Andrey, We start the caches only at server start up time. 2 of our caches are replicated, all other caches are partitioned with 1 backup. Do you think replicated caches might be causing this issue, when clients leave and join the cluster?
All server nodes start with the same cache config. All client nodes start with the same config but have no knowledge of cache creation. We don't use any node filters. For IgniteCompute we start the Runnable on server nodes, but for cache creation there are no filters. Any luck reproducing? Has anyone else had an issue where a txn seems like it is finished but the key lock is not released? Thanks, Binti -- View this message in context: http://apache-ignite-users.70518.x6.nabble.com/Pessimistic-TXN-did-not-release-lock-on-a-key-all-subsequent-txns-failed-tp10536p11784.html Sent from the Apache Ignite Users mailing list archive at Nabble.com.