<!--#include virtual="header.txt"-->

<h1><a name="top">Job Submit Plugin API</a></h1>

<h2 id="overview">Overview<a class="slurm_link" href="#overview"></a></h2>
<p> This document describes Slurm job submit plugins and the API that
defines them. It is intended as a resource to programmers wishing to write
their own Slurm job submit plugins. This is version 100 of the API.</p>

<p>Slurm job submit plugins must conform to the
Slurm Plugin API with the following specifications:</p>

<p><span class="commandline">const char
plugin_name[]="<i>full&nbsp;text&nbsp;name</i>"</span>
<p style="margin-left:.2in">
A free-formatted ASCII text string that identifies the plugin.

<p><span class="commandline">const char
plugin_type[]="<i>major/minor</i>"</span><br>
<p style="margin-left:.2in">
The major type must be &quot;job_submit.&quot;
The minor type can be any suitable name for the type of job submission package.
We include samples in the Slurm distribution for
<ul>
<li><b>all_partitions</b> &mdash; Set default partition to all partitions on
the cluster.</li>
<li><b>defaults</b> &mdash; Set default values for job submission or modify
requests.</li>
<li><b>logging</b> &mdash; Log select job submission and modification
parameters.</li>
<li><b>lua</b> &mdash; Interface to <a href="http://www.lua.org">Lua</a> scripts
implementing these functions (actually a slight variation of them). Sample Lua
scripts can be found with the Slurm distribution in the directory
<i>contribs/lua</i>. The Lua script must be named "job_submit.lua" and must
be located in the default configuration directory (typically the subdirectory
"etc" of the installation directory). Slurmctld will fatal on startup if the
configured lua script is invalid. Slurm will try to load the script for each
job submission. If the script is broken or removed while slurmctld is running,
Slurm will fallback to the previous working version of the script.
<b>Warning</b>: slurmctld runs this script while holding internal locks, and
only a single copy of this script can run at a time. This blocks most
concurrency in slurmctld. Therefore, this script should run to completion as
quickly as possible.</li>
<li><b>partition</b> &mdash; Sets a job's default partition based upon job
submission parameters and available partitions.</li>
<li><b>pbs</b> &mdash; Translate PBS job submission options to Slurm equivalent
(if possible).</li>
<li><b>require_timelimit</b> &mdash; Force job submissions to specify a
timelimit.</li>
</ul>

<p><span class="commandline">const uint32_t plugin_version</span><br>
If specified, identifies the version of Slurm used to build this plugin and
any attempt to load the plugin from a different version of Slurm will result
in an error.
If not specified, then the plugin may be loaded by Slurm commands and
daemons from any version, however this may result in difficult to diagnose
failures due to changes in the arguments to plugin functions or changes
in other Slurm functions used by the plugin.</p>

<p>Slurm can be configured to use multiple job_submit plugins if desired,
however the lua plugin will only execute one lua script named "job_submit.lua"
located in the default script directory (typically the subdirectory "etc" of
the installation directory).</p>


<h2 id="api">API Functions<a class="slurm_link" href="#api"></a></h2>
<p>All of the following functions are required. Functions which are not
implemented must be stubbed.

<p class="commandline"> int init (void)
<p style="margin-left:.2in"><b>Description</b>:<br>
  Called when the plugin is loaded, before any other functions are
  called. Put global initialization here.
<p style="margin-left:.2in"><b>Returns</b>: <br>
  <span class="commandline">SLURM_SUCCESS</span> on success, or<br>
  <span class="commandline">SLURM_ERROR</span> on failure.</p>

<p class="commandline"> void fini (void)
<p style="margin-left:.2in"><b>Description</b>:<br>
  Called when the plugin is removed. Clear any allocated storage here.
<p style="margin-left:.2in"><b>Returns</b>: None.</p>

