0

I am trying to understand the metrics emited by argo workflow but their explination isn't helping enough:

For example argo_workflows_pods_count

It is possible for a workflow to start, but no pods be running (e.g. cluster is too busy to run them). This metric sheds light on actual work being done.

Does it mean the count of all the running pods for all the workflows (if this is the case, then, at least for me, doesn't seem correct) from all the namespaces?

There is a difference between this metric and kubernetes_state.pod.* metrics (which would give me the pods with different states, eg: running)?

Razvan Tudorica
  • 475
  • 5
  • 13

1 Answers1

0

Enabling and scraping the endpoint shows the following data exposed:

# HELP argo_workflows_pods_count Number of Pods from Workflows currently accessible by the controller by status (refreshed every 15s)
# TYPE argo_workflows_pods_count gauge
argo_workflows_pods_count{status="Pending"} 0
argo_workflows_pods_count{status="Running"} 0

As we are querying the workflow controller here and there are no additional labels attached to the metric, we can assume that this is indeed the total number of pods created by Argo. However, this is not necessarily the same as kubernetes_state.pod.* as this will also include pods created by other processes.

Sudesh Jethoe
  • 86
  • 5