hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Gary Helmling (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-4263) Prevent RegionObservers from interfering with system tables.
Date Sat, 27 Aug 2011 05:32:29 GMT

    [ https://issues.apache.org/jira/browse/HBASE-4263?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13092200#comment-13092200
] 

Gary Helmling commented on HBASE-4263:
--------------------------------------

See:

HBASE-1697
HBASE-3025
HBASE-1712

I completely disagree about access control not being a "strong" use case.  And I'm sure it's
not the only possible or valid use case.

I also disagree that this would be a desirable restriction and I think the downsides are far
worse than the benefits.  There are still countless ways a coprocessor can crash a server.
 What are we buying here?

If the underlying need is to have an easy way for users to load RegionObservers on all *user*
tables, then I think it would be easy to add another configuration property to enable that.
 Coprocessors loaded through that property could just skip loading on -ROOT- and .META.  If
that's the issue, then let's fix it.

But removing the current property that allows loading RegionObservers on *all* tables or restricting
the ability to run RegionObservers on -ROOT- and .META. severely undermines coprocessors as
an extension mechanism and I'll be -1 on changes in that direction.

> Prevent RegionObservers from interfering with system tables.
> ------------------------------------------------------------
>
>                 Key: HBASE-4263
>                 URL: https://issues.apache.org/jira/browse/HBASE-4263
>             Project: HBase
>          Issue Type: Bug
>          Components: coprocessors
>    Affects Versions: 0.92.0
>            Reporter: Lars Hofhansl
>            Priority: Minor
>             Fix For: 0.92.0
>
>
> It turns out that a RegionObserver can interfere with - ROOT - and .META.
> That seems weird and should be prevented.
> (The one use case for this that I could come up with is access control by Region by intercepting
actions on .META., and I don't think that's a particularly strong use case).
> I'll attach a patch as soon as I get to it.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message