zhangmeng916 opened a new issue #1067: URL: https://github.com/apache/helix/issues/1067
In Helix design, relay message (P2P message) is leveraged to improve mastership handoff latency. However, from our customer's using experience, we found that a large percent of messages that should be sent as relay messages are not, which caused high mastership handoff latency. The reason is that in message generation phase, we hold on the generation of SLAVE -> MASTER message to the future master if there is already a pending message for that instance on the same partition. But in many cases, the pending message is already processed, and the current state of that partition is also updated, and only the message itself has not been deleted yet. In these scenarios, we should not wait on the pending message, instead, the new message from S->M should be generated, and then in message selection phase the two messages will be combined as a relay message. ---------------------------------------------------------------- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
