hadoop-common-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Arun Suresh (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HADOOP-11824) HDFS client and YARN should check for elected master with Zookeeper
Date Fri, 10 Apr 2015 19:18:12 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-11824?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14490178#comment-14490178
] 

Arun Suresh commented on HADOOP-11824:
--------------------------------------

[~nkedel], As far as HDFS is concerned, it is possible, atleast currently that the cluster
is deployed without ZK (manual failover). There has been some discussion around that in HDFS-7858

> HDFS client and YARN should check for elected master with Zookeeper
> -------------------------------------------------------------------
>
>                 Key: HADOOP-11824
>                 URL: https://issues.apache.org/jira/browse/HADOOP-11824
>             Project: Hadoop Common
>          Issue Type: Improvement
>            Reporter: Nate Edel
>            Assignee: Nate Edel
>
> When trying to connect to a NameNode (DFS) or Resource Manager (Yarn), clients will try
multiple connections in order to find the elected master node.  Clients should instead check
the relevant znode in zookeeper.
> This will be especially valuable in DFS after HDFS-6440 is adopted, as the worst case
scenario in the DFS client will then be testing more than one standby nodes in order before
finding the master.



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

Mime
View raw message