This is a bespoke Prometheus exporter used to enable the monitoring of Pacemaker based HA clusters.
The exporter is a stateless HTTP endpoint. On each HTTP request, it locally inspects the cluster status by parsing pre-existing distributed data, provided by the tools of the various cluster components.
Exported data include:
- Pacemaker cluster summary, nodes and resources stats
- Corosync ring errors and quorum votes
- SBD devices health status
- DRBD resources and connections stats
(note: only DBRD v9 is supported; for v8.4, please refer to the Prometheus Node Exporter project)
A comprehensive list of all the metrics can be found in the metrics document.
The project can be installed in many ways, including but not limited to:
git clone https://github.com/ClusterLabs/ha_cluster_exporter
cd ha_cluster_exporter
make
make install
go get github.com/ClusterLabs/ha_cluster_exporter
You can find the repositories for RPM based distributions in SUSE's Open Build Service.
On openSUSE or SUSE Linux Enterprise you can just use the zypper
system package manager:
export DISTRO=SLE_15_SP1 # change as desired
zypper addrepo https://download.opensuse.org/repositories/server:/monitoring/$DISTRO/server:monitoring.repo
zypper install prometheus-ha_cluster_exporter
You can run the exporter in any of the cluster nodes.
$ ./ha_cluster_exporter
INFO[0000] Serving metrics on 0.0.0.0:9664
Though not strictly required, it is strongly advised to run it in all the nodes.
It will export the metrics under the /metrics
path, on port 9664
by default.
While the exporter can run outside a HA cluster node, it won't export any metric it can't collect; e.g. it won't export DRBD metrics if it can't be locally inspected with drbdsetup
.
A warning message will inform the user of such cases.
Hint: You can deploy a full HA Cluster via Terraform with SUSE/ha-sap-terraform-deployments.
All the runtime parameters can be configured either via CLI flags or via a configuration file, both or which are completely optional.
For more details, refer to the help message via ha_cluster_exporter --help
.
Note: the built-in defaults are tailored for the latest version of SUSE Linux Enterprise and openSUSE.
The program will scan, in order, the current working directory, $HOME/.config
, /etc
and /usr/etc
for files named ha_cluster_exporter.(yaml|json|toml)
.
The first match has precedence, and the CLI flags have precedence over the config file.
Please refer to the example YAML configuration for more details.
A systemd unit file is provided with the RPM packages. You can enable and start it as usual:
systemctl --now enable prometheus-ha_cluster_exporter
Pull requests are more than welcome!
We recommend having a look at the design document before contributing.
Most development tasks can be accomplished via make.
The default target will clean, analyse, test and build the amd64 binary into the build/bin
directory.
You can also cross-compile to the various architectures we support with make build-all
.
The CI will automatically publish GitHub releases to SUSE's Open Build Service: to perform a new release, just publish a new GH release or push a git tag. Tags must always follow the SemVer scheme.
If you wish to produce an OBS working directory locally, after you have configured osc
locally, you can run:
make obs-workdir
This will checkout the OBS project and prepare a release in the build/obs
directory.
Note that, by default, dev
is used as the RPM Version
field, as well as a suffix for all the binary file names.
To prepare an actual release, you can use the VERSION
environment variable to set this value to an actual release tag.
To commit the release to OBS, run make obs-commit
.