CODEWITHSUNDEEP
replicationhazelcastrecoverydistributed-cachingwan
Kin Cheung
Kin Cheung

Reputation: 860

Hazelcast 3.5.5 - Wan Replication Recovery

Say, we have two Hazelcast clusters located in two locations, namely London and Tokyo and we configure WAN replication in active-active mode to replicate a distributed map between them.

If one of the clusters, say London, is completely down. After resuming to services, we found that Hazelcast will not repopulate the data from Tokyo to London and leave London with an empty map.

My question is, if we want to repopulate the data to the downed site after coming back up, what can we do?

ps 1: Do you normally expect a distributed cache to behave like this?
ps 2: In this similar situation, what do you normally do to "recover"?

Thank you!!

Upvotes: 0

Views: 93

Answers (1)

noctarius
noctarius

Reputation: 6094

You will have to provision data on startup. This is normally done using a database or a custom operation to send / retrieve the data. Hazelcast does not provide a generic implementation since requirements are too diverse to just handle them all in a simple implementation, even WAN Replication needs are very different and the default implementation sometimes won't fit.

With Hazelcast 3.6 there is now a new feature called HotRestart which keeps a persistent snapshot of the dataset on (hopefully non-spinning) disk and can reload it almost immediately after cluster failures / restarts.

Upvotes: 1

Related Questions