Background - When I am doing rolling updates on my hazelcast cluster I see higher latencies and throughput. I use docker/kubernetes for my hazelcast deployment and I bring one pod down gracefully and deploy the new changes and continue. Since I do a graceful shutdown I believe there is a lot of partition migration happening and in one of the documentation links for hazelcast I saw :
NOTE: During a partition migration, no other operations are executed on the related partition.
This I believe is the reason for higher latencies as some of the operations are hung. What I want to know is if backups will help me mitigate this issue?
I went through a bunch of documentation to get the purpose of backups in Hazelcast. Everywhere I see it is for preventing data loss. Data loss is not my concern here since I back up all the data to DB using mapstores.