l2cache and then bound to socket will simply be bound
to all the processors in the socket where they are located. In this manner, users can
exert detailed control over relative MCW rank location and binding.
.
.PP
Finally, \fI--report-bindings\fP can be used to report bindings.
.
.PP
As an example, consider a node with two processor sockets, each comprising
four cores. We run \fIompi-submit\fP with \fI-np 4 --report-bindings\fP and
the following additional options:
.
% ompi-submit ... --map-by core --bind-to core
[...] ... binding child [...,0] to cpus 0001
[...] ... binding child [...,1] to cpus 0002
[...] ... binding child [...,2] to cpus 0004
[...] ... binding child [...,3] to cpus 0008
% ompi-submit ... --map-by socket --bind-to socket
[...] ... binding child [...,0] to socket 0 cpus 000f
[...] ... binding child [...,1] to socket 1 cpus 00f0
[...] ... binding child [...,2] to socket 0 cpus 000f
[...] ... binding child [...,3] to socket 1 cpus 00f0
% ompi-submit ... --map-by core:PE=2 --bind-to core
[...] ... binding child [...,0] to cpus 0003
[...] ... binding child [...,1] to cpus 000c
[...] ... binding child [...,2] to cpus 0030
[...] ... binding child [...,3] to cpus 00c0
% ompi-submit ... --bind-to none
.
.PP
Here, \fI--report-bindings\fP shows the binding of each process as a mask.
In the first case, the processes bind to successive cores as indicated by
the masks 0001, 0002, 0004, and 0008. In the second case, processes bind
to all cores on successive sockets as indicated by the masks 000f and 00f0.
The processes cycle through the processor sockets in a round-robin fashion
as many times as are needed. In the third case, the masks show us that
2 cores have been bound per process. In the fourth case, binding is
turned off and no bindings are reported.
.
.PP
Open MPI's support for process binding depends on the underlying
operating system. Therefore, certain process binding options may not be available
on every system.
.
.PP
Process binding can also be set with MCA parameters.
Their usage is less convenient than that of \fIompi-submit\fP options.
On the other hand, MCA parameters can be set not only on the \fIompi-submit\fP
command line, but alternatively in a system or user mca-params.conf file
or as environment variables, as described in the MCA section below.
Some examples include:
.
.PP
ompi-submit option MCA parameter key value
--map-by core rmaps_base_mapping_policy core
--map-by socket rmaps_base_mapping_policy socket
--rank-by core rmaps_base_ranking_policy core
--bind-to core hwloc_base_binding_policy core
--bind-to socket hwloc_base_binding_policy socket
--bind-to none hwloc_base_binding_policy none
.
.
.SS Rankfiles
.
Rankfiles are text files that specify detailed information about how
individual processes should be mapped to nodes, and to which
processor(s) they should be bound. Each line of a rankfile specifies
the location of one process (for MPI jobs, the process' "rank" refers
to its rank in MPI_COMM_WORLD). The general form of each line in the
rankfile is:
.
rank <N>=<hostname> slot=<slot list>
.
.PP
For example:
.
$ cat myrankfile
rank 0=aa slot=1:0-2
rank 1=bb slot=0:0,1
rank 2=cc slot=1-2
$ ompi-submit -H aa,bb,cc,dd -rf myrankfile ./a.out
.
.PP
Means that
.
Rank 0 runs on node aa, bound to logical socket 1, cores 0-2.
Rank 1 runs on node bb, bound to logical socket 0, cores 0 and 1.
Rank 2 runs on node cc, bound to logical cores 1 and 2.
.
.PP
Rankfiles can alternatively be used to specify \fIphysical\fP processor
locations. In this case, the syntax is somewhat different. Sockets are
no longer recognized, and the slot number given must be the number of
the physical PU as most OS's do not assign a unique physical identifier
to each core in the node. Thus, a proper physical rankfile looks something
like the following:
.
$ cat myphysicalrankfile
rank 0=aa slot=1
rank 1=bb slot=8
rank 2=cc slot=6
.
.PP
This means that
.
Rank 0 will run on node aa, bound to the core that contains physical PU 1
Rank 1 will run on node bb, bound to the core that contains physical PU 8
Rank 2 will run on node cc, bound to the core that contains physical PU 6
.
