When a Worker resource is created / updated / deleted we publish a WorkerEvent which has the operation type and the serialized resource. All replicas, including the one that performed the update consume the events and reply them into the worker cache the same way, from the DB. There is a potential performance optimization by making the replica that performed the update publish the notification locally (and faster).
The event from the DB will arrive later and be ignored (because the version is not newer). This way the staleness of the workercache on each replica gets reduced.
When a
Workerresource is created / updated / deleted we publish aWorkerEventwhich has the operation type and the serialized resource. All replicas, including the one that performed the update consume the events and reply them into the worker cache the same way, from the DB. There is a potential performance optimization by making the replica that performed the update publish the notification locally (and faster).The event from the DB will arrive later and be ignored (because the version is not newer). This way the staleness of the workercache on each replica gets reduced.