Monitor Argo workflows

Algorithms in Tator are accomplished with Argo Workflows. This document will show some useful examples for monitoring workflows that have been launched on a Kubernetes cluster.

Retrieve current workflows

To retrieve current workflows, use:

argo list

This will show all workflows. You can use a grep to search for a particular name:

argo list | grep my-workflow-

To get a count of workflows you can pipe into wc:

argo list | wc -l
argo list | grep my-workflow- | wc -l

Delete old workflows

To delete workflows older than a day:

argo delete --older=1d

The older parameter accepts much shorter values as well (e.g., 5m, 3h).

Monitor one particular workflow

To watch a workflow use watch:

argo watch my-workflow-12345

Stream logs from a workflow

To stream logs from a workflow use logs:

argo logs -f my-workflow-12345

The -f flag makes logging continuous.

Get a workflow description

If a workflow is not starting for some reason, you can check if it is because of a workflow error or resource limitation using kubectl:

kubectl describe wf my-workflow-12345