Why are replicas schedule to other nodes when deployment replicas > 1 ? about hwameistor

[SSmallMonster]

This field has no meaning for the user and will only be used during the initialization and update configuration stages.

During the scheduling period, the nodes where all replicas with a Ready state are located can be used as the Primary node, which is different from the Primary node mentioned by the application layer(e.g. mysql). In the case of consistent data, the master node can completely drift.

[qilongqiu]

I don't think this is right, hwameistor should do this scheduling function

The drdb state of lvr is master-slave related

[SSmallMonster]

DRBD status is changed to Primary/Secondary after scheduling, when no workload is using this volume, the DRBD volume status also changed to Secondary/Secondary.

So, DRBD status is changing Primary/Secondary after scheduling rather scheduling acording to DRBD status - Secondary/Secondary.

[SSmallMonster]

After research, the scheduler may select a node which is Secondary when workload's replica >= 2.

This is a bug, the correct schedule result should be the same node which first scheduled.