Skip to main content
Removed the need for apologies
Source Link
Bruce Becker
  • 3.7k
  • 4
  • 19
  • 41

Apologies for the formatting, typing from mobile is a pain.

  1. I do not see any obstacles. All communication is done through the apiserver over https. 2. 
  2. Multiple replicas of the controller and scheduler are allowed, leader election is built in. All you need is load balancing in front of the apiservers. 3. 
  3. I find this to work best, since usually the failure domain is a node, rather than a component. 4. 
  4. Yes.

Apologies for the formatting, typing from mobile is a pain.

  1. I do not see any obstacles. All communication is done through the apiserver over https. 2. Multiple replicas of the controller and scheduler are allowed, leader election is built in. All you need is load balancing in front of the apiservers. 3. I find this to work best, since usually the failure domain is a node, rather than a component. 4. Yes.
  1. I do not see any obstacles. All communication is done through the apiserver over https. 
  2. Multiple replicas of the controller and scheduler are allowed, leader election is built in. All you need is load balancing in front of the apiservers. 
  3. I find this to work best, since usually the failure domain is a node, rather than a component. 
  4. Yes.
Source Link
LLlAMnYP
  • 285
  • 1
  • 9

Apologies for the formatting, typing from mobile is a pain.

  1. I do not see any obstacles. All communication is done through the apiserver over https. 2. Multiple replicas of the controller and scheduler are allowed, leader election is built in. All you need is load balancing in front of the apiservers. 3. I find this to work best, since usually the failure domain is a node, rather than a component. 4. Yes.