I have 20 servers, with 10 nodes in each datacenter, A and B. each node is having label.
During maintenance whenever we bring down two nodes from datacenter A, most of the pods will start running in the nodes from other datacenter B, as per kubernetes placement policy. Once we finish with maintenance, we have to manually restart the pods such that pods start running in datacenter A, so that numbers of pods are balanced accross datacetner A and B.
Is it possible to start pods in other three nodes from same datacenter A, if they have enough resource to run it. And only run pods on nodes from DataCenter B if nodes from A lack resources.
Do we have to implement antiaffinity rule for this? or we will have to split zone? What is the best way to implment it such that we dont have to do manual rebalancing.
Can pods be rebalance accross the nodes by itself without manual restart?