[
https://issues.apache.org/jira/browse/HBASE-12904?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14287768#comment-14287768
]
Andrew Purtell commented on HBASE-12904:
----------------------------------------
I also don't think caching HTable references will buy us much. These are tables
referring to regions moving around (except META) so we are going to be going to
META for region relocation anyway, the caching doesn't give us a benefit.
> Threading issues in region_mover.rb
> -----------------------------------
>
> Key: HBASE-12904
> URL: https://issues.apache.org/jira/browse/HBASE-12904
> Project: HBase
> Issue Type: Bug
> Affects Versions: 0.98.9
> Reporter: Lars Hofhansl
> Assignee: Andrew Purtell
> Fix For: 1.0.0, 2.0.0, 0.98.10, 1.1.0
>
> Attachments: HBASE-12904-0.98.patch, HBASE-12904.patch
>
>
> We've seen various race conditions when using region_mover with multiple
> threads.
> {code}
> NoMethodError: undefined method `getScanner' for nil:NilClass
> isSuccessfulScan at
> /home/sfdc/current//bigdata-hbase/hbase/hbase/bin/region_mover.rb:138
> unloadRegions at
> /home/sfdc/current//bigdata-hbase/hbase/hbase/bin/region_mover.rb:360
> {code}
> {code}
> NoMethodError: undefined method `[]=' for nil:NilClass
> getTable at
> /home/sfdc/current//bigdata-hbase/hbase/hbase/bin/region_mover.rb:64
> unloadRegions at
> /home/sfdc/current//bigdata-hbase/hbase/hbase/bin/region_mover.rb:359
> {code}
> Looking at getTable, it's not thread safe. So the multithreaded that was
> added is incorrect.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
