AutoScaling Azure Virtual Machine Scale Sets

Just deployed a test swarm using the Docker for Azure 1.12.0-beta12 ‘Deploy to Azure’ template. Right away I noticed that the number of VMs added to each scale set (Manager and Worker) is set upon deployment and new VMs can’t be added later on. (eg if I create my initial swarm and choose 5 VMs for my worker nodes, that is the most worker nodes I can have and can’t add more to the swarm later on). Is this the case? Is there no way to use the built in autoscaling features of Azure VM Scale Sets to automatically spin up more worker nodes if the aggregate cpu usage crosses a threshold?

Hi @kevinmcgarry, I think this is added now (in the latest template, azure-v1.13.0-1 at time of writing).

You should be able to increase (or even decrease) the number of worker nodes in the VM scale set manually with the provided slider bar in the Azure web console. Docker for Azure will take care of adding/removing them to the Swarm.

by default autoscaling is off

Auto scaling is indeed off by default for now. We will look into auto scaling in a future release. At the moment, you can manually increase (or decrease) the number of nodes in the VM Scale Set group for workers post deployment and the new nodes should come up and join the swarm: