[
https://issues.apache.org/jira/browse/HBASE-20592?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16493088#comment-16493088
]
Appy commented on HBASE-20592:
------------------------------
bq. Why I'd prefer to have this tool in 2.0+ is to reduce the number of
different HBase versions the users have to use during upgrade. Their existing
1.x cluster won't have the tool but it could be part of the new version so they
wouldn't need a third one.
Well if it could work this way, it'd be best.
But there are few considerations:
- new admin api is incompatible with 1.x. So while 2.0 uses
TableDescriptor/ColumnFamilyDescriptor, 1.x uses HTableDescriptor,
HColumnDescriptor. In the case (might want to check the code) that the wire
bits are same (PBs), and what's written using one is readable by another, it
might work.
- I think that's how CM team wanted it, these tools in separate standalone jars
(so shading and stuff to make then not need any external when running)
> Create a tool to verify tables do not have prefix tree encoding
> ---------------------------------------------------------------
>
> Key: HBASE-20592
> URL: https://issues.apache.org/jira/browse/HBASE-20592
> Project: HBase
> Issue Type: New Feature
> Components: Operability, tooling
> Reporter: Peter Somogyi
> Assignee: Peter Somogyi
> Priority: Minor
> Fix For: 2.1.0
>
> Attachments: HBASE-20592.master.001.patch,
> HBASE-20592.master.002.patch
>
>
> HBase 2.0.0 removed PREFIX_TREE encoding so users need to modify data block
> encoding to something else before upgrading to HBase 2.0+. A tool would help
> users to verify that there are no tables left with PREFIX_TREE encoding.
> The tool needs to check the following:
> * There are no tables where DATA_BLOCK_ENCODING => 'PREFIX_TREE'
> * -Check existing hfiles that none of them have PREFIX_TREE encoding (in
> case table description is changed but hfiles were not rewritten)-
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
