[banjoh]

Summary

When a cluster stays queued, pollForStatus times out with only Cluster did not reach state running within N seconds. The API already reports why the cluster is waiting (last_scheduling_status, for example exceeds-limits), and pollForStatus already reads it, but it was only shown for the error state. This makes the reason visible while waiting and on timeout.

Changes

• Include the last scheduling status, with the same explanation the vendor portal shows, in the per-poll debug line and in the timeout error.
• Fail immediately when the cluster enters upgrade_error, as it already does for error, instead of waiting for the timeout.
• Retry on 5xx and failure on other HTTP errors are unchanged, as is the function signature.

Example timeout error:

Cluster did not reach state running within 1200 seconds. Last status: queued, last scheduling status: exceeds-limits (quota exceeded, manage your quota on the Compatibility Matrix settings page in the vendor portal)

Testing

• Added tests for the timeout error with a scheduling status and for upgrade_error. npm test passes.

Once released, bumping replicated-lib in replicated-actions surfaces this in the create-cluster and upgrade-cluster actions.

[kriscoleman]

🧼 clean!