Azure#

In certain cases, it might be helpful to ‘scale up’ a node group in a cluster before an event, to test cloud provider quotas or to make user server startup faster.

Azure Console UI instructions#

Scaling up a node pool#

  1. Login to https://portal.azure.com using the appropriate cluster credentials

  2. Search the Kubernetes service section for the appropriate cluster

  3. Click on the cluster name, then go to Settings and select the Node pools option

  4. Click on the appropriate node pool that you would like to scale up

  5. Then select the Scale node pool option from the top of the page

  6. A new window should pop up that looks like this

    ../../../_images/azure-scale-node-pool-window.png

    Scale node pool#

  7. If the Autoscale option is selected like in the screenshot above (the default, recommended option), then in order to scale up the node pool to an exact number of nodes, temporarily deactivate the autoscaler, by selecting the Manual option, introduce the desired number of nodes then click Apply.

  8. After the Apply succeded, you should see the new nodes coming up. You can then click on Scale node pool option again, enable the Autoscale, and set the Min number of nodes to the desired one the you set in the step before.

Warning

Don’t forget to turn the autoscaler back on after the manual modification of the node pool size! This is really important, otherwise a scale up from the max manual limit, won’t be able to happen automatically, and the hub won’t be able to spawn new user servers.

Scaling down a node pool#

  1. Follow the first six steps in the scaling up guide above, until you get to the Scale node pool window.

  2. This time, do not activate the Manual mode, and just adjust the Min number of nodes for the autoscaler. As users stop their servers after the event, eventually a scale down event will be triggered, and the autoscaler will adjust the node pool size according to the limits that are set.

Note

The cluster autoscaler doesn’t enforce the node pool size after updating the Min or Max counts. The limits will be taken into accounts for future scaling decisions. A new scaling decision happens after a scale up/down event is triggered.

So, because we usually want nodes to be ready and waiting before an event, and not wait for a scale up/down event, we need to temporarily disable the autoscaler.

More about the Azure autoscaler in the docs here.

Terraform instructions#

TODO