0

This information eluded me when searching through the documentation. How does Apache Ignite, if used in cluster mode, synchronize between nodes when these nodes become inactive and active again?

Assume the following scenario:

I have 2 nodes. Node A is elected Leader. Everything works until node B crashes. Node A continues to work, more transactions are executed. Before node B comes back up, node A goes down. After that, node B comes back up. What happens in this case? Does node B know it was not the master and has to wait until node A comes back up? Does it throw an exception and fails to start? Does it simply not care?

Is there any information on the docs for such scenarios?

PentaKon
  • 3,359
  • 3
  • 32
  • 64

1 Answers1

0

There is a rebalance process, which synchronizes data between nodes.

There is no such thing as a Leader in Ignite. All data in the cache divided by partitions and each node have its own subset of primary partitions.

Before node B comes back up, node A goes down

Do you have persistence enabled? If no, then you just lost data, since all nodes in cluster went down.

Evgenii Zhuravlev
  • 2,817
  • 1
  • 7
  • 15
  • Even if I have persistence enabled, the scenario I'm describing is still a valid question. If a node in the topology goes down early and all other nodes do some more *writes* before they all go down and then the first node comes back up by itself it will have a *deprecated* view of the data and no other node to synchronize with. How is that scenario handled? – PentaKon May 28 '19 at 07:46