Zookeeper backend timeout and no re-auth

[afterwords]

I'm using zookeeper as my HA backend in production and have run into the following issue:

When zookeeper restarts or expires the vault client session, vault doesn't try to reauth.

2016-10-11 00:20:36.295243 I | Authentication failed: zk: session has been expired by the server 2016/10/11 00:20:36.295456 [WRN] core: leadership lost, stopping active operation 2016-10-11 00:20:36.296029 I | Connected to xxx.xxx.xxx.xxx:2181 2016/10/11 00:20:36.298505 [INF] core: pre-seal teardown starting 2016/10/11 00:20:36.298674 [INF] core/stopClusterListener: stopping listeners 2016/10/11 00:20:36.298898 [INF] core/startClusterListener: shutting down listeners 2016-10-11 00:20:36.358971 I | Authenticated: id=1234567890, timeout=4000 2016/10/11 00:20:36.459922 [INF] core/startClusterListener: listeners successfully shut down 2016/10/11 00:20:36.459974 [INF] core/stopClusterListener: success 2016/10/11 00:20:36.460096 [INF] rollback: stopping rollback manager 2016/10/11 00:20:36.460749 [INF] core: pre-seal teardown complete 2016/10/11 00:20:36.586379 [ERR] core: failed to acquire lock error=zk: not authenticated

You can imagine this is pretty problematic for HA.

[jefferai]

Is this 0.6.2? I seem to remember something similar to this being fixed in it.

[afterwords]

This is 0.6.1. Reviewing 0.6.2 notes now, but not seeing anything about it.

[afterwords]

Also, all of the other nodes had already timed out since they weren't getting any traffic and so I ended up with a bunch of Vault nodes that were all reporting they were in standby with no leader, but couldn't connect to Zookeeper.

[afterwords]

Just tested 0.6.2 in my sandbox and it seems that you are correct @jefferai . Thanks again.

[jefferai]

For completeness, this wasn't in the notes because it was solved via a normal upstream library sync -- see #1933