hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Xiao Chen (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-9698) Long running Balancer should renew TGT
Date Fri, 12 Feb 2016 22:15:18 GMT

    [ https://issues.apache.org/jira/browse/HDFS-9698?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15145403#comment-15145403
] 

Xiao Chen commented on HDFS-9698:
---------------------------------

FWIW, I think I'd briefly update here for future reference. Thanks all for the helpful comments
above!

I've seen similar problems, that the Balancer fails with {{Failed to find any Kerberos tgt}}
after several hours. The problem turns out to be Kerberos usage IMHO, and not a bug in hadoop.

According to [Kerberos docs|http://web.mit.edu/kerberos/krb5-1.13/doc/admin/conf_files/krb5_conf.html],
there're {{ticket_lifetime}} and {{renew_lifetime}}. The former being the lifetime of the
TGT, which it can be renewed to extend to a maximum value of the later.
In the failure scenario, a TGT is generated by the user and provided to the balancer (which
means in the balancer context, {{UserGroupInformation.isLoginTicketBased() == true}}). {{client#handleSaslConnectionFailure}}
is behaving correctly on extending the {{ticket_lifetime}}. But there's no way to extend beyond
the {{renew_lifetime}}, and I think a new TGT has to be generated which should not be hadoop's
responsibility in this case.

> Long running Balancer should renew TGT
> --------------------------------------
>
>                 Key: HDFS-9698
>                 URL: https://issues.apache.org/jira/browse/HDFS-9698
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: balancer & mover, security
>    Affects Versions: 2.6.3
>            Reporter: Zhe Zhang
>            Assignee: Zhe Zhang
>         Attachments: HDFS-9698.00.patch
>
>
> When the {{Balancer}} runs beyond the configured TGT lifetime, the current logic won't
renew TGT.



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

Mime
View raw message