.PP
Rankfiles are treated as \fIlogical\fP by default, and the MCA parameter
rmaps_rank_file_physical must be set to 1 to indicate that the rankfile
is to be considered as \fIphysical\fP.
.
.PP
The hostnames listed above are "absolute," meaning that actual
resolveable hostnames are specified. However, hostnames can also be
specified as "relative," meaning that they are specified in relation
to an externally-specified list of hostnames (e.g., by ompi-submit's --host
argument, a hostfile, or a job scheduler).
.
.PP
The "relative" specification is of the form "+n<X>", where X is an
integer specifying the Xth hostname in the set of all available
hostnames, indexed from 0. For example:
.
$ cat myrankfile
rank 0=+n0 slot=1:0-2
rank 1=+n1 slot=0:0,1
rank 2=+n2 slot=1-2
$ ompi-submit -H aa,bb,cc,dd -rf myrankfile ./a.out
.
.PP
Starting with Open MPI v1.7, all socket/core slot locations are be
specified as
.I logical
indexes (the Open MPI v1.6 series used
.I physical
indexes). You can use tools such as HWLOC's "lstopo" to find the
logical indexes of socket and cores.
.
.
.SS Application Context or Executable Program?
.
To distinguish the two different forms, \fIompi-submit\fP
looks on the command line for \fI--app\fP option. If
it is specified, then the file named on the command line is
assumed to be an application context. If it is not
specified, then the file is assumed to be an executable program.
.
.
.
.SS Locating Files
.
If no relative or absolute path is specified for a file, Open
MPI will first look for files by searching the directories specified
by the \fI--path\fP option. If there is no \fI--path\fP option set or
if the file is not found at the \fI--path\fP location, then Open MPI
will search the user's PATH environment variable as defined on the
source node(s).
.PP
If a relative directory is specified, it must be relative to the initial
working directory determined by the specific starter used. For example when
using the rsh or ssh starters, the initial directory is $HOME by default. Other
starters may set the initial directory to the current working directory from
the invocation of \fIompi-submit\fP.
.
.
.
.SS Current Working Directory
.
The \fI\-wdir\fP ompi-submit option (and its synonym, \fI\-wd\fP) allows
the user to change to an arbitrary directory before the program is
invoked. It can also be used in application context files to specify
working directories on specific nodes and/or for specific
applications.
.PP
If the \fI\-wdir\fP option appears both in a context file and on the
command line, the context file directory will override the command
line value.
.PP
If the \fI-wdir\fP option is specified, Open MPI will attempt to
change to the specified directory on all of the remote nodes. If this
fails, \fIompi-submit\fP will abort.
.PP
If the \fI-wdir\fP option is \fBnot\fP specified, Open MPI will send
the directory name where \fIompi-submit\fP was invoked to each of the
remote nodes. The remote nodes will try to change to that
directory. If they are unable (e.g., if the directory does not exist on
that node), then Open MPI will use the default directory determined by
the starter.
.PP
All directory changing occurs before the user's program is invoked; it
does not wait until \fIMPI_INIT\fP is called.
.
.
.
.SS Standard I/O
.
Open MPI directs UNIX standard input to /dev/null on all processes
except the MPI_COMM_WORLD rank 0 process. The MPI_COMM_WORLD rank 0 process
inherits standard input from \fIompi-submit\fP.
.B Note:
The node that invoked \fIompi-submit\fP need not be the same as the node where the
MPI_COMM_WORLD rank 0 process resides. Open MPI handles the redirection of
\fIompi-submit\fP's standard input to the rank 0 process.
.PP
Open MPI directs UNIX standard output and error from remote nodes to the node
that invoked \fIompi-submit\fP and prints it on the standard output/error of
\fIompi-submit\fP.
Local processes inherit the standard output/error of \fIompi-submit\fP and transfer
to it directly.
.PP
Thus it is possible to redirect standard I/O for Open MPI applications by
using the typical shell redirection procedure on \fIompi-submit\fP.
\fB%\fP ompi-submit -np 2 my_app < my_input > my_output
Note that in this example \fIonly\fP the MPI_COMM_WORLD rank 0 process will
receive the stream from \fImy_input\fP on stdin. The stdin on all the other
nodes will be tied to /dev/null. However, the stdout from all nodes will
be collected into the \fImy_output\fP file.
