Status metrics for gluster daemon

[cloudbehl]

1. gluster_prometheus_up
2. gluster_server_up
3. gluster_gd2_up
4. gluster_csi_up
5. gluster_block_csi_up
6. gluster_block_up
7. gluster_operator_up

[shtripat]

@cloudbehl I feel the one named gluster_promethues_up should rather be called gluster_exporter_up

[cloudbehl]

@cloudbehl I feel the one named gluster_promethues_up should rather be called gluster_exporter_up

Ack!

[shtripat]

IIUC we can do the below things for marking the running status of services

• gluster_exporter - when exporterstarts, it will set the value of gluster_exporter_up = 1 in prometheus
• gluster_server - run as ps command and if brick processes are running set the value of gluster_server_up = 1
• gluster_gd2 - Make a REST call http://{IP}:24007/v1/hello and if it returns 200 then set the value gluster_gd2_up = 1
• gluster_csi - @Madhu-1 is there some kind of REST call using which we can make out if the CSI driver is up or not. If not do we have to run ps command to get the running status?
• gluster_block_csi - @Madhu-1 is there some kind of REST call using which we can make out if the CSI driver is up or not. If not do we have to run ps command to get the running status?
• gluster_block - @Madhu-1 is there some kind of REST call using which we can make out if the CSI driver is up or not. If not do we have to run ps command to get the running status?
• gluster_opretaor - @JohnStrunk @Madhu-1 is there some kind of REST call using which we can make out if the CSI driver is up or not. If not do we have to run ps command to get the running status?

To mark these services down, we can have recording rules written in prometheus which checks if the status of these services updated for last (say 30 mins). If not updated for last 30 mins, mark the service status as down.

@JohnStrunk @aravindavk @Madhu-1 does this make sense and good to go ahead this way?

[Madhu-1]

gluster_csi - @Madhu-1 is there some kind of REST call using which we can make out if the CSI driver is up or not. If not do we have to run ps command to get the running status?

CSI driver does not provide REST calls but it does provide RPC calls, you can check the status by sending probe request. but I personally do not prefer to do this way, instead of that you can make Kube call and check the pod status (but this also not ensure that CSI driver is healthy or not :( )

[aravindavk]

gluster_gd2 - Make a REST call http://{IP}:24007/v1/hello and if it returns 200 then set the value gluster_gd2_up = 1

metrics_ps plugin already returning glusterd2's status, can upstate derived from that? https://github.com/gluster/gluster-prometheus/blob/master/gluster-exporter/metric_ps.go#L18

Alternatively glusterd2 has ping API(GET http://{IP}:24007/ping)

gluster_csi - @Madhu-1 is there some kind of REST call using which we can make out if the CSI driver is up or not. If not do we have to run ps command to get the running status?

csi driver process will not run in same pod. ps command can't be used. Also gluster-exporter need not export CSI metrics.

gluster_block_csi - @Madhu-1 is there some kind of REST call using which we can make out if the CSI driver is up or not. If not do we have to run ps command to get the running status?

Same as above. ps command is not useful.

gluster_block - @Madhu-1 is there some kind of REST call using which we can make out if the CSI driver is up or not. If not do we have to run ps command to get the running status?

Again ps command is not useful. But https://github.com/gluster/gluster-block-restapi project can expose REST API to provide health details.