I wrote the ForwardingMetricsConsumer for some performance tests.  It does 
work, I just didn't really test it all that well and don't feel comfortable 
with its stability.  If others want to try and harden it they are welcome to, 
but we are also in the process of redesigning how metrics work to fix some 
fundamental flaws in the current metrics system.

LinearDRPCTopologyBuilder is deprecated, but DRPC itself is not.  It has some 
quirks, especially around security, but it is fully supported and at least for 
us we have been using it in production for quite a while.  We are even open 
sourcing an interesting query engine that uses it.  
https://github.com/yahoo/bullet-storm  I also have a pull request up for 2.x 
that I need to finish that improves the scalability of the DRPC server 
drastically for HTTP requests.  So feel free to use DRPC, file bugs against it 
if you find anything, and if you have improvements please let us know.

- Bobby

On Saturday, January 7, 2017, 2:19:23 PM CST, S G <[email protected]> 
wrote:HttpForwardingMetricsConsumer is another such thing mentioned in
http://storm.apache.org/releases/current/Metrics.html

Bobby Evans confirmed on the mailing list that it is not meant to be used
in production.

But the docs do not mention it anywhere and this can cause some innocent
users to try it out and be disappoined.

We should mention the same in the above docs too that its not production
ready.

Thanks
SG



On Sat, Jan 7, 2017 at 12:13 PM, S G <[email protected]> wrote:

> Hi,
>
> I came across https://storm.apache.org/releases/current/
> Distributed-RPC.html and thought of diving into it.
>
> But some research showed me that it is a deprecated feature.
> http://storm.apache.org/2012/08/02/storm080-released.html
>
> Transactional topologies documentation says that explicitly
> https://storm.apache.org/releases/current/Transactional-topologies.html
>
> But DRPC docs do not say so.
> If it is indeed deprecated, we should say so explicitly to avoid users
> wasting time on it.
>
> Thanks
> SG
>
>

Reply via email to