[
https://issues.apache.org/jira/browse/HBASE-17717?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15893854#comment-15893854
]
Hudson commented on HBASE-17717:
--------------------------------
SUCCESS: Integrated in Jenkins build HBase-1.1-JDK8 #1936 (See
[https://builds.apache.org/job/HBase-1.1-JDK8/1936/])
HBASE-17717 Explicitly use "sasl" ACL scheme for hbase superuser (elserj: rev
54747ccb285484dbbf823e30d08bace16bbc10bb)
* (add)
hbase-client/src/test/java/org/apache/hadoop/hbase/zookeeper/TestZKUtil.java
* (edit)
hbase-client/src/main/java/org/apache/hadoop/hbase/zookeeper/ZKUtil.java
> Incorrect ZK ACL set for HBase superuser
> ----------------------------------------
>
> Key: HBASE-17717
> URL: https://issues.apache.org/jira/browse/HBASE-17717
> Project: HBase
> Issue Type: Bug
> Components: security, Zookeeper
> Reporter: Shreya Bhat
> Assignee: Josh Elser
> Fix For: 2.0.0, 1.4.0, 1.3.1, 1.1.10, 1.2.6
>
> Attachments: HBASE-17717.001.0.98.patch,
> HBASE-17717.001.branch-1.1.patch, HBASE-17717.001.patch
>
>
> Shreya was doing some testing of a deploy of HBase, verifying that the ZK
> ACLs were actually set as we expect (yay, security).
> She noticed that, in some cases, we were seeing multiple ACLs for the same
> user.
> {noformat}
> 'world,'anyone
> : r
> 'sasl,'hbase
> : cdrwa
> 'sasl,'hbase
> : cdrwa
> {noformat}
> After digging into this (and some insight from the mighty [~enis]), we
> realized that this was happening because of an overridden value for
> {{hbase.superuser}}. However, the ACL value doesn't match what we'd expect to
> see (as hbase.superuser was set to {{cstm-hbase}}).
> After digging into this code, it seems like the {{auth}} ACL scheme in
> ZooKeeper does not work as we expect.
> {code}
> if (superUser != null) {
> acls.add(new ACL(Perms.ALL, new Id("auth", superUser)));
> }
> {code}
> In the above, the {{"auth"}} scheme ignores any provided "subject" in the
> {{Id}} object. It *only* considers the authentication of the current
> connection. As such, our usage of this never actually sets the ACL for the
> superuser correctly.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
