yuexihan opened a new issue, #11583:
URL: https://github.com/apache/inlong/issues/11583

   ### What happened
   
   consumption suddenly stop
   <img width="1866" alt="企业微信截图_17331083438836" 
src="https://github.com/user-attachments/assets/f721a8b2-33d3-4405-9551-c3a382de9369";>
   
   
   ### What you expected to happen
   
   1. If the old code fails to register to Master again, there will be no 
further actions. It needs to be changed to continue retrying indefinitely upon 
failure.
   <img width="716" alt="企业微信截图_17334052584263" 
src="https://github.com/user-attachments/assets/2227aa5c-4449-4d0a-88e2-1926e9508490";>
   
   2. OfferEventAndNotify is responsible for receiving rebalance events, while 
OfferEventResult handles the results of those events. Therefore, the former 
should definitely not operate on rebalanceResults, nor should it reprocess the 
header of rebalanceResults as if it were a newly received event. Otherwise, if 
the result of the previous event has not been reported yet when the second 
event arrives, the subsequent processRebalanceEvent will end up processing the 
first event again, causing the second received event to be discarded.
   <img width="684" alt="企业微信截图_1733405369322" 
src="https://github.com/user-attachments/assets/981fceea-28cc-4371-a360-edcd00fc7d58";>
   
   3. ip_selector directly returns the domain name, while the connection pool 
maintains one connection per address, and it will not reconnect as long as the 
packet sending and receiving are normal. If there is a cluster switch, as long 
as the old master is sending and receiving packets normally, even if the 
registration fails, the TCP connection will not reconnect. Therefore, it is 
necessary to resolve the domain name into an IP address before returning it. 
During a cluster switch, a new IP can be returned by re-selecting. 
   <img width="537" alt="企业微信截图_17334056241261" 
src="https://github.com/user-attachments/assets/05bfeeb0-847c-42e0-b417-7928b4e9e83d";>
   
   4. Here, there is no check to see if rsp's success is true. If a cluster 
change causes the master to no longer recognize the client, it will continue to 
send heartbeats without re-registering.
   <img width="697" alt="企业微信截图_17334061728723" 
src="https://github.com/user-attachments/assets/097f42f7-0147-49bb-8525-e1c6f3dff393";>
   
   
   
   
   ### How to reproduce
   
   when master changes
   
   ### Environment
   
   _No response_
   
   ### InLong version
   
   master
   
   ### InLong Component
   
   InLong TubeMQ
   
   ### Are you willing to submit PR?
   
   - [X] Yes, I am willing to submit a PR!
   
   ### Code of Conduct
   
   - [X] I agree to follow this project's [Code of 
Conduct](https://www.apache.org/foundation/policies/conduct)
   


-- 
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.

To unsubscribe, e-mail: dev-unsubscr...@inlong.apache.org.apache.org

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org

Reply via email to