Github user afine commented on the issue:
https://github.com/apache/zookeeper/pull/273
@phunt I'm trying to understand the problem you are describing. Please let
me know what I am missing.
Since `QuorumZooKeeperServer` reimplements `setState` without
`ZooKeeperServerShutdownHandler` I think this is only an issue when dealing
with standalone ZooKeeper servers.
Standalone servers are only instantiated (outside of tests) in
`ZooKeeperServerMain#runFromConfig` (which also does the work of setting the
`ZooKeeperServerShutdownHandler`).
In other words, I think we are dealing with an unfortunate usage of
inheritance. It would be ideal to add this `ZooKeeperServerShutdownHandler` in
a subclass of `ZooKeeperServer` used only for standalone purposes but it looks
like we did the opposite, putting in the `ZooKeeperServerShutdownHandler` in
the default case and reimplementing `setState` for the purpose of ignoring the
`ZooKeeperServerShutdownHandler` it in `QuorumZooKeeperServer`.
When the `ZooKeeperServerShutdownHandler` is needed it is always set (in
production code).
So I guess my question is, what problem would an operator be debugging with
this log message?
---
