Skip to content
Helmhive Helmhive Helmhive

Reconcile an uncertain run

Once a run has recorded its mutation boundary, Helmhive refuses to guess. A lost lease, a cancellation, an adapter failure, or a lost result after that point puts the run into Reconciliation required, and the plan with it. Nothing is retried automatically.

Record the reconciliation

Section titled “Record the reconciliation”
  1. Verify the remote system independently: the deployed revision, the service’s health, or the provider’s own run record.
  2. On the run, choose Reconcile outcome.
  3. Pick the Observed outcome: Succeeded, Failed, Not applied, or Manual intervention.
  4. Enter an Evidence reference (a run URL, remote revision, or incident reference) and a Reason of at least ten characters.
  5. Choose Record reconciliation.

The decision is immutable and cannot be replaced. The run becomes Reconciled with a certain outcome, and the plan becomes Succeeded only when you observed success; every other outcome marks it Failed.

Why it works this way

Section titled “Why it works this way”

Read-only steps keep a bounded retry policy because they cannot have partially changed anything. A mutating step can, and a second attempt on top of an unknown state is more dangerous than a human check. See Deployment model.