Add named exception to detect when a cluster node has been quarantined by others

[garyiwu]

When a cluster node gets auto-downed/quarantined, it has to restart its own actor system before it can rejoin the cluster. To do this we need a way to reliably detect when a node has been quarantined by other nodes.

Currently, the only way to do this appears to be doing string matching on the exception message of an AssociationError, which is fragile. Ideally there should be a named exception that we can check, so that the node can restart itself as appropriate.

[jdevelop]

👍

[jdevelop]

I faced the same problem and came out with the following solution:

• add the special case class AssociationHandle.ActorSystemQuarantinedEvent that will contain the local address of the system being quarantined, and remote address of the system that has quarantined the current one.
• upon receive of the quarantine message in the Endpoint - send a new AssociationHandle.ActorSystemQuarantinedEvent to the event stream of the current ActorSystem.

Hence any actor could subscribe to those events and in case of quarantine - do something, e.g - trigger system restart.

I found this pretty flexible in terms of usage, however need somebody to review the approach and the proposed patch. Thanks.

[jdevelop]

Perhaps this one could be mark as fixed?

[rkuhn]

Indeed, thanks!