<p><b>Note</b>: These init and fini functions are not the same as those
described in the <span class="commandline">dlopen (3)</span> system library.
The C run-time system co-opts those symbols for its own initialization.
The system <span class="commandline">_init()</span> is called before the Slurm
<span class="commandline">init()</span>, and the Slurm
<span class="commandline">fini()</span> is called before the system's
<span class="commandline">_fini()</span>.</p>

<p class="commandline">
int job_submit(struct job_descriptor *job_desc, uint32_t submit_uid, char **error_msg)
<p style="margin-left:.2in"><b>Description</b>:<br>
This function is called by the slurmctld daemon with the job submission
parameters supplied by the user regardless of the command used (e.g.
salloc, sbatch, slurmrestd). Only explicitly
defined values will be represented. For values not defined at submit time
<span class="commandline">slurm.NO_VAL/16/64</span> or
<span class="commandline">nil</span> will be set. It can be used to log and/or
modify the job parameters supplied by the user as desired. Note that this
function has access to the slurmctld's global data structures, for example
to examine the available partitions, reservations, etc.
<p style="margin-left:.2in"><b>Arguments</b>: <br>
<span class="commandline">job_desc</span>
(input/output) the job allocation request specifications, before job defaults
are set.<br>
<span class="commandline">submit_uid</span>
(input) user ID initiating the request.<br>
<span class="commandline">error_msg</span>
(output) If the argument is not null, then a plugin generated error message
can be stored here. The error message is expected to have allocated memory
which Slurm will release using the xfree function. The error message is always
propagated to the caller, no matter the return code.<br>
<p style="margin-left:.2in"><b>Returns</b>: <br>
<span class="commandline">SLURM_SUCCESS</span> on success, or<br>
<span class="commandline">SLURM_ERROR</span> on failure.

<p class="commandline">
int job_modify(struct job_descriptor *job_desc, job_record_t *job_ptr, uint32_t modify_uid)
<p style="margin-left:.2in"><b>Description</b>:<br>
This function is called by the slurmctld daemon with job modification parameters
supplied by the user regardless of the command used (e.g. scontrol, sview,
slurmrestd). It can be used to log and/or
modify the job parameters supplied by the user as desired. Note that this
function has access to the slurmctld's global data structures, for example to
examine the available partitions, reservations, etc.
<p style="margin-left:.2in"><b>Arguments</b>: <br>
<span class="commandline">job_desc</span>
(input/output) the job allocation request specifications, before job defaults
are set.<br>
<span class="commandline">job_ptr</span>
(input/output) slurmctld daemon's current data structure for the job to
be modified.<br>
<span class="commandline">modify_uid</span>
(input) user ID initiating the request.<br>
<p style="margin-left:.2in"><b>Returns</b>: <br>
<span class="commandline">SLURM_SUCCESS</span> on success, or<br>
<span class="commandline">SLURM_ERROR</span> on failure.

<h2 id="lua">Lua Functions<a class="slurm_link" href="#lua"></a></h2>
<p>The Lua functions differ slightly from those implemented in C for
better ease of use. Sample Lua scripts can be found with the Slurm distribution
in the directory <i>contribs/lua</i>. The default installation location of
the Lua scripts is the same location as the Slurm configuration file,
<i>slurm.conf</i>.
Reading and writing of job environment variables using Lua is possible
by referencing the environment variables as a data structure containing
named elements.</p>
<p><b>NOTE</b>: Only sbatch sends the environment to slurmctld. salloc and srun
do not send the environment to slurmctld, so <i>job_desc.environment</i> is not
available in the job_submit plugin for these jobs.</p>
<p>For example:</p>
<pre>
...
	-- job_desc.environment is only available for batch jobs.
	if (job_desc.script) then
		if (job_desc.environment ~= nil) then
			if (job_desc.environment["FOO"] ~= nil) then
				slurm.log_user("Found env FOO=%s",
					       job_desc.environment["FOO"])
			end
		end
	end
...
</pre>
<p><b>NOTE</b>: To get/set the environment for all types of jobs, an alternate
approach is to use <a href="cli_filter_plugins.html">CliFilterPlugins</a>.</p>

