[
https://issues.apache.org/jira/browse/ZOOKEEPER-1987?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14073345#comment-14073345
]
Patrick Hunt commented on ZOOKEEPER-1987:
-----------------------------------------
Hi Alex
bq. Client port/adress are now part of the dynamic information that can be
changed by reconfig.
That's fine. However I see a few issues we'll need to resolve:
1) backward compatibility. We support n to n+1 backward compatibility
(upgrading 3.4 to 3.5). This currently breaks that compatibility. What you are
saying in your comment is fine, however the conversion from old, non-dynamic,
(but valid) config to new, dynamic config is not working properly. There's even
an existing jira for this same issue that a user faced - ZOOKEEPER-1950.
2) The server is not properly validating the configuration.
bq. with 3.5.0 you have to include the ports when you list the servers
if that's the case (missing required parameter) the server needs to log an
error and refuse to come up
3) The server needs to output XX.bak# files when modifying the configuration
files. Currently they are overwritten, in order to support this in a production
environment we need .bak files to be created on each change, so that we have a
history to look at if things go wrong, or if we need to revert if something
goes wrong.
4) afaict there is no documentation. existing documentation (zk admin guide,
examples, etc...) are all based on the "pre-reconfig" requirements. If this
were an optional feature it would be less important (people could just ignore
it). However reconfig is effecting every single user due to the change in
configuration requirements. The config file format has changed, this is a
requirement to run the server.
It looks to me like 1 and 4 need to be addressed asap, otw folks won't really
be able to try out a release. They will run into the same issue I did, the same
issue that was reported in ZOOKEEPER-1950.
Perhaps we can create some sub-tasks on this jira and work through the issues
individually?
> unable to restart 3 node cluster
> --------------------------------
>
> Key: ZOOKEEPER-1987
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-1987
> Project: ZooKeeper
> Issue Type: Bug
> Components: tests
> Affects Versions: 3.5.0
> Reporter: Patrick Hunt
> Assignee: Alexander Shraer
> Priority: Blocker
> Fix For: 3.5.0
>
> Attachments: ZOOKEEPER-1987-ver1.patch, ZOOKEEPER-1987-ver2.patch,
> ZOOKEEPER-1987-ver3.patch, ZOOKEEPER-1987.patch, f1.jstack, l3.jstack,
> test3.tar.gz
>
>
> I tried a fairly simple test, start a three node cluster, bring it down, then
> restart it. On restart the servers elect the leader and send updates, however
> the negotiation never completes - the client ports are never bound for
> example.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
