Prometheus exporter for Lustre metrics.
go get github.com/dnhodgson/lustre_exporter
cd $GOPATH/src/github.com/dnhodgson/lustre_exporter
make
./lustre_exporter <flags>
- collector.ost=disabled/core/extended
- collector.mdt=disabled/core/extended
- collector.mgs=disabled/core/extended
- collector.mds=disabled/core/extended
- collector.client=disabled/core/extended
- collector.generic=disabled/core/extended
- collector.lnet=disabled/core/extended
- collector.health=disabled/core/extended
All above flags default to the value "extended" when no argument is submitted by the user.
Example: ./lustre_exporter --collector.ost=disabled --collector.mdt=core --collector.mgs=extended
The above example will result in a running instance of the Lustre Exporter with the following statuses:
- collector.ost=disabled
- collector.mdt=core
- collector.mgs=extended
- collector.mds=extended
- collector.client=extended
- collector.generic=extended
- collector.lnet=extended
- collector.health=extended
Flag Option Detailed Description
- disabled - Completely disable all metrics for this portion of a source.
- core - Enable this source, but only for metrics considered to be particularly useful.
- extended - Enable this source and include all metrics that the Lustre Exporter is aware of within it.
All Lustre procfs and procsys data from all nodes running the Lustre Exporter that we perceive as valuable data is exported or can be added to be exported (we don't have any known major gaps that anyone cares about, so if you see something missing, please file an issue!).
See the issues tab for all known issues.
In the event that you encounter issues with specific metrics (especially on versions of Lustre older than 2.7), please try disabling those specific troublesome metrics using the documented collector flags in the 'disabled' or 'core' state. Users have encountered bugs within Lustre where specific sysfs and procfs files miscommunicate their sizes, causing read calls to fail.