hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Eric Friesen (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-4484) 0.18.1 breaks SOCKS server setting
Date Tue, 21 Oct 2008 23:26:44 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-4484?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12641661#action_12641661
] 

Eric Friesen commented on HADOOP-4484:
--------------------------------------

The change specifically relates to DNS resolution. While 17.2 would connect to the jobtracker
and namenode using the IP addresses (as they are listed in the hadoop-site.xml file, 18.1
somehow gets an internal hostname from some communication through the proxy server, then tries
to resolve it outside of the proxy server. I don't see why hadoop should ever insist on using
hostnames when hadoop-site.xml file specifically lists the jobtracker and namenode using IP
addresses.

> 0.18.1 breaks SOCKS server setting
> ----------------------------------
>
>                 Key: HADOOP-4484
>                 URL: https://issues.apache.org/jira/browse/HADOOP-4484
>             Project: Hadoop Core
>          Issue Type: Sub-task
>    Affects Versions: 0.18.1
>            Reporter: Eugene Hung
>            Priority: Minor
>
> Our cluster is behind a gateway, and we asked users to use hadoop.socks.server to access
the cluster.  
> Recently, we "upgraded" our cluster from 0.17.2 to 0.18.1.  However, this creates the
following problem:
> Wrote input for Map #4
> Starting Job
> java.net.UnknownHostException: unknown host: hadoop-master
> at org.apache.hadoop.ipc.Client$Connection.<init>(Client.java:195)
> at org.apache.hadoop.ipc.Client.getConnection(Client.java:779)
> at org.apache.hadoop.ipc.Client.call(Client.java:704)
> at org.apache.hadoop.ipc.RPC$Invoker.invoke(RPC.java:216)
> at org.apache.hadoop.dfs.$Proxy0.getProtocolVersion(Unknown Source)
> at org.apache.hadoop.ipc.RPC.getProxy(RPC.java:319)
> Nothing else has changed and we were able to duplicate this error on a separate cluster
setup, so we
> believe 0.18.1 has broken this feature.  

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message