How to connect your cluster’s workload manager to COS¶
This how-to guide shows you how to connect your cluster’s workload manager to the Canonical Observability Stack to observe the workload manager’s logs, metrics, and a alerts.
Prerequisites¶
To successfully connect your cluster’s workload manager to COS, you must have:
The Juju CLI client installed on your machine.
Deploy an agent¶
First, in the model hosting your Charmed HPC cluster’s workload manager, deploy a Grafana agent:
juju deploy grafana-agent
Connect the workload manager to the agent¶
After deploying the Grafana agent, connect the agent to the workload manager controller:
juju integrate slurmctld:cos-agent grafana-agent:cos-agent
Make COS available to the workload manager¶
With the agent connected to the workload manager controller, make COS available to the model hosting the cluster’s workload manager:
Important
For the instructions below to succeed, you must have deployed the
offers overlay
as part of your COS cloud deployment.
juju consume microk8s:admin/cos.prometheus-receive-remote-write
juju consume microk8s:admin/cos.loki-logging
juju consume microk8s:admin/cos.grafana-dashboards
Connect the workload manager to COS¶
Now connect the Grafana agent connected to the workload manager controller to COS:
juju relate grafana-agent prometheus-receive-remote-write
juju relate grafana-agent loki-logging
juju relate grafana-agent grafana-dashboards