[
https://issues.apache.org/jira/browse/RANGER-1368?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16388081#comment-16388081
]
Zsombor Gegesy commented on RANGER-1368:
----------------------------------------
This should be handled by the installation scripts, not by Ranger startup code.
And making a directory world readable is generally a very bad idea.
> different service on the same machine would make the shared audit local
> buffer directory access confilct when using hdfs to store audit log
> -------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: RANGER-1368
> URL: https://issues.apache.org/jira/browse/RANGER-1368
> Project: Ranger
> Issue Type: Bug
> Components: audit
> Affects Versions: 0.6.0, 0.7.0
> Environment: linux centos 6,ranger 0.6.0
> Reporter: zhangxiong
> Priority: Major
> Attachments:
> 0002-make-the-common-dir-created-for-audit-log-local-file.patch
>
>
> when configured hdfs to store ranger audit log, different service on the same
> machine would make the shared audit local buffer directory access confilct.
> detail:
> suppose that we have a server S with one hdfs NameNode NN and yarn
> ResourceManager RM deployed on it, and ranger configured hdfs as audit log
> storage backend and configured a local log buffer directory /tmp/ranger/log/
> via 'xasecure.audit.hdfs.config.local.buffer.directory' .
> first ,namenode started and find the directory /tmp/ranger/log/ doesn't
> exists, so NN created the dir with access mod 700 and continues its own work
> . then resourcemanager started and need to store its audit log to the already
> exist dir ,unfortunately RM has no access authz to the dir and cannot work
> well.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
