Minor Changes

• #5606 24a1685 Thanks @tim-smart! - backport @effect/cluster from effect v4

  @effect/cluster no longer requires a Shard Manager, and instead relies on the
  RunnerStorage service to track runner state.

  To migrate, remove any Shard Manager deployments and use the updated layers in
  @effect/platform-node or @effect/platform-bun.

  Breaking Changes

  • ShardManager module has been removed
  • EntityNotManagedByRunner error has been removed
  • Shard locks now use database advisory locks, which requires stable sessions
    for database connections. This means load balancers or proxies that rotate
    connections may cause issues.
  • @effect/platform-node/NodeClusterSocketRunner is now
    @effect/cluster/NodeClusterSocket
  • @effect/platform-node/NodeClusterHttpRunner is now
    @effect/cluster/NodeClusterHttp
  • @effect/platform-bun/BunClusterSocketRunner is now
    @effect/cluster/BunClusterSocket
  • @effect/platform-bun/BunClusterHttpRunner is now
    @effect/cluster/BunClusterHttp

  New Features

  • RunnerHealth.layerK8s has been added, which uses the Kubernetes API to track
    runner health and liveness. To use it, you will need a service account with
    permissions to read pod information.

Patch Changes

• Updated dependencies [3c15d5f, 3863fa8, 2a03c76, 24a1685, 24a1685]:
  • effect@3.19.0
  • @effect/cluster@0.51.0
  • @effect/rpc@0.72.0
  • @effect/platform@0.93.0
  • @effect/sql@0.47.0