Hello! I have tried your reproducer, and indeed I observe the same NPE. I will also update your ticket.
It can even be simplified. It is enough to remove cacheConfigurations section from client node config (default-config.xml in your case) while keeping it in server node config to observe this problem reliably. + irudyak@ I'm asking the cassandra contributor to chime in. Is it a known limitation that cacheConfiguration should be set up on all nodes for cassandra-backed caches? Indeed PojoField.accessor is transient and it is never populated when it have to travel over the network. There's an immediate workaround: you can set clientReconnectDisabled setting to true, re-create client node manually in case of disconnect. This way you are guaranteed to have "good" instance of cache store every time. Regards, -- Ilya Kasnacheev 2018-06-14 9:57 GMT+03:00 kotamrajuyashasvi <[email protected]>: > Created Ignite ticket for the issue: > https://issues.apache.org/jira/browse/IGNITE-8788 > > > > -- > Sent from: http://apache-ignite-users.70518.x6.nabble.com/ >