.
.
.
.SS Signal Propagation
.
When orte-submit receives a SIGTERM and SIGINT, it will attempt to kill
the entire job by sending all processes in the job a SIGTERM, waiting
a small number of seconds, then sending all processes in the job a
SIGKILL.
.
.PP
SIGUSR1 and SIGUSR2 signals received by orte-submit are propagated to
all processes in the job.
.
.PP
One can turn on forwarding of SIGSTOP and SIGCONT to the program executed
by ompi-submit by setting the MCA parameter orte_forward_job_control to 1.
A SIGTSTOP signal to ompi-submit will then cause a SIGSTOP signal to be sent
to all of the programs started by ompi-submit and likewise a SIGCONT signal
to ompi-submit will cause a SIGCONT sent.
.
.PP
Other signals are not currently propagated
by orte-submit.
.
.
.SS Process Termination / Signal Handling
.
During the run of an MPI application, if any process dies abnormally
(either exiting before invoking \fIMPI_FINALIZE\fP, or dying as the result of a
signal), \fIompi-submit\fP will print out an error message and kill the rest of the
MPI application.
.PP
User signal handlers should probably avoid trying to cleanup MPI state
(Open MPI is currently not async-signal-safe; see MPI_Init_thread(3)
for details about
.I MPI_THREAD_MULTIPLE
and thread safety). For example, if a segmentation fault occurs in
\fIMPI_SEND\fP (perhaps because a bad buffer was passed in) and a user
signal handler is invoked, if this user handler attempts to invoke
\fIMPI_FINALIZE\fP, Bad Things could happen since Open MPI was already
"in" MPI when the error occurred. Since \fIompi-submit\fP will notice that
the process died due to a signal, it is probably not necessary (and
safest) for the user to only clean up non-MPI state.
.
.
.
.SS Process Environment
.
Processes in the MPI application inherit their environment from the
Open RTE daemon upon the node on which they are running. The
environment is typically inherited from the user's shell. On remote
nodes, the exact environment is determined by the boot MCA module
used. The \fIrsh\fR launch module, for example, uses either
\fIrsh\fR/\fIssh\fR to launch the Open RTE daemon on remote nodes, and
typically executes one or more of the user's shell-setup files before
launching the Open RTE daemon. When running dynamically linked
applications which require the \fILD_LIBRARY_PATH\fR environment
variable to be set, care must be taken to ensure that it is correctly
set when booting Open MPI.
.PP
See the "Remote Execution" section for more details.
.
.
.SS Remote Execution
.
Open MPI requires that the \fIPATH\fR environment variable be set to
find executables on remote nodes (this is typically only necessary in
\fIrsh\fR- or \fIssh\fR-based environments -- batch/scheduled
environments typically copy the current environment to the execution
of remote jobs, so if the current environment has \fIPATH\fR and/or
\fILD_LIBRARY_PATH\fR set properly, the remote nodes will also have it
set properly). If Open MPI was compiled with shared library support,
it may also be necessary to have the \fILD_LIBRARY_PATH\fR environment
variable set on remote nodes as well (especially to find the shared
libraries required to run user MPI applications).
.PP
However, it is not always desirable or possible to edit shell
startup files to set \fIPATH\fR and/or \fILD_LIBRARY_PATH\fR. The
\fI--prefix\fR option is provided for some simple configurations where
this is not possible.
.PP
The \fI--prefix\fR option takes a single argument: the base directory
on the remote node where Open MPI is installed. Open MPI will use
this directory to set the remote \fIPATH\fR and \fILD_LIBRARY_PATH\fR
before executing any Open MPI or user applications. This allows
running Open MPI jobs without having pre-configured the \fIPATH\fR and
\fILD_LIBRARY_PATH\fR on the remote nodes.
