Merge pull request #5605 from garlick/rexec_shell

shell: add rexec plugin

Author: mergify[bot]

doc/Makefile.am

@@ -382,7 +382,6 @@ if ENABLE_DOCS
 man_MANS = $(MAN1_FILES) $(MAN3_FILES) $(MAN5_FILES) $(MAN7_FILES)
 $(RST_FILES): \
 	man1/common/resources.rst \
-	man1/common/nodeset.rst \
 	man1/common/job-param-additional.rst \
 	man1/common/job-param-batch.rst \
 	man1/common/job-param-common.rst \
@@ -461,7 +460,6 @@ EXTRA_DIST = \
 	$(RST_FILES) \
 	man1/index.rst \
 	man1/common/resources.rst \
-	man1/common/nodeset.rst \
 	man1/common/job-param-additional.rst \
 	man1/common/job-param-batch.rst \
 	man1/common/job-param-common.rst \

doc/man1/common/nodeset.rst

@@ -2,66 +2,85 @@
 flux-exec(1)
 ============
 
-
 SYNOPSIS
---------
-**flux** **exec** [--noinput] [*--label-io] [*—dir=DIR'] [*--rank=NODESET*] [*--verbose*] COMMANDS...
+========
+
+**flux** **exec** [*--noinput*] [*--label-io*] [*—dir=DIR*] [*--rank=IDSET*] [*--verbose*] *COMMAND...*
 
 DESCRIPTION
 ===========
 
 .. program:: flux exec
 
-:program:`flux exec` runs commands across one or more Flux broker ranks using
-the *broker.exec* service. The commands are executed as direct children
-of the broker, and the broker handles buffering stdout and stderr and
-sends the output back to :program:`flux exec` which copies output to its own
-stdout and stderr.
-
-On receipt of SIGINT and SIGTERM signals, :program:`flux exec` shall forward
-the received signal to all currently running remote processes.
-
-In the event subprocesses are hanging or ignoring SIGINT, two SIGINT
-signals (typically sent via Ctrl+C) in short succession can force
-:program:`flux exec` to exit.
-
-:program:`flux exec` is meant as an administrative and test utility, and cannot
-be used to launch Flux jobs.
+:program:`flux exec` remotely executes one or more copies of *COMMAND*,
+similar to :linux:man1:`pdsh`.  It bypasses the scheduler and is intended
+for launching administrative commands or tool daemons, not for launching
+parallel jobs.  For that, see :man1:`flux-run`.
 
+By default, *COMMAND* runs across all :man1:`flux-broker` processes.  If the
+:option:`--jobid` option is specified, the commands are run across a job's
+:man1:`flux-shell` processes.  Normally this means that one copy of *COMMAND*
+is executed per node, but in unusual cases it could mean more (e.g. if the
+Flux instance was started with multiple brokers per node).
 
-EXIT STATUS
-===========
-
-In the case that all processes are successfully launched, the exit status
-of :program:`flux exec` is the largest of the remote process exit codes.
-
-If a non-existent rank is targeted, :program:`flux exec` will return with
-code 68 (EX_NOHOST from sysexits.h).
-
-If one or more remote commands are terminated by a signal, then
-:program:`flux exec` exits with exit code 128+signo.
+Standard output and standard error of the remote commands are captured
+and combined on the :program:`flux exec` standard output and standard error.
+Standard input of :program:`flux exec` is captured and broadcast to standard
+input of the remote commands.
 
+On receipt of SIGINT and SIGTERM signals, :program:`flux exec` forwards
+the received signal to the remote processes.  When standard input of
+:program:`flux exec` is a terminal, :kbd:`Control-C` may be used to send
+SIGINT.  Two of those in short succession can force :program:`flux exec`
+to exit in the event that remote processes are hanging.
 
 OPTIONS
 =======
 
 .. option:: -l, --label-io
 
-   Label lines of output with the source RANK.
+   Label lines of output with the source broker RANK.  This option is not
+   affected by :option:`--jobid`.
 
