Skip to content
pipe pipe

Resuming Failed Runs

How resume works

When a pipeline fails, Pipe saves the run state to ~/.pipe/state/<pipeline>/<run-id>.json. You can resume from the last failure:

Terminal window
pipe deploy --resume <run-id>

Pipe reads the state file, skips all steps that completed successfully, and re-executes from the first failed step onward.

Finding the run ID

When a run fails, Pipe prints the run ID in the error output. You can also list state files:

Terminal window
ls ~/.pipe/state/<pipeline>/

What gets skipped

  • Steps that completed successfully in the original run are skipped.
  • Their outputs are restored from the state file and re-injected into the environment so downstream steps have the values they need.

Sensitive steps

Steps marked sensitive: true are always re-executed on resume, even if they succeeded in the original run. This is because their output was not saved to the state file. The step runs again so that downstream steps receive the value via PIPE_* environment variables.

- id: get-token
  run: "vault read -field=token secret/deploy"
  sensitive: true

State rotation

Control how many state files to keep with PIPE_STATE_ROTATE:

Terminal window
export PIPE_STATE_ROTATE=5   # keep last 5 state files per pipeline

Set to 0 to disable rotation (keep all state files). See Environment Variables.