Migrating from Rundeck

There is no importer. A migration means rewriting each job as a YAML file, which is usually mechanical. Plan to move a few jobs at a time and run them in parallel with Rundeck before cutting over.

How the concepts map

Rundeck	Dagu
Job	DAG, one YAML file
Job step	Step
Sequential workflow	Steps linked with depends
Job option	params
Error handler	handler_on.failure, or a step's retry_policy and continue_on
Schedule	schedule, with CRON_TZ= for the time zone
Node dispatch over SSH	ssh, per DAG or per step
Agent on a node	Worker selected by worker_selector
Job reference	dag.run step
Key Storage	Secrets
Webhook	Webhooks
Project	Workspace
Job definitions in the database	Files in Git, checked with dagu validate

Example

A nightly deployment job with an option, a retry, and an error handler:

type: graph
schedule: "CRON_TZ=Asia/Tokyo 0 2 * * *"
calendar: weekdays

params:
  - name: environment
    default: staging

ssh:
  user: deploy
  host: app01.example.internal
  key: /etc/dagu/keys/deploy

steps:
  - id: stop_app
    run: sudo systemctl stop myapp

  - id: migrate
    run: /opt/myapp/bin/migrate --env "${params.environment}"
    depends: stop_app
    retry_policy:
      limit: 3
      interval_sec: 30

  - id: start_app
    run: sudo systemctl start myapp
    depends: migrate

handler_on:
  failure:
    run: /opt/ops/page-oncall.sh

What does not carry over

• Node filters and resource models. Dagu has no fleet inventory. Name hosts directly, or use worker labels, and express a fan-out with parallel.
• Per-job ACLs. Roles apply to a whole server or a workspace, so you cannot hand a single job to a group. To limit a group to certain jobs, give those jobs their own workspace.
• Plugins. Rundeck plugins do not run on Dagu. Use a step type or call the same tool from a command.
• Cluster mode. Dagu has one active scheduler with standbys on shared storage, which is a different design from a Rundeck cluster.

SSO, roles, and audit logs need a self-host license.

Steps

1. List jobs by how often they run and who depends on them. Start with low-risk ones.
2. Export the job definitions as YAML from the Rundeck web console or API.
3. Write a script that turns the export into DAG files, or convert a few by hand to learn the mapping.
4. Run dagu validate on every file in CI.
5. Move secrets out of Key Storage.
6. Run the new DAG alongside the Rundeck job, on a test target or with the Rundeck job's side effects turned off, and compare results.
7. Disable the Rundeck schedule and enable the Dagu one.
8. Set up backups and alerts before you retire Rundeck.