Skip to content

Runs you did not start

The connector reports every Argo Workflow in the namespaces its ArgoConnection covers, whoever created it. The ones Dagweave didn’t start show up as adopted runs. You can open them and get failure alerts for them without moving the workflow into Dagweave.

This needs the connection’s status feature, which is on by default when you connect a cluster.

Anything that creates a Workflow in a watched namespace:

  • a CronWorkflow’s runs, including one you deployed with Deploy schedule
  • argo submit or kubectl create -f
  • CI jobs, operators, or anything else with cluster credentials
  • dagweave run on your own machine, once you’ve signed it in. Only you see those until you share one.

The CronWorkflow itself shows up too. It reads Running while the schedule is active and Error once it’s stopped.

A run Dagweave started can show as adopted for a moment, before its dispatch finishes. The duplicate goes away on its own.

Fewer people than for your other runs:

  • Team admins and owners.
  • Anyone given namespace:dispatch on the reporting connector, directly or through a group. No role includes it, so someone has to grant it on purpose.

Nobody else, including Editors and Viewers who can read every other run. They don’t see adopted runs in the list, and opening one by ID returns 404. This is stricter than Teams and access control.

In the Runs panel’s Runs view, adopted runs have their own section below the table, “Not started from Dagweave”. It’s hidden if there are none you can read. Each is named <namespace>/<name>, and where other runs show what started them, it says Adopted.

A run from a schedule you deployed with Deploy schedule is different in one way: Dagweave knows which workflow it came from, because Argo names each one after its CronWorkflow. It says Deployed schedule instead, it’s listed under that workflow’s runs too, and the run view links back to the workflow. The same people can read it as any other adopted run.

The public API leaves them out of GET /api/v1/runs and lists them at GET /api/v1/runs/adopted. Over MCP that’s dagweave_list_adopted_runs. GET /api/v1/runs/{id} returns one if you’re allowed to read it.

Adopted runs don’t count toward health totals, success rates or your team’s limit on runs at once. The sweep that fails runs stuck in Pending skips them. They do count when Dagweave checks whether a cluster you ranked is full.

You see what Argo recorded: the phase, the message, and each step’s start and finish times, with an ADOPTED badge by the title. There are no step logs.

If it fails and your team has failure alerts on, you get the usual alert once, by email or Slack, naming the Dagweave run ID.

You can’t re-run or stop it from Dagweave. Use argo stop or kubectl. You also can’t link it to a workflow you built in Dagweave.