Do we intend for our thrift RPC implementation to be consumable as public
API?
This question is inspired by recent "How to get the Accumulo statistic
data" thread over on user@, where we point a user at the GetMasterStats
class.
How stable is that in theory? I understand that in practice it proba
IMO we do not. The thrift interface that we currently have is not easily
consumable (you have to understand a lot about the metadata and cluster
layout), and we don't include it in SemVer. Additional features like
GetMasterStats should be plumbed through to the Java API if we want to
expose them.
Agreed.
I think in the case of GetMasterStats, we don't have any other way to
actually do it via public API now. It would be nice to have something in
the future, but that wouldn't help in the present.
Adam Fuchs wrote:
IMO we do not. The thrift interface that we currently have is not easily
I agree. The thrift API should not be public API (except the proxy, of course).
--
Christopher L Tubbs II
http://gravatar.com/ctubbsii
On Thu, Jun 4, 2015 at 12:48 PM, Adam Fuchs wrote:
> IMO we do not. The thrift interface that we currently have is not easily
> consumable (you have to understa