@guozhangwang I was thinking about your compatibility concerns. Could we fix is 
with the following approach: we encode which lock structure to use in rebalance 
protocol (we can simply pump up the version) -- if at least one instance is on 
old version, we still use old locks -- after all instances are on new version, 
we switch from old lock files to new lock files (for this, code must hold old 
lock, get new lock, releases old lock).

Thoughts?

[ Full content available at: https://github.com/apache/kafka/pull/5650 ]
This message was relayed via gitbox.apache.org for [email protected]

Reply via email to