Resilient Task Orchestrator Example

[ricardozanini]

An educational example demonstrating event-driven task choreography patterns using Quarkus Flow and Kafka.

What It Demonstrates

Event-Driven Choreography: A coordinator workflow decomposes a build pipeline into independent tasks (lint, test, build, deploy) and emits events. Each task is handled by a separate workflow instance that executes, retries on failure, and emits completion events.

Idempotent Phase Execution: Tasks are divided into phases (e.g., compile → package → verify). Completed phases are tracked and skipped on retry, allowing safe resume from the exact failure point.

Automatic Retry: Failed tasks retry up to 5 times. The retry logic properly returns FAILED status instead of throwing exceptions, allowing the workflow to continue through the retry loop.

Pattern: Saga Choreography - tasks communicate via events rather than direct orchestration, providing fault isolation and independent scaling.

Use Case

Build pipeline orchestration where:

• Tasks can fail and retry independently
• One task failure doesn't block others
• Tasks execute idempotently (safe to re-run)
• Phases resume from last checkpoint

[github-actions]

🙈 The PR is closed and the preview is expired.

[ricardozanini]

Blocked by:

• Fix #1354 - Isolate variables in ForExecutor to avoid racing condition to overwrite loop variables serverlessworkflow/sdk-java#1363