Alerts - OCS Cluster and Cluster Nodes health

[umangachapagain]

Need following status alerts:

Node status (Up/Down) Container status (Up/Down) Gluster peer in cluster status (Connected/Disconnected) Glusterd2 service status (Up/Down) Cluster status

[umangachapagain]

@shtripat How do we get these metrics? @JohnStrunk I was thinking if these metrics should come from an exporter in anthill as it would have the real updates about the Cluster or Node health which it maintains for reconciliation.

[JohnStrunk]

@JohnStrunk I was thinking if these metrics should come from an exporter in anthill as it would have the real updates about the Cluster or Node health which it maintains for reconciliation.

I'm hesitant to get these items from Anthill. It will have its own view of each, but we then get a dependency... If the operator is down or malfunctioning, the alerts are potentially wrong.

I would expect many of these to come via data from gluster-prometheus or health checks on labeled pods. The benefit of using g-p is that as long as 1 gd2 pod is ready, the exporter should be available through the gd2 client service.

[cloudbehl]

Node status (Up/Down)

It can come from K8s(node exporter). We can add a recording rule and set it under a gluster namespace.

Container status (Up/Down)

It can come from K8s. but I don't know how useful this will be.

Gluster peer in cluster status (Connected/Disconnected)

It can be provided by glusterd2 API

Glusterd2 service status (Up/Down)

It can be provided by glusterd2 api /ping endpoint

Cluster status

It can be provided by v1/cluster/{cluster_id}/status

@umangachapagain @JohnStrunk