I am testing rolling upgrade scenario and cache become unresonsive for around 3 minutes during rolling upgrade. I observed that query doesn't failover to other running member. I used following properties to improve the situation and need your inputs if i run into other failure scenarios:
"hazelcast.client.invocation.timeout.seconds" = 10;
"hazelcast.client.heartbeat.timeout" = 5000;
"hazelcast.client.heartbeat.interval" = 1000;
member properties:
"hazelcast.max.no.heartbeat.seconds" =20;
"hazelcast.client.heartbeat.timeout" =5000;
"hazelcast.client.heartbeat.interval" =1000;
"hazelcast.member.list.publish.interval.seconds" =10;
"hazelcast.client.invocation.timeout.seconds" =10;
I can see lots of disconnection logs. I also saw failover happening to other member when invocation timeout in 10 seconds. I have seen some request timeouts which didn't occur before.
Pls. someone throw some light if above configuration is good?
Regards,
Anup