.PP
Open MPI adds the basename of the current
node's "bindir" (the directory where Open MPI's executables are
installed) to the prefix and uses that to set the \fIPATH\fR on the
remote node. Similarly, Open MPI adds the basename of the current
node's "libdir" (the directory where Open MPI's libraries are
installed) to the prefix and uses that to set the
\fILD_LIBRARY_PATH\fR on the remote node. For example:
.TP 15
Local bindir:
/local/node/directory/bin
.TP
Local libdir:
/local/node/directory/lib64
.PP
If the following command line is used:
\fB%\fP ompi-submit --prefix /remote/node/directory
Open MPI will add "/remote/node/directory/bin" to the \fIPATH\fR
and "/remote/node/directory/lib64" to the \fLD_LIBRARY_PATH\fR on the
remote node before attempting to execute anything.
.PP
The \fI--prefix\fR option is not sufficient if the installation paths
on the remote node are different than the local node (e.g., if "/lib"
is used on the local node, but "/lib64" is used on the remote node),
or if the installation paths are something other than a subdirectory
under a common prefix.
.PP
Note that executing \fIompi-submit\fR via an absolute pathname is
equivalent to specifying \fI--prefix\fR without the last subdirectory
in the absolute pathname to \fIompi-submit\fR. For example:
\fB%\fP /usr/local/bin/ompi-submit ...
is equivalent to
\fB%\fP ompi-submit --prefix /usr/local
.
.
.
.SS Exported Environment Variables
.
All environment variables that are named in the form OMPI_* will automatically
be exported to new processes on the local and remote nodes. Environmental
parameters can also be set/forwarded to the new processes using the MCA
parameter \fImca_base_env_list\fP. The \fI\-x\fP option to \fIompi-submit\fP has
been deprecated, but the syntax of the MCA param follows that prior
example. While the syntax of the \fI\-x\fP option and MCA param
allows the definition of new variables, note that the parser
for these options are currently not very sophisticated - it does not even
understand quoted values. Users are advised to set variables in the
environment and use the option to export them; not to define them.
.
.
.
.SS Setting MCA Parameters
.
The \fI-mca\fP switch allows the passing of parameters to various MCA
(Modular Component Architecture) modules.
.\" Open MPI's MCA modules are described in detail in ompimca(7).
MCA modules have direct impact on MPI programs because they allow tunable
parameters to be set at run time (such as which BTL communication device driver
to use, what parameters to pass to that BTL, etc.).
.PP
The \fI-mca\fP switch takes two arguments: \fI<key>\fP and \fI<value>\fP.
The \fI<key>\fP argument generally specifies which MCA module will receive the value.
For example, the \fI<key>\fP "btl" is used to select which BTL to be used for
transporting MPI messages. The \fI<value>\fP argument is the value that is
passed.
For example:
.
.TP 4
ompi-submit -mca btl tcp,self -np 1 foo
Tells Open MPI to use the "tcp" and "self" BTLs, and to run a single copy of
"foo" an allocated node.
.
.TP
ompi-submit -mca btl self -np 1 foo
Tells Open MPI to use the "self" BTL, and to run a single copy of "foo" an
allocated node.
.\" And so on. Open MPI's BTL MCA modules are described in ompimca_btl(7).
.PP
The \fI-mca\fP switch can be used multiple times to specify different
\fI<key>\fP and/or \fI<value>\fP arguments. If the same \fI<key>\fP is
specified more than once, the \fI<value>\fPs are concatenated with a comma
(",") separating them.
.PP
Note that the \fI-mca\fP switch is simply a shortcut for setting environment variables.
The same effect may be accomplished by setting corresponding environment
variables before running \fIompi-submit\fP.
The form of the environment variables that Open MPI sets is:
OMPI_MCA_<key>=<value>
.PP
Thus, the \fI-mca\fP switch overrides any previously set environment
variables. The \fI-mca\fP settings similarly override MCA parameters set
in the
$OPAL_PREFIX/etc/openmpi-mca-params.conf or $HOME/.openmpi/mca-params.conf
file.
.
.PP
Unknown \fI<key>\fP arguments are still set as
environment variable -- they are not checked (by \fIompi-submit\fP) for correctness.
Illegal or incorrect \fI<value>\fP arguments may or may not be reported -- it
depends on the specific MCA module.
.PP
To find the available component types under the MCA architecture, or to find the
available parameters for a specific component, use the \fIompi_info\fP command.
See the \fIompi_info(1)\fP man page for detailed information on the command.
.
.SS Running as root
.
