[
https://issues.apache.org/jira/browse/HADOOP-5033?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12663904#action_12663904
]
Jerome Boulon commented on HADOOP-5033:
---------------------------------------
>> an HTTP post to the collector doesn't have interesting semantics. So
>> ditching add(List<>) will prevent mistaken assumptions.
No totally true.
The ChukwaHTTPSender is an implementation of the ChukwaSender interface and if
you were using a JDBCCollector you will open a transaction, insert the list of
chunks in batch mode then commit. I think that it could be the same for an HTTP
post.
It best to be able to acknowledge at the HTTP post level that at each
individual chunk, at least it gave us the flexibility to do it at this level.
So +1 for remove add(Chunk) and not ditching add(List<>) .
The next question is do we want to report any warning/error in a well defined
Object so the collector can report that in a unified way.
Right now the HTTPSender is not using the HTTP response.
> chukwa writer API is confusing
> ------------------------------
>
> Key: HADOOP-5033
> URL: https://issues.apache.org/jira/browse/HADOOP-5033
> Project: Hadoop Core
> Issue Type: Improvement
> Components: contrib/chukwa
> Reporter: Ari Rabkin
> Assignee: Ari Rabkin
> Priority: Minor
>
> The ChukwaWriter interface has both add(Chunk) and add(List<Chunk>). The
> code doesn't actually use the former. I'd like to remove it. Thoughts?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
