lordcheng10 edited a comment on pull request #14970: URL: https://github.com/apache/pulsar/pull/14970#issuecomment-1084941134
In fact, the problem here may be the problem of consumer equality, similar to the previous producer problem. When there are duplicate consumers in the consumerList, it is possible to remove the actually successfully registered consumers when removing them. In fact, in our scenario, we also found a problem: when the consumer reconnects, although we can see that the number of consumers has not changed through monitoring, the broker no longer pushes messages to these consumers. Through the stats command, we see these The consumer's permit is 0, and it is a consumer that has been removed from the consumerSet and consumers. <img width="787" alt="image" src="https://user-images.githubusercontent.com/19296967/161125305-faacaad2-7b7c-4d86-8164-85514603355b.png"> <img width="1335" alt="image" src="https://user-images.githubusercontent.com/19296967/161125415-3172dbf7-3106-4fbf-a801-0d80e8108362.png"> <img width="1547" alt="image" src="https://user-images.githubusercontent.com/19296967/161037021-eea6e429-e38d-4baa-adf7-07e80dce4e28.png"> So the fundamental problem here is actually the definition of consumer equality. I think we might need to do something like https://github.com/apache/pulsar/pull/12846 on the definition of equality for consumers. I see concerns about the definition of consumer equality in this PR ,and this worry seems right: <img width="944" alt="image" src="https://user-images.githubusercontent.com/19296967/161124672-1c45b767-f4af-42f9-aa77-40707a008818.png"> -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
