MoonResilience

MoonResilience is a MoonBit resilience toolkit for service calls, job execution, message handling, and other infrastructure-facing workloads. It provides retry policies, circuit breaking, rate limiting, bulkhead isolation, and a unified execution entry for composing those controls.

The first version is intentionally kept as a pure MoonBit core library. It does not depend on any framework runtime, so it can be reused in CLI tools, backend components, scheduling logic, and teaching examples around stability control.

Why this project

Resilience controls are a common requirement in mature backend systems:

• retry transient failures without hiding permanent faults
• stop cascading failures with circuit breaking
• smooth burst traffic with rate limiting
• isolate expensive or unstable work with bulkheads

MoonResilience packages these controls as a small reusable MoonBit library instead of scattering ad hoc logic across projects.

Core capabilities

• fixed and exponential backoff retry policies
• circuit breaker with Closed, Open, and HalfOpen states
• token-bucket style rate limiter
• bulkhead isolation with explicit in-flight capacity
• event recording hooks for retries, breaker transitions, limiter rejection, and bulkhead rejection
• unified execute_with flow for composing the controls

Repository links

• GitLink: https://www.gitlink.org.cn/q2weasd/MoonResilience
• GitHub: https://github.com/liying-han/MoonBit-

Package layout

• root package: @q2weasd/moonresilience
• demo CLI: cmd/main
• competition materials: docs/competition

Quick start

moon test
moon run cmd/main

Example

let policy = @moonresilience.default_policy()
let hooks = @moonresilience.empty_hooks()
let result = @moonresilience.execute_with(
  policy,
  0,
  hooks,
  fn(attempt : Int) -> @moonresilience.ActionOutcome[String] {
    if attempt < 3 {
      Failure({
        code: "upstream_timeout",
        message: "temporary timeout",
        retryable: true,
      })
    } else {
      Success("ok")
    }
  },
)

Current scope

The current version focuses on a stable core model:

• synchronous execution model with explicit now_ms
• deterministic tests without sleeping or external clocks
• event recording through ResilienceHooks

Planned follow-up work includes metrics adapters, configuration loading, and integration wrappers for HTTP clients or job runners.

License

Apache-2.0