forked from Controls/gitea-pages
88 lines
4.5 KiB
ReStructuredText
88 lines
4.5 KiB
ReStructuredText
``profile::icinga::checks::slurm``
|
|
==========================================
|
|
|
|
This profile should be called when **Slurm** (server, cn, ui) is being configured.
|
|
|
|
Setup specific checks for Slurm nodes:
|
|
* It can check ``slurmd`` service. This check makes sense on computing nodes running ``slurmd``.
|
|
* It can check ``munge`` service. This check makes sense on any node with a configured **Slurm** client/server.
|
|
* It can check ``slurmctld`` service. This check makes sense on server nodes, will disable ``slurmd`` check.
|
|
* It can check ``slurmdbd`` service. This check makes sense on server nodes.
|
|
* It can check ``sinfo`` status, detecting miss-behaving nodes.
|
|
|
|
By default, since the most common node in a **Slurm** cluster is a computing node, this check will report about
|
|
``slurmd`` and ``munge`` status.
|
|
|
|
Parameters
|
|
----------
|
|
|
|
===================== ========== =========================================================================
|
|
**Name** **Type** **Default**
|
|
--------------------- ---------- -------------------------------------------------------------------------
|
|
skip_munge Boolean ``icinga::checks::options::slurm::skip_munge`` **``false``**
|
|
skip_slurmd Boolean ``icinga::checks::options::slurm::skip_slurmd``: **``false``**
|
|
check_slurmdbd Boolean ``icinga::checks::options::slurm::check_slurmdbd`` **``false``**
|
|
check_slurmctld Boolean ``icinga::checks::options::slurm::check_slurmctld`` **``false``**
|
|
ignore_draining Boolean ``icinga::checks::options::slurm::ignore_draining`` **``false``**
|
|
ignore_drained Boolean ``icinga::checks::options::slurm::ignore_drained`` **``false``**
|
|
check_nodes Boolean ``icinga::checks::options::slurm::check_nodes`` **``false``**
|
|
no_reason Boolean ``icinga::checks::options::slurm::no_reason`` **``true``**
|
|
no_timestamp Boolean ``icinga::checks::options::slurm::no_timestamp`` **``false``**
|
|
===================== ========== =========================================================================
|
|
|
|
``skip_munge``
|
|
~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
By default ``munge`` service is checked. Can be disabled by setting ``icinga::checks::options::slurm::skip_munge: false``.
|
|
|
|
``skip_slurmd``
|
|
~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
By default ``slurmd`` service is checked. Can be disabled by setting ``icinga::checks::options::slurm::skip_slurmd: false``.
|
|
|
|
|
|
``check_slurmdbd``
|
|
~~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
By default ``slurmdbd`` service is disabled. Can be enabled by setting ``icinga::checks::options::slurm::check_slurmdbd: true``.
|
|
|
|
``check_slurmctld``
|
|
~~~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
By default ``slurmctld`` service is disabled. Can be enabled by setting ``icinga::checks::options::slurm::check_slurmctld: true``, in this case ``icinga::checks::options::slurm::skip_slurmd``
|
|
should be also set to ``true``.
|
|
|
|
`ignore_draining``
|
|
~~~~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
By default nodes in ``draining`` state are checked and will be reported as ``[WARNING]``. This can be disabled by setting ``icinga::checks::options::slurm::ignore_draining: true``.
|
|
When set to ``true``, ``draining`` nodes will be reported anyway but will be considered as ``[OK]``
|
|
|
|
``ignore_drained``
|
|
~~~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
By default nodes in ``drained`` state are checked and will be reported as ``[WARNING]``. This can be disabled by setting ``icinga::checks::options::slurm::ignore_drained: true``.
|
|
When set to ``true``, ``drained`` nodes will be reported anyway but will be considered as ``[OK]``
|
|
|
|
``no_reason``
|
|
~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
By default **no** ``Reason`` is reported. This is because it will add extra text to the alarm output when nodes are ``drained``/``draining``/``down``/``failed``.
|
|
Printing ``Reason`` can be turned on by setting ``icinga::checks::options::slurm::no_reason: false``.
|
|
|
|
``no_timestamp``
|
|
~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
By default ``Timestamp`` is being reported. Timestamp shows the date and time when a node was set ``drained``/``draining``/``down``/``failed``.
|
|
As this setting adds extra text in the alarm, it can be disabled by setting ``icinga::checks::options::slurm::no_reason: true``.
|
|
|
|
Facts
|
|
-----
|
|
|
|
When some specific facts are detected, it will trigger some actions.
|
|
|
|
===================================== =============== ==============================================
|
|
**Fact** **Value(s)** **Action description**
|
|
------------------------------------- --------------- ----------------------------------------------
|
|
===================================== =============== ==============================================
|