[
https://issues.apache.org/jira/browse/HBASE-21072?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16589423#comment-16589423
]
stack commented on HBASE-21072:
-------------------------------
Took a look at this one. The code for putting the lock in place is still
around inside HBaseFsck, the name of the HBCK class in hbase1 -- see
FileLockCallable. Tough part is that the hbck1 tool has been made to work
against hbase2; it can read state in hbase2, just not set it. The help output
is good on messaging this and in the refguide we have text that says the below:
{code}
HBCK must match HBase server version
You must not use an HBase 1.x version of HBCK against an HBase 2.0+ cluster.
HBCK is strongly tied to the HBase server version. Using the HBCK tool from an
earlier release against an HBase 2.0+ cluster will destructively alter said
cluster in unrecoverable ways.
As of HBase 2.0, HBCK is a read-only tool that can report the status of some
non-public system internals. You should not rely on the format nor content of
these internals to remain consistent across HBase releases.
{code}
So, we should let hbase2 versions of hbck1 run and block hbase1 versions of
hbck1. Let me have hbase2 hbck1 tools put the lock in a different place. That
should do it.
> Block out HBCK1 in hbase2
> -------------------------
>
> Key: HBASE-21072
> URL: https://issues.apache.org/jira/browse/HBASE-21072
> Project: HBase
> Issue Type: Sub-task
> Components: hbck
> Affects Versions: 2.0.1
> Reporter: stack
> Assignee: stack
> Priority: Major
>
> [~busbey] left a note in the parent issue that I only just read which has a
> prescription for how we might block hbck1 from running against an hbase-2.x
> (hbck1 could damage a hbase-2....Its disabled in hbase-2 but an errant hbck1
> from an hbase-1.x install might run).
> Here is quote from parent issue:
> {code}
> I was idly thinking about how to stop HBase v1 HBCK. Thanks to HBASE-11405,
> we know that all HBase 1.y.z hbck instances should refuse to run if there's a
> lock file at '/hbase/hbase-hbck.lock' (given defaults). How about HBase v2
> places that file permanently in place and replace the contents (usually just
> an IP address) with a note about how you must not run HBase v1 HBCK against
> the cluster?
> {code}
> There is also the below:
> {code}
> We could pick another location for locking on HBase version 2 and start
> building in a version check of some kind?
> {code}
> ... to which I'd answer, lets see. hbck2 is a different beast. It asks the
> master to do stuff. It doesn't do it itself, as hbck1 did. So no need of a
> lock/version.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
