[CONTROLLER-1941] Controller does not quarantine on isolation/unisolation in cluster

Relates — Wed, 2 Jun 2021 09:31:15 +0000

In a 3 node or a geo-cluster, consider the following scenario:

1. geo-cluster - odl1 to odl6
2. set odl1-odl3 as voting, odl4-odl6 as non-voting
3. isolate odl6 using iptables
4. wait for 5 mins
5. unisolate using iptables -F
6. the odl1/2/3 and odl6 will go into a quarantine loop as below:

2020-05-27T09:06:32,036 | INFO  | opendaylight-cluster-data-akka.actor.default-dispatcher-40 | Remoting                         | 48 - com.typesafe.akka.slf4j - 2.5.22 | Quarantined address [akka.tcp://opendaylight-cluster-data@10.18.165.103:2550] is still unreachable or has not been restarted. Keeping it quarantined.
2020-05-27T09:06:32,037 | INFO  | opendaylight-cluster-data-akka.actor.default-dispatcher-40 | Remoting                         | 48 - com.typesafe.akka.slf4j - 2.5.22 | Quarantined address [akka.tcp://opendaylight-cluster-data@10.18.165.101:2550] is still unreachable or has not been restarted. Keeping it quarantined.
2020-05-27T09:06:32,037 | INFO  | opendaylight-cluster-data-akka.actor.default-dispatcher-40 | Remoting                         | 48 - com.typesafe.akka.slf4j - 2.5.22 | Quarantined address [akka.tcp://opendaylight-cluster-data@10.18.165.102:2550] is still unreachable or has not been restarted. Keeping it quarantined.

This is actually an issue in akka upstream - https://github.com/akka/akka/issues/24764.
But the solution for this is only present in the artery based akka.
Hence, till we move to artery we need to add this workaround to make this scenario work.

OpenDaylight JIRA

[CONTROLLER-1941] Controller does not quarantine on isolation/unisolation in cluster