bharathv commented on pull request #1593: URL: https://github.com/apache/hbase/pull/1593#issuecomment-622193641
> On other usage of hedged reads, the first thing is the read replica feature. If you can prove that, you can use less code to support same feature in rpc framework than RpcRetryingCaller, then I'm OK with implementing the hedged reads feature in rpc. Ok, let me read the code for read replicas. I had a brief look at it just now and I see that the timeline reads are very much intertwined with AsyncSingleRequestRpcRetryingCaller (which is common for all gets/puts/...etc). We need to separate it out for hedging gets and I think we can port it to use hedging rpc framework but I don't know about the number of lines, it may be more or less. Anyway, I don't think number of lines is always the right metric to go by. > So my suggestion here is still, remove the hedged reads feature to not block the 2.3.0 release, and open another issue for implementing it, as it has much more long term impact than the MasterRegistry feature. Well if you already made your decision, I don't think I can do much to change it. The least we could do is to fix it to make it actually async (which is #1601). If you still have other concerns, I think we'd end up blocking 2.3.0 (which is very close AFAICT, based on jiras from @ndimiduk). Let's see what other reviewers think? Thanks. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
