hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Andrew Purtell (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-13768) ZooKeeper znodes are bootstrapped with insecure ACLs in a secure configuration
Date Tue, 26 May 2015 20:54:24 GMT

    [ https://issues.apache.org/jira/browse/HBASE-13768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14559850#comment-14559850

Andrew Purtell commented on HBASE-13768:

bq. Today we release this patch and we decide to not include backup-master and region-in-transition,
so they are still world readable. next month we decide to restrict them to be non-client readable.
what is the upgrade and auto-fix path

Yeah, we should make a whitelist of world-readable znode paths. Then, enumerate znodes under
the root at master startup, recursively make world readable only if appearing in the whitelist,
or recursively make private if not. I think checking only the children of the root znode is
sufficient for now. The code could be updated later if it needs to check down additional levels.

> ZooKeeper znodes are bootstrapped with insecure ACLs in a secure configuration
> ------------------------------------------------------------------------------
>                 Key: HBASE-13768
>                 URL: https://issues.apache.org/jira/browse/HBASE-13768
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Andrew Purtell
>            Assignee: Enis Soztutar
>            Priority: Blocker
>             Fix For: 2.0.0, 0.98.13, 1.0.2, 1.2.0, 1.1.1,,,
>         Attachments: HBASE-13768_v1.patch, HBASE-13768_v2.patch
> A logic error causes HBase in most secure configuration deployments to handle its coordination
state in ZooKeeper via insecure ACLs. Anyone with remote unauthenticated network access to
the ZooKeeper quorum, which by definition includes all HBase clients, can make use of this
opening to violate the operational integrity of the system. For example, critical znodes can
be deleted, causing outages. It is possible to introduce rogue replication endpoints. It is
possible to direct the distributed log splitting facility to split arbitrary files in HDFS.

This message was sent by Atlassian JIRA

View raw message