[
https://issues.apache.org/jira/browse/HADOOP-12775?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15137327#comment-15137327
]
Daniel Templeton commented on HADOOP-12775:
-------------------------------------------
Thanks, [~andrew.wang]. The metrics system takes a class name as the
source/sink to start. It's then instantiated in complete isolation. That
works fine in reality, but it's ugly for testing. The alternative to the
static {{suppliedConf}} is to have the test write out a configuration file and
then convince the metrics system to read it. That's also pretty ugly.
On the {{suppliedFilesystem}} variable, I still don't exactly understand why I
need it. When using a secure mini-cluster, if I try to use
{{FileSystem.get()}}, all operations fail with a checksum error. Know anything
about that?
I have thought about making the interval configurable. And now that you've
asked the question, I guess I gotta. :) Would number of minutes be a
reasonable way to configure the rollover interval?
> RollingFileSystemSink doesn't work on secure clusters
> -----------------------------------------------------
>
> Key: HADOOP-12775
> URL: https://issues.apache.org/jira/browse/HADOOP-12775
> Project: Hadoop Common
> Issue Type: Bug
> Affects Versions: 2.9.0
> Reporter: Daniel Templeton
> Assignee: Daniel Templeton
> Priority: Critical
> Attachments: HADOOP-12775.001.patch, HADOOP-12775.002.patch,
> HADOOP-12775.003.patch
>
>
> If HDFS has kerberos enabled, the sink cannot write its logs.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
