[
https://issues.apache.org/jira/browse/SOLR-7074?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15049121#comment-15049121
]
Mark Miller commented on SOLR-7074:
-----------------------------------
bq. And perhaps Solr's usage of ZK is so basic that using the embedded ZK is
more than good enough, with decent defaults.
Embedded comes with a lot of limitations. Unless you are toying around, simply
makes no sense to use it. With some of the new dynamic cluster stuff newer ZK
versions have, it could probably be improved, but even if it was, I still think
anyone serious is going to want to run ZK as a separate service.
bq. I'm prepared to make it happen.
+1, would be nice to not have to go dig up the right version and what not.
> Simple script to start external Zookeeper
> -----------------------------------------
>
> Key: SOLR-7074
> URL: https://issues.apache.org/jira/browse/SOLR-7074
> Project: Solr
> Issue Type: Improvement
> Components: SolrCloud
> Reporter: Upayavira
>
> When starting Zookeeper for SolrCloud, we have two choices, bin/solr -c
> (without -z) and run Zookeeper embedded. Whilst this works, it is suggestive
> of bad practice, something you will have to unlearn later.
> The second option is to download Zookeeper, and install it manually. This
> isn't hugely hard, but it is quite manual. However, try it on Windows, and
> you'll find the Windows start script for Zookeeper is extremely thin (no
> start/stop options, etc).
> Solr contains everything needed to start Zookeeper. Why not have a bin/zk
> script that can start it for us. Thus, first time user instructions for
> ZolrCloud could be:
> bin/zk start
> bin/zk uploadconf examples/techproducts/conf
> bin/solr start -c -Z
> -Z here hypothetically says "look in the standard place, localhost:2181, for
> Zookeeper".
> or even:
> bin/zk start -config=examples/techproducts/conf
> bin/ssolr start -Z
> This should support both Windows and Unix.
> bin/zk would support all the functionality of the current zkCli scripts as
> well as being able to start/stop zookeeper.
> This would make initial experience of using Solr a lot simpler, in my view,
> which would be a good thing.
> Not only that, deploying Zookeeper in production, for Solr, need only be push
> a Solr app dir to the server, then run bin/zk start, reducing the complexity
> of installation.
> Thoughts? If this seems like a good idea, and don't share valid objections,
> I'm prepared to make it happen.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
