[ 
https://issues.apache.org/jira/browse/FLUME-1715?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14276180#comment-14276180
 ] 

Jonathan Creasy commented on FLUME-1715:
----------------------------------------

>From the OpenTSDB Mailing list:

On Thursday, September 25, 2014 9:37:38 AM UTC-7, Jonathan Creasy wrote:
I haven't really gotten too familiar with the code, but could the right classes 
be extracted from OpenTSDB and included as a jar within a Flume sink to have 
Flume write to HBase as if it were OpenTSDB 2.0 doing the writing?

On Tue, Oct 14, 2014 at 12:11 AM, ManOLamancha <[email protected]> wrote:
Super simple to use, just include the tsdb jar file and either overload the 
Config class or load it from a file, then instantiate a TSDB object and start 
calling addDatapoint(). Boom! We're looking at doing the same thing with Kafka. 

> OpenTSDB Sink
> -------------
>
>                 Key: FLUME-1715
>                 URL: https://issues.apache.org/jira/browse/FLUME-1715
>             Project: Flume
>          Issue Type: New Feature
>          Components: Sinks+Sources
>            Reporter: Mike Percy
>            Assignee: Jayant Shekhar
>
> (Cloned from JIRA issue tracking OpenTSDB sink impl against Flume 0.9)
> It would be useful to have an OpenTSDB sink for Flume 1.x. Since the protocol 
> is a text-based TCP protocol, basically a netcat-type protocol, I don't see 
> any concern about licensing issues (OpenTSDB is LGPL licensed).
> A couple of ideas on an implementation:
> * Have one mode that works "out of the box" where we assume that events sent 
> to this sink are already fully formatted as OpenTSDB "put" queries. Then, 
> this sink only really has to handle error conditions sent back from OpenTSDB, 
> not serialization
> * Have another mode where the user can do some serialization of arbitrary 
> events as OpenTSDB queries. The rest of the logic remains the same as above.
> More info: http://opentsdb.net/metrics.html
> Not sure where the docs are on wire protocol error handling logic. I am 
> pretty sure it uses a "no news is good news" policy, whereby it returns 
> nothing on success, and sends some message back on error.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to