yangw

Hi @tangweichun , It seems timeout when catch up relay log after 10s. You can set ReasonableReplicationLagSeconds greater than 10s(if not set, it default to 10s).

> But the SQL thread is actually stopped by the orchestrator and needs to be manually opened to continue. Why orchestrator stop the promoted instance's SQL thread? is there any...

Is there any reproduce steps? In my opinion, the related user and password have passed to init container actually. Here is code: https://github.com/bitpoke/mysql-operator/blob/b7ea748da61515346b69d7d8314be8e672b5c915/pkg/controller/mysqlcluster/internal/syncer/statefullset.go#L618-L626 From above, the ***init*** and ***sidecar*** container...

Maybe set cluster.Spec.RcloneExtraArgs=[]string{"--progress"} will take effect.

PTAL, @cndoit18

@cndoit18 ready to merge, please

maybe one case: https://ci.bitpoke.io/bitpoke/mysql-operator/319/1/6 ``` • Failure [0.164 seconds] MySQL user controller when creating the MysqlUser without permissions [It] reconciles the mysql user, but not before the finalizer has been...

Nice catch! May be the first method is better, because it's better to unit test and it's more consistent with already code.

Maybe we should watch the related resource, if any changes on it, then we reconcile it.

We could use hostnetwork as another lightly solution. We could add hostnetwork && port to crd.