-observed dataloss while replacing zookeeper severs (zookeeper 3.5.0)
I am experimenting with zookeeper server node replacement using 3.5.0
In my setup I have 3 zookeeper servers forming a quorum and want to replace
all 3 zookeeper servers with brand new instances of servers.
If I replace all 3 nodes in quick succession (using reconfig -remove x -add
Is there a possibility zookeeper data may not be resynced completely on
the new instances of zookeeper servers due to quick succession. As I
observed that some data nodes (znodes) are not present in the zookeeper
database after replacing all members in the quorum.
Is there any way I can avoid this and how should I ensure data has been
resynced before proceeding for next server replacement.
Appreciate you help.
Thanks & Regards,