[
https://issues.apache.org/jira/browse/AMBARI-7896?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14640092#comment-14640092
]
Tyler Hawkes commented on AMBARI-7896:
--------------------------------------
As a first step to this (mainly because of Hortonworks' support) it would be
great to have ambari server installed on 2 machines and have them both use the
same database. Because the database is separate from the cluster I've got it
installed on another VM that is more highly available than my hadoop nodes, but
ambari is installed on the hadoop nodes because we only have support for so
many machines and keeping in-line with that agreement in a cost-effective way
encourages users to install ambari on a hadoop nodes that may or may not be
highly available.
> High Availability Ambari Server
> -------------------------------
>
> Key: AMBARI-7896
> URL: https://issues.apache.org/jira/browse/AMBARI-7896
> Project: Ambari
> Issue Type: Improvement
> Components: ambari-server
> Environment: HDP 2.1
> Reporter: Hari Sekhon
>
> Ambari server needs high availability as loss of the Ambari server machine
> would leave it's cluster unmanageable.
> This could reuse ZooKeeper for Active/Standby election. I'm not sure if it's
> practical in scalability terms to store the configurations in ZK for
> management of larger clusters, perhaps that could use HA databases such as
> MySQL multi-master or replicate the config itself at the application layer to
> the standby Ambari server.
> Right now I guess you can roll your own with Linux HA + DRBD but it feels
> like this should be native as per other components that leverage ZooKeeper.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
