doc/html/jobcomp_kafka.shtml

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

<h1>Job Completion Kafka plugin guide</h1>

<p>When configured, the <b>jobcomp/kafka</b> plugin attempts to publish a subset
of the fields for each completed job record to a Kafka server. Since Slurm
25.05 it's also possible to optionally send a subset of job record fields when
the job first starts running.</p>

<h2 id="prereq">Requirements<a class="slurm_link" href="#prereq"></a></h2>
<p>The plugin serializes the subset of fields to JSON before each produce
attempt. The serialization is done using the Slurm serialization plugins, so the
<b>libjson-c</b> development files are an indirect prerequisite for this plugin.
</p>

<p>The plugin offloads part of the client producer work to <b>librdkafka</b> and
consumes its API, thus the library development files are another prerequisite.
</p>
<ul>
	<li><a href="https://github.com/confluentinc/librdkafka">librdkafka</a>
		 development files</li>
	<li><a href="related_software.html#json">libjson-c</a> development
		 files</li>
</ul>

<h2 id="config">Configuration<a class="slurm_link" href="#config"></a></h2>

<p>The plugin is configured with the following
<a href="slurm.conf.html">slurm.conf</a> options:</p>

<ul>
<li>
<a href="slurm.conf.html#OPT_JobCompType"><b>JobCompType</b></a>
Should be set to <b>jobcomp/kafka</b>.
<pre>JobCompType=jobcomp/kafka</pre>
</li>

<li>
<a href="slurm.conf.html#OPT_JobCompLoc"><b>JobCompLoc</b></a> This string
represents an absolute path to a file containing 'key=value' pairs configuring
<a href="https://github.com/confluentinc/librdkafka/blob/master/CONFIGURATION.md">
	librdkafka properties</a>. For the plugin to work properly, this file
needs to exist and at least the <b>bootstrap.servers</b> property needs to be
be configured.
<pre>JobCompLoc=/arbitrary/path/to/rdkafka.conf</pre>

<p><b>NOTE</b>: There is no default value for JobCompLoc when this plugin is
configured, and thus it needs to be explicitly set.</p>

<p><b>NOTE</b>: The <b>librdkafka</b> parameters configured in the file
referenced by this option take effect upon slurmctld restart.</p>

<p><b>NOTE</b>: The plugin doesn't validate these parameters, but just logs
an error and fails if any parameter passed to the library API function
rd_kafka_conf_set() fails.</p>

<p>An example configuration file could look like this:</p>

<pre>
bootstrap.servers=kafkahost1:9092
debug=broker,topic,msg
linger.ms=400
log_level=7
</pre>
</li>

<li>
<a href="slurm.conf.html#OPT_JobCompParams"><b>JobCompParams</b></a> Comma
separated list of extra configurable parameters. Please refer to the slurm.conf
man page for specific details. Example:
<pre>JobCompParams=flush_timeout=200,poll_interval=3,requeue_on_msg_timeout,
topic=mycluster</pre>
<p><b>NOTE</b>: Changes to this option do not require a slurmctld restart.
Reconfiguration or SIGHUP is sufficient for them to take effect.</p>
<p><b>NOTE</b>: Please, refer to the man page to configure job start events.</p>
</li>

<li>
<a href="slurm.conf.html#OPT_DebugFlags"><b>DebugFlags</b></a> Optional
<b>JobComp</b> debug flag for extra plugin specific logging.
<pre>DebugFlags=JobComp</pre>
</li>
</ul>

<h2 id="functionality">Plugin Functionality<a class="slurm_link"
		href="#functionality"></a></h2>
<p>For each finished (or optionally start running) job, the plugin
jobcomp_p_record_job_[end|start] operation is executed. A subset of the job
record fields are serialized into a JSON string via the Slurm serialization
plugins. Then the serialized string is attempted to be produced using the
librdkafka rd_kafka_producev() API call.</p>

<p>Producing a message to librdkafka can be done even if the Kafka server is
down. But an error returned from this call makes it so the message is discarded.
Produced messages accumulate in the librdkafka out queue for up to "linger.ms"
milliseconds (a configurable librdkafka parameter) before building
a message set from the accumulated messages.</p>

<p>Then the librdkafka library transmits a produce request. While no "ACK" is
received, messages are conceptually considered to be "in-flight" according to
the library documentation. The library then receives a produce response, which
can be handled in one of two ways:</p>
<dl>
<dt><b>retriable error</b></dt>
<dd>The library will automatically attempt a retry if
no library limit parameter is hit.</dd>
<dt><b>permanent error</b> or <b>success</b></dt>
<dd>The message will be removed
from the library out queue and is staged to the library delivery report queue.
</dd></dl>

<p>The following diagram illustrates the functionality being described:</p>

<img src="jobcomp_kafka_fig1.png">

<p>The <b>jobcomp/kafka</b> plugin has a background poll handler thread that
periodically calls the librdkafka API rd_kafka_poll() function. How frequent
the thread makes the call is configurable via JobCompParams=poll_interval.
The call makes it so that messages in the library delivery report queue are
pulled and handled back to the plugin delivery report callback, which takes
different actions depending on the error message the library set. By default,
successful messages are just logged if DebugFlags=JobComp is enabled, and
messages with permanent errors are discarded, unless the error is message timed
out and JobCompParams is configured with "requeue_on_msg_timeout", which would
instruct the callback to attempt to produce the message again.</p>

<p>On plugin termination, the fini() operation is executed. The rd_kafka_purge()
library API function is called which purges librdkafka out queue messages. The
rd_kafka_flush() API call is also called, which waits until all outstanding
produce requests (and potentially other types of requests) are completed.
How much to wait is also configurable via JobCompParams=flush_timeout
parameter. Purged messages are always saved to the plugin state file in the
Slurm StateSaveLocation, and messages purged while "in-flight" are discarded.
</p>

<p><b>NOTE</b>: You must be using librdkafka v1.0.0 or later in order to take
advantage of the purge functionality described above. With previous versions
the outgoing queue can not be purged to the state file on shutdown, which
means that any messages that weren't delivered before the termination of the
kafka plugin will be lost.</p>

<p>On plugin initialization, after parsing the configuration, saved messages in
the state are loaded and attempted to be produced again. So undelivered messages
should be resilient to slurmctld restarts.</p>

<p>The Kafka broker "host:port" list should be explicitly configured in the
file referenced by JobCompLoc option explained above. The default topic is
the configured Slurm ClusterName, but it can also be configured via
JobCompParams=topic parameter.</p>

<p>The <b>jobcomp/kafka</b> plugin mostly logs informational messages to the
JobComp DebugFlag, except for error messages. The librdkafka by default logs
to the application stderr, but the plugin configures the library to forcefully
log to syslog instead. The library logging level and debug contexts are also
configurable via the file referenced by JobCompLoc, as well as the rest of the
library configuration parameters.</p>

<p style="text-align:center;">Last modified 03 February 2025</p>

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