<p class="commandline">
int slurm_job_submit(job_desc_msg_t *job_desc, List part_list, uint32_t
submit_uid)
<p style="margin-left:.2in"><b>Description</b>:<br>
This function is called by the slurmctld daemon with the job submission
parameters supplied by the user regardless of the command used (e.g.
salloc, sbatch, slurmrestd). Only explicitly
defined values will be represented. For values not defined at submit time
<span class="commandline">slurm.NO_VAL/16/64</span> or
<span class="commandline">nil</span> will be set. It can be used to log and/or
modify the job parameters supplied by the user as desired. Note that this
function has access to the slurmctld's global data structures, for example
to examine the available partitions, reservations, etc.
<p style="margin-left:.2in"><b>Arguments</b>: <br>
<span class="commandline">job_desc</span>
(input/output) the job allocation request specifications.<br>
<span class="commandline">part_list</span>
(input) List of pointer to partitions which this user is authorized to use.<br>
<span class="commandline">submit_uid</span>
(input) user ID initiating the request.<br>
<a name="job_modify_returns"></a><p style="margin-left:.2in"><b>Returns</b>: <br>
<span class="commandline">slurm.SUCCESS</span> &mdash;
Job submission accepted by plugin.<br>
<span class="commandline">slurm.FAILURE</span> &mdash;
Job submission rejected due to error (Deprecated in 19.05).<br>
<span class="commandline">slurm.ERROR</span> &mdash;
Job submission rejected due to error.<br>
<span class="commandline">slurm.ESLURM_*</span> &mdash;
Job submission rejected due to error as defined by
<i>slurm/slurm_errno.h</i> and <i>src/common/slurm_errno.c</i>.<br>

<p><b>NOTE</b>: As <span class="commandline">job_desc</span> contains only
user-specified values, undefined values can be recognized (before defaults
are set) by either checking for <span class="commandline">nil</span> or for
the corresponding <span class="commandline">slurm.NO_VAL/16/64</span>. This
allows sites to apply policies, such as requiring users to define the number
of nodes, as in the example below:</p>
<pre>
...
	-- Number of nodes must be defined at submit time
	if (job_desc.max_nodes == slurm.NO_VAL) then
		slurm.log_user("No max_nodes specified, please specify a number of nodes")
		return slurm.ERROR
	end
...
</pre>
<p class="commandline">
int slurm_job_modify(job_desc_msg_t *job_desc, job_record_t *job_ptr,
List part_list, int modify_uid)
<p style="margin-left:.2in"><b>Description</b>:<br>
This function is called by the slurmctld daemon with job modification parameters
supplied by the user regardless of the command used (e.g. scontrol, sview,
slurmrestd). It can be used to log and/or
modify the job parameters supplied by the user as desired. Note that this
function has access to the slurmctld's global data structures, for example to
examine the available partitions, reservations, etc.
<p style="margin-left:.2in"><b>Arguments</b>: <br>
<span class="commandline">job_desc</span>
(input/output) the job allocation request specifications.<br>
<span class="commandline">job_ptr</span>
(input/output) slurmctld daemon's current data structure for the job to
be modified.<br>
<span class="commandline">part_list</span>
(input) List of pointer to partitions which this user is authorized to use.<br>
<span class="commandline">modify_uid</span>
(input) user ID initiating the request.<br>
<p style="margin-left:.2in"><b>Returns</b>: <br>
<a href="#job_modify_returns">Returns from job_modify() are the same as the returns from job_submit().</a>