 .. option:: -n, --noinput
 
    Do not attempt to forward stdin. Send EOF to remote process stdin.
 
 .. option:: -d, --dir=DIR
 
-   Set the working directory of remote *COMMANDS* to *DIR*. The default is to
+   Set the working directory of remote *COMMAND* to *DIR*. The default is to
    propagate the current working directory of flux-exec(1).
 
-.. option:: -r, --rank=NODESET
+.. option:: -r, --rank=IDSET
+
+   Target specific ranks, where *IDSET* is a set of zero-origin node ranks in
+   RFC 22 format.  If :option:`--jobid` is specified, the ranks are interpreted
+   as an index into the list of nodes assigned to the job.  Otherwise, they
+   refer to the nodes assigned to the Flux instance.
+
+   The default is to target all ranks.  As a special case, :option:`--rank=all`
+   is accepted and behaves the same as the default.
+
+.. option:: -x, --exclude=IDSET
 
-   Target specific ranks in *NODESET*. Default is to target "all" ranks.
-   See `NODESET FORMAT`_ below for more information.
+   Exclude specific ranks.  *IDSET* is as described in :option:`--rank`.
+
+.. option:: -j, --jobid=JOBID
+
+   Run *COMMAND* on the nodes allocated to *JOBID* instead of the nodes
+   assigned to the Flux instance.
+
+   This uses the exec service embedded in :man1:`flux-shell` rather than
+   :man1:`flux-broker`.
+
+   The interpretation of :option:`--rank` and :option:`--exclude` is adjusted
+   as noted in their descriptions.  For example, :option:`flux exec -j ID -r 0`
+   will run only on the first node assigned to *JOBID*, and
+   :option:`flux exec -j ID -x 0` will run on all nodes assigned to *JOBID*
+   except the first node.
+
+   This option is only available when the job owner is the same as the Flux
+   instance owner.
 
 .. option:: -v, --verbose
 
@@ -73,22 +92,40 @@ OPTIONS
 
 .. option:: --with-imp
 
-   Prepend the full path to :program:`flux-imp run` to *COMMANDS*. This option
+   Prepend the full path to :program:`flux-imp run` to *COMMAND*. This option
    is mostly meant for testing or as a convenience to execute a configured
    ``prolog`` or ``epilog`` command under the IMP. Note: When this option is
    used, or if :program:`flux-imp` is detected as the first argument of
-   *COMMANDS*, :program:`flux exec` will use :program:`flux-imp kill` to
+   *COMMAND*, :program:`flux exec` will use :program:`flux-imp kill` to
    signal remote commands instead of the normal builtin subprocess signaling
    mechanism.
 
+CAVEATS
+=======
+
+In a multi-user flux instance, access to the rank 0 broker execution
+service is restricted to requests that originate from the local broker.
+Therefore, :program:`flux exec` (without :option:`--jobid`) must be run
+from the rank 0 broker if rank 0 is included in the target *IDSET*.
+
+EXIT STATUS
+===========
 
-NODESET FORMAT
-==============
+In the case that all processes are successfully launched, the exit status
+of :program:`flux exec` is the largest of the remote process exit codes.
 
-.. include:: common/nodeset.rst
+If a non-existent rank is targeted, :program:`flux exec` will return with
+code 68 (EX_NOHOST from sysexits.h).
 
+If one or more remote commands are terminated by a signal, then
+:program:`flux exec` exits with exit code 128+signo.
 
 RESOURCES
 =========
 
 .. include:: common/resources.rst
+
+FLUX RFC
+========
+
+:doc:`rfc:spec_22`

doc/man1/flux-exec.rst

@@ -65,7 +65,6 @@ graphviz
 prepended
 multipart
 Tpng
-nodesets
 keepalives
 emerg
 gettimeofday
@@ -117,7 +116,6 @@ ary
 baz
 EPGM
 modopts
-nodeset
 noexec
 pre
 slurm
@@ -831,3 +829,4 @@ unlinks
 VM
 fred
 unmapped
+kbd