Data loss during detection period

[superchee0526]

I have a question about the master/slave switch process.

During the controller’s detection period, the old master can still accept write requests, but it cannot replicate the data to the slaves. After the failure is detected, one of the slaves is promoted to master. Even the old master later rejoins the cluster, the data written during that detection period is lost.

Is my understanding correct?

[git-hulk]

@superchee0526

Is my understanding correct?

Yes, that's correct for the current implementation. But we can mitigate this issue after introducing the WAIT command in the next release.

[superchee0526]

Thank you!