[ 
https://issues.apache.org/jira/browse/HBASE-18864?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16374941#comment-16374941
 ] 

stack commented on HBASE-18864:
-------------------------------

[~jatsakthi] See the dev-tools/submit-patch.py tool. It will format your 
patches properly.

On this patch, you need a test sir where you pass a type that is unknown so 
your new addtion gets exercised.

If an unknown type, should default be setting scope to '1'; i.e. global, i.e. 
replication everything, everywhere?

I'd say if unknown type, safer to shutdown replication... set it to '0'?

nit: This code path is exercised when we are writing WAL too, right? If so, 
what if a user puts 1M edits in with bad scope -- we'll have 1M lines in file?

Should we include the walkey in the log so an operator has something to use if 
they want to go look for source of the bad scope?

Thanks.



> NullPointerException thrown when adding rows to a table from peer cluster, 
> table with replication factor other than 0 or 1
> --------------------------------------------------------------------------------------------------------------------------
>
>                 Key: HBASE-18864
>                 URL: https://issues.apache.org/jira/browse/HBASE-18864
>             Project: HBase
>          Issue Type: Bug
>          Components: hbase
>    Affects Versions: 1.3.0
>            Reporter: smita
>            Assignee: Sakthi
>            Priority: Major
>              Labels: beginner
>         Attachments: hbase-18864.branch-1.2.001.patch
>
>
> Scenario:
> =============
> add_peer
> create a table
> alter table with REPLICATION_SCOPE => '5'
> enable table replication
> login to peer cluster and try putting data to the table 



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to