[Bug]When a single computer room is isolated from the network, it may cause message loss.

[iamqq23ue]

Before Creating the Bug Report

• I found a bug, not just asking a question, which should be created in GitHub Discussions.

• I have searched the GitHub Issues and GitHub Discussions of this repository and believe that this is not a duplicate.

• I have confirmed that this bug belongs to the current repository, not other repositories of RocketMQ.

Runtime platform environment

suse linux

RocketMQ version

5.1.1

JDK Version

1.8

Describe the Bug

The architecture consists of three data centers, three brokers, one master, and two slaves. Three controllers are also one for each data center (the same as the nameserver).

Simulate the isolation of the broker master and one controller network from the brokers and controllers of the other two data centers.

During network failures, it can be observed that two broker masters will appear, and when the network is restored, the original master will become a slave.

However, messages sent to the original master during network failures will be lost.

Steps to Reproduce

1,Simulate the isolation of the broker master and one controller network from the brokers and controllers of the other two data centers.
2,send message to the original master

What Did You Expect to See?

Sending a message to the original master that has been quarantined should result in an error

What Did You See Instead?

Sending a message to the original master that has been quarantined did not report an error

Additional Context

No response

[RongtongJin]

Hi @iamqq23ue, I have previously simulated a similar scenario, but it seems that this situation did not occur (assuming that allAckInSyncStateSet is enabled). This is because the SyncStateSet on the original master broker cannot be updated at this time, and messages sent to this master broker will fail to be send successfully.

[RongtongJin]

Additionally, it is suggested to ask such questions on Github Discussion.

[iamqq23ue]

Thank you for your reply. I will test it.

[iamqq23ue]

hi @RongtongJin ,when allAckInSyncStateSet is true,tps will significantly decrease.
in my case,tps dropped from 30000 to 3000.

I only have one slave.

[iamqq23ue]

The performance issue has been resolved in the latest version 5.1.3