Mailing-List: contact notifications-help@accumulo.apache.org; run by ezmlm
Precedence: bulk
Reply-To: jira@apache.org
Date: Thu, 18 Aug 2016 03:07:21 +0000 (UTC)
From: "Mike Drob (JIRA)" <jira@apache.org>
To: notifications@accumulo.apache.org
Message-ID: <JIRA.12997630.1471363751000.339725.1471489641568@Atlassian.JIRA>
In-Reply-To: <JIRA.12997630.1471363751000@Atlassian.JIRA>
References: <JIRA.12997630.1471363751000@Atlassian.JIRA> <JIRA.12997630.1471363751409@arcas>
Subject: [jira] [Commented] (ACCUMULO-4407) Use zk chroot instead of
 instance id
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
archived-at: Thu, 18 Aug 2016 03:07:24 -0000


    [ https://issues.apache.org/jira/browse/ACCUMULO-4407?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15425843#comment-15425843 ] 

Mike Drob commented on ACCUMULO-4407:
-------------------------------------

In a perfect world, we wouldn't need the instance ID stuff anymore at all. But, as an initial pass, it might make sense to just chroot ourselves to {{/accumulo/UUID}} in a lot of places and take advantage of the simplifications that that would get us.

> Use zk chroot instead of instance id
> ------------------------------------
>
>                 Key: ACCUMULO-4407
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-4407
>             Project: Accumulo
>          Issue Type: Improvement
>          Components: core, zookeeper
>            Reporter: Mike Drob
>             Fix For: 2.0.0
>
>
> This is something I'd like to see considered for Accumulo 2.0. Probably can't move on it any sooner because it would break a lot of assumptions, but might be possible to make work in a backwards compatible way.
> Currently we use instance id in zookeeper to allow for multiple installations sharing the same zk quorum. This is both a confusing amount of indirection and a headache for operators when they need to intervene because a uuid is not very human-memory friendly.
> Instead, if we specify a chroot path then we can assume the given accumulo instance is the only one we will see. This will simplify a lot of APIs because in essence we will be combining the zk and instance id arguments into one thing.
> I suspect that either we were not aware of zk chroot or it did not exist when we started, hence not using it. But there's not really any good excuses left today to avoid it.


--
This message was sent by Atlassian JIRA
(v6.3.4#6332)