<h2 id="attr">Lua Job Attributes<a class="slurm_link" href="#attr"></a></h2>
<p>The available job attributes change occasionally with different versions of
Slurm. To find the job attributes that are available for the version of Slurm
you're using, go to the <a href="https://github.com/SchedMD/slurm"> SchedMD
github page</a>. Navigate to <b>src/lua/slurm_lua.c</b> and look for function
<b>slurm_lua_job_record_field()</b>, which contains the list
of attributes available for the job_record (e.g. current record in Slurm).
Navigate to <b>src/plugins/job_submit/lua/job_submit_lua.c</b> and look for
function <b>_get_job_req_field()</b>, which contains the list of attributes
available for the job_descriptor (e.g. submission or modification request).
</p>

<h2 id="building">Building<a class="slurm_link" href="#building"></a></h2>
<p>Generally using a LUA interface for a job submit plugin is best:
It is simple to write and maintain with minimal dependencies upon the Slurm
data structures.
However using C does provide a mechanism to get more information than available
using LUA including full access to all of the data structures and functions
in the slurmctld daemon.
The simplest way to build a C program would be to just replace one of the
job submit plugins included in the Slurm distribution with your own code
(i.e. use a patch with your own code).
Then just build and install Slurm with that new code.
Building a new plugin outside of the Slurm distribution is possible, but
far more complex.
It also requires access to a multitude of Slurm header files as shown in the
procedure below.</p>

<ol>
<li>You will need to at least partly build Slurm first. The "configure" command
must be executed in order to build the "config.h" file in the build directory.</li>

<li>Create a local directory somewhere for your files to build with.
Also create subdirectories named ".libs" and ".deps".</li>

<li>Copy a ".deps/job_submit_*Plo" file from another job_submit plugin's ".deps"
directory (made as part of the build process) into your local ".deps" subdirectory.
Rename the file as appropriate to reflect your plugins name (e.g. rename
"job_submit_partition.Plo" to be something like "job_submit_mine.Plo").</li>

<li>Compile and link your plugin. Those options might differ depending
upon your build environment. Check the options used for building the
other job_submit plugins and modify the example below as required.</li>

<li>Install the plugin.</li>
</ol>

<pre>
# Example:
# The Slurm source is in ~/SLURM/slurm.git
# The Slurm build directory is ~/SLURM/slurm.build
# The plugin build is to take place in the directory
#   "~/SLURM/my_submit"
# The installation location is "/usr/local"

# Build Slurm from ~/SLURM/slurm.build
# (or at least run "~/SLURM/slurm.git/configure")

# Set up your plugin files
cd ~/SLURM
mkdir my_submit
cd my_submit
mkdir .libs
mkdir .deps
# Create your plugin code
vi job_submit_mine.c

# Copy up a dependency file
cp ~/SLURM/slurm.build/src/plugins/job_submit/partition/.deps/job_submit_partition.Plo \
   .deps/job_submit_mine.Plo

# Compile
gcc -DHAVE_CONFIG_H -I~/SLURM/slurm.build -I~/slurm.git \
   -g -O2 -pthread -fno-gcse -Werror -Wall -g -O0       \
   -fno-strict-aliasing -MT job_submit_mine.lo          \
   -MD -MP -MF .deps/job_submit_mine.Tpo                \
   -c job_submit_mine.c -o .libs/job_submit_mine.o

# Some clean up
mv -f .deps/job_submit_mine.Tpo .deps/job_submit_mine.Plo
rm -fr .libs/job_submit_mine.a .libs/job_submit_mine.la \
   .libs/job_submit_mine.lai job_submit_mine.so

# Link
gcc -shared -fPIC -DPIC .libs/job_submit_mine.o -O2         \
   -pthread -O0 -pthread -Wl,-soname -Wl,job_submit_mine.so \
   -o job_submit_mine.so

# Install
cp job_submit_mine.so file \
   /usr/local/lib/slurm/job_submit_mine.so
</pre>

<p style="text-align:center;">Last modified 30 October 2024</p>

<!--#include virtual="footer.txt"-->