CODEWITHSUNDEEP
amazon-web-servicesamazon-ec2amazon-elbautoscaling
red888
red888

Reputation: 31652

What happens if I disable AZRebalance and an AZ becomes unhealthy?

Curious about this quote:

If you suspend AZRebalance and a scale out or scale in event occurs, the scaling process still tries to balance the Availability Zones. For example, during scale out, it launches the instance in the Availability Zone with the fewest instances.

What happens if an AZ is unhealthy in this case? Will it still failover or will it still try to launch in that region because I've disabled AZRebalance?

Upvotes: 1

Views: 5874

Answers (1)

Michael - sqlbot
Michael - sqlbot

Reputation: 179344

After certain actions occur, your Auto Scaling group can become unbalanced between Availability Zones. Auto Scaling compensates by rebalancing the Availability Zones.

https://docs.aws.amazon.com/autoscaling/ec2/userguide/auto-scaling-benefits.html#AutoScalingBehavior.InstanceUsage

When an AZ is in an unhealthy state, or otherwise has no capacity, it isn't factored into the "unbalanced" calculation -- unbalanced refers to a condition where the workload isn't distributed evenly among healthy AZs with adequate capacity.

Note the word after in the quote. In your scenario, AZRebalance refers something that happens after recovery, not during an event.

If an AZ is unhealthy, replacement instances will be launched in a different AZ, regardless of the setting.

Later, when the AZ returns to a healthy status, the system will try to migrate some of the instances back into the previously-unhealthy AZ to spread your workload back out... but without AZRebalance, the system won't try to even out the workload by launching new instances into the now-healthy AZ and then terminating some of the instances elsewhere -- it will just remain as it was, unbalanced, until other routine or recovery scaling events rebalance things over time.

Think of this as enabling a strictly proactive "best practice" rebalancing strategy that rebalances the instances just for the purpose of keeping the workload even across AZs -- which of course is where you want to be in a normal, healthy environment so that a subsequent problem will disrupt the minimum number of instances possible -- it does not impact the purposes of scaling in due to demand or to recover from the loss of some instances. Disabling it doesn't impact those recovery actions.

Upvotes: 4

Related Questions