[
https://issues.apache.org/jira/browse/HBASE-8462?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13645217#comment-13645217
]
Enis Soztutar commented on HBASE-8462:
--------------------------------------
bq. If we are not allowing the negative timestamps, should we not allow through
bulk load also?
Yes, but AFAIK, bulk load creates KVs directly, so there is no easy way to
check the conditions there.
bq. This change should go in the trunk version only right?
I think this should go to 0.94 as well, but this will break compat if anybody
relies on it somehow. But, users should not have been relying on it, since if
you put a value with negative ts, you might not be able to read it back. We can
opt to not have in 0.94 as well.
bq. What are the cases to have a negative timestamp in any usecase- is it to
have some special entries?
There should be no use case for this, since it will not work as expected.
> Custom timestamps should not be allowed to be negative
> ------------------------------------------------------
>
> Key: HBASE-8462
> URL: https://issues.apache.org/jira/browse/HBASE-8462
> Project: HBase
> Issue Type: Bug
> Components: Client
> Reporter: Enis Soztutar
> Assignee: Enis Soztutar
> Fix For: 0.98.0, 0.94.8, 0.95.1
>
> Attachments: hbase-8462_v1.patch
>
>
> Client supplied timestamps should not be allowed to be negative, otherwise
> unpredictable results will follow. Especially, since we are encoding the ts
> using Bytes.Bytes(long), negative timestamps are sorted after positive ones.
> Plus, the new PB messages define ts' as uint64.
> Credit goes to Huned Lokhandwala for reporting this.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
