Mailing-List: contact common-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Reply-To: common-issues@hadoop.apache.org
Date: Thu, 9 Feb 2012 05:25:59 +0000 (UTC)
From: "Todd Lipcon (Commented) (JIRA)" <jira@apache.org>
To: common-issues@hadoop.apache.org
Message-ID: 
 <88552533.18160.1328765159438.JavaMail.tomcat@hel.zones.apache.org>
In-Reply-To: 
 <1770088487.17728.1328753279374.JavaMail.tomcat@hel.zones.apache.org>
Subject: [jira] [Commented] (HADOOP-8041) HA: log a warning when a failover
 is first attempted
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/HADOOP-8041?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13204274#comment-13204274 ] 

Todd Lipcon commented on HADOOP-8041:
-------------------------------------

My thinking is the following:
For a given proxy object, when we first have a successful RPC, we can set an internal flag in the failover proxy provider indicating that it has connected once. Then, whenever we do a failover, if that flag is set, then we should print a warning. Otherwise, only print it at DEBUG level.

This would allow FsShell commands to not print warnings due to an "already been failed over for a while" situation, but still cause an INFO msg to be printed in MR tasks or HBase servers if a failover takes place while they're accessing DFS.
                
> HA: log a warning when a failover is first attempted 
> -----------------------------------------------------
>
>                 Key: HADOOP-8041
>                 URL: https://issues.apache.org/jira/browse/HADOOP-8041
>             Project: Hadoop Common
>          Issue Type: Sub-task
>          Components: ha
>    Affects Versions: HA Branch (HDFS-1623)
>            Reporter: Eli Collins
>
> Currently we always warn for each client operation made to a NN we've failed over to:
> {noformat}
> hadoop-0.24.0-SNAPSHOT $ ./bin/hdfs dfs -lsr /
> 12/02/08 17:43:04 WARN retry.RetryInvocationHandler: Exception while invoking getFileInfo of class org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolTranslatorPB after 0 fail over attempts. Trying to fail over immediately.
> {noformat}
> I'm going to remove this warning in HDFS-2918 since we shouldn't warn every time a client performs an operation, eg could be weeks after the failover. But we should eg log a warning eg the client first does a failover.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira