Hi Ariel, Yes, using TopologyValidator is looks like what you are looking for.
PARTITIONED with number of backups as a size of the cluster and REPLICATED are similar, but not same. Replicated cache has backups on all node, partitioned cache always has fixed number of backups. Also, Ignite has some optimizations for Replicated caches based on knowledge that every node contains full dataset. On Wed, Mar 28, 2018 at 7:12 AM, Ariel Tubaltsev <[email protected]> wrote: > I would like to have a strongly consistent (K,V) cluster of 3 nodes, with a > quorum of 2. > 1 node can be down / partitioned, but it won't survive the failure of 2 > nodes. > > In my mind, if I start single node when there are 3 nodes specified in > TcpDiscoveryVmIpFinder, it should not process requests till it joins at > least one additional node. > But it doesn't happen: I tried both PARTITIONED mode, FULL SYNC, number of > backups 2 and also REPLICATED mode. > > Looks like TopologyValidator can help here - I can check that there are at > least 2 nodes - is that a right approach? > > Also, will be there any difference between PARTITIONED mode + FULL_SYNC + > number of replicas as a size of the cluster and REPLICATED mode? > > Thank you > Ariel > > > > > > -- > Sent from: http://apache-ignite-users.70518.x6.nabble.com/ > -- Best regards, Andrey V. Mashenkov
