[GitHub] [hbase] Reidddddd commented on pull request #2585: HBASE-25217 [Metrics] Add metrics for Call in IPC response queue

[GitBox <gi...@apache.org>]

Reidddddd commented on pull request #2585:
URL: https://github.com/apache/hbase/pull/2585#issuecomment-718341999


   > Curious whats the context for this change? Noticed something fishy there in the call queue?
   
   The main purpose is to gain better insight of the whole rpc process procedure. Currently, we have in request queue metrics, call processed metrics (queue time, wait time, total time etc), but we don't have metrics for calls in response queue which is the last part of the procedure.
   
   We have some clients in production fond of doing big joins or big scans, so I suspect if any possible that responses would be accumulated in such scenarios, knowing that socket channel only processed 64KB per chunk.
   ```
     protected long channelWrite(GatheringByteChannel channel, BufferChain bufferChain)
     throws IOException {
       long count =  bufferChain.write(channel, NIO_BUFFER_LIMIT); // 64KB
       if (count > 0) this.metrics.sentBytes(count);
       return count;
     }
   ```


----------------------------------------------------------------
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:
users@infra.apache.org