[
https://issues.apache.org/jira/browse/HBASE-19774?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16322957#comment-16322957
]
Sergey Soldatov commented on HBASE-19774:
-----------------------------------------
Well, that problem persists for quite period of time. But for 1.x releases this
message was not that scary (that was something like "Table X not found, got
Y"). The problem is that in locateRegionInMeta we are using ReverseScan without
setting stopRow. So if the given table doesn't exist we may return a record any
other table. In branch-2 the exception message was changed and now it actually
looks scary and may lead the user to think that HBase has problems with meta
table.
I see two ways how to fix it :
1. Add stop row to the corresponding scan. Not sure whether we have any corner
cases that may be affected.
2. Consider that the returned record as an expected result for the missing
table and just report in a regular way without that scary message.
Personally, I would prefer the first way. Any suggestions, [~stack], [~elserj],
[~te...@apache.org] ?
> incorrect behavior of locateRegionInMeta
> ----------------------------------------
>
> Key: HBASE-19774
> URL: https://issues.apache.org/jira/browse/HBASE-19774
> Project: HBase
> Issue Type: Bug
> Affects Versions: 2.0.0-beta-1
> Reporter: Romil Choksi
> Assignee: Sergey Soldatov
> Fix For: 2.0.0-beta-1
>
>
> When we try to operate with not existing table, in some circumstances we get
> an incorrect report about the not existing table:
> {noformat}
> ERROR: Region of
> 'hbase:namespace,,1510363071508.0d8ddea7654f95130959218e9bc9c89c.' is
> expected in the table of 'nonExistentUsertable', but hbase:meta says it is in
> the table of 'hbase:namespace'. hbase:meta might be damaged.
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