The Open MPI team strongly advises against executing
.I ompi-submit
as the root user. MPI applications should be run as regular
(non-root) users.
.
.PP
Reflecting this advice, ompi-submit will refuse to run as root by default.
To override this default, you can add the
.I --allow-run-as-root
option to the
.I ompi-submit
command line.
.
.SS Exit status
.
There is no standard definition for what \fIompi-submit\fP should return as an exit
status. After considerable discussion, we settled on the following method for
assigning the \fIompi-submit\fP exit status (note: in the following description,
the "primary" job is the initial application started by ompi-submit - all jobs that
are spawned by that job are designated "secondary" jobs):
.
.IP \[bu] 2
if all processes in the primary job normally terminate with exit status 0, we return 0
.IP \[bu]
if one or more processes in the primary job normally terminate with non-zero exit status,
we return the exit status of the process with the lowest MPI_COMM_WORLD rank to have a non-zero status
.IP \[bu]
if all processes in the primary job normally terminate with exit status 0, and one or more
processes in a secondary job normally terminate with non-zero exit status, we (a) return
the exit status of the process with the lowest MPI_COMM_WORLD rank in the lowest jobid to have a non-zero status, and (b)
output a message summarizing the exit status of the primary and all secondary jobs.
.IP \[bu]
if the cmd line option --report-child-jobs-separately is set, we will return -only- the
exit status of the primary job. Any non-zero exit status in secondary jobs will be
reported solely in a summary print statement.
.
.PP
By default, OMPI records and notes that MPI processes exited with non-zero termination status.
This is generally not considered an "abnormal termination" - i.e., OMPI will not abort an MPI
job if one or more processes return a non-zero status. Instead, the default behavior simply
reports the number of processes terminating with non-zero status upon completion of the job.
.PP
However, in some cases it can be desirable to have the job abort when any process terminates
with non-zero status. For example, a non-MPI job might detect a bad result from a calculation
and want to abort, but doesn't want to generate a core file. Or an MPI job might continue past
a call to MPI_Finalize, but indicate that all processes should abort due to some post-MPI result.
.PP
It is not anticipated that this situation will occur frequently. However, in the interest of
serving the broader community, OMPI now has a means for allowing users to direct that jobs be
aborted upon any process exiting with non-zero status. Setting the MCA parameter
"orte_abort_on_non_zero_status" to 1 will cause OMPI to abort all processes once any process
exits with non-zero status.
.PP
Terminations caused in this manner will be reported on the console as an "abnormal termination",
with the first process to so exit identified along with its exit status.
.PP
.
.\" **************************
.\" Examples Section
.\" **************************
.SH EXAMPLES
Be sure also to see the examples throughout the sections above.
.
.TP 4
ompi-submit -np 4 -mca btl ib,tcp,self prog1
Run 4 copies of prog1 using the "ib", "tcp", and "self" BTL's for the
transport of MPI messages.
.
.
.TP 4
ompi-submit -np 4 -mca btl tcp,sm,self
.br
--mca btl_tcp_if_include eth0 prog1
.br
Run 4 copies of prog1 using the "tcp", "sm" and "self" BTLs for the
transport of MPI messages, with TCP using only the eth0 interface to
communicate. Note that other BTLs have similar if_include MCA
parameters.
.
.\" **************************
.\" Diagnostics Section
.\" **************************
.
.\" .SH DIAGNOSTICS
.\" .TP 4
.\" Error Msg:
.\" Description
.
.\" **************************
.\" Return Value Section
.\" **************************
.
.SH RETURN VALUE
.
\fIompi-submit\fP returns 0 if all processes started by \fIompi-submit\fP exit after calling
MPI_FINALIZE. A non-zero value is returned if an internal error occurred in
ompi-submit, or one or more processes exited before calling MPI_FINALIZE. If an
internal error occurred in ompi-submit, the corresponding error code is returned.
In the event that one or more processes exit before calling MPI_FINALIZE, the
return value of the MPI_COMM_WORLD rank of the process that \fIompi-submit\fP first notices died
before calling MPI_FINALIZE will be returned. Note that, in general, this will
be the first process that died but is not guaranteed to be so.
.
.\" **************************
.\" See Also Section
.\" **************************
.
.SH SEE ALSO
MPI_Init_thread(3)