https://cwiki.apache.org/confluence/display/IGNITE/%28Partition+Map%29+Exchange+-+under+the+hood
As per the documentation, if we are using apache ignite thick client 2.8 and above, it should not trigger PME in server nodes. But it does not look like it is skipping, I see below logs INFO 2021-02-01 18:56:20,362 [sys-#45%TRCAP%] org.apache.ignite.internal.processors.cache.distributed.dht.preloader.GridDhtPartitionsExchangeFuture: Received full message, will finish exchange [node=7b01dd15-f5b3-48c4-8d7a-2765d9de13e8, resVer=AffinityTopologyVersion [topVer=18, minorTopVer=0]] INFO 2021-02-01 18:56:20,877 [sys-#45%TRCAP%] org.apache.ignite.internal.processors.cache.distributed.dht.preloader.GridDhtPartitionsExchangeFuture: Finish exchange future [startVer=AffinityTopologyVersion [topVer=18, minorTopVer=0], resVer=AffinityTopologyVersion [topVer=18, minorTopVer=0], err=null, rebalanced=true, wasRebalanced=false] INFO 2021-02-01 18:56:21,424 [exchange-worker-#38%TRCAP%] org.apache.ignite.internal.processors.cache.GridCachePartitionExchangeManager: Skipping rebalancing (no affinity changes)org.apache.ignite.internal.processors.cache.distributed.dht.preloader.GridDhtPartitionsExchangeFuture: Completed partition exchange I see "Skipping rebalancing" on GridCachePartition but it says "Completed partition exchange" in GridDhtPartitions. What is the difference between GridDhtPartitionsExchangeFuture and GridCachePartitionExchangeManager. At present our thick clients are taking more than 2 minutes to connect to servers nodes (8 jvms) and most of the time has been taken here. Wondering if we skip this GridDhtPartitionsExchange then it will boost up the connectivity time. Is there a way to skip it ? and also is it valid to skip it ? -- Sent from: http://apache-ignite-users.70518.x6.nabble.com/