Mailing-List: contact dev-help@zookeeper.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@zookeeper.apache.org
Date: Mon, 21 Aug 2017 18:35:00 +0000 (UTC)
From: "Dan Benediktson (JIRA)" <jira@apache.org>
To: dev@zookeeper.apache.org
Message-ID: <JIRA.13096501.1503336935000.81860.1503340500177@Atlassian.JIRA>
In-Reply-To: <JIRA.13096501.1503336935000@Atlassian.JIRA>
References: <JIRA.13096501.1503336935000@Atlassian.JIRA> <JIRA.13096501.1503336935102@jira-lw-us.apache.org>
Subject: [jira] [Commented] (ZOOKEEPER-2879) Adding observers dynamically
 without server id
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
archived-at: Mon, 21 Aug 2017 18:35:08 -0000


    [ https://issues.apache.org/jira/browse/ZOOKEEPER-2879?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16135568#comment-16135568 ] 

Dan Benediktson commented on ZOOKEEPER-2879:
--------------------------------------------

It's also unclear to me what problem this feature is intended to solve.
- Efficiency of dynamic config should be completely unimportant: the cost of another < 100 bytes of configuration data per observer is vanishingly small. This doesn't seem to be the real goal, just an added benefit, but let's be clear that it isn't a benefit.
- The real goal seems to be ease of management, but I don't really understand that. Can you elaborate in what way it is easier to not have server IDs? FWIW, we operate dozens of ZK clusters, some of them with more than 50 members, so I'm not unfamiliar with the problems of managing "large" ZK ensembles, for some definition of large; having to assign unique server IDs isn't really one of them, at least at our scale.

> Adding observers dynamically without server id
> ----------------------------------------------
>
>                 Key: ZOOKEEPER-2879
>                 URL: https://issues.apache.org/jira/browse/ZOOKEEPER-2879
>             Project: ZooKeeper
>          Issue Type: Improvement
>          Components: quorum
>    Affects Versions: 3.6.0
>            Reporter: Fangmin Lv
>            Assignee: Fangmin Lv
>
> Dynamic config requires observer has unique server id, which means we cannot simply add observer with dynamic server id -1. For large observer cluster, it's much more easier to add observer without unique server id if it doesn't need to be promoted to participant. Also, it will make dynamic config more efficient, we don't need to store and send the long list of observers during re-config.


--
This message was sent by Atlassian JIRA
(v6.4.14#64029)