Skip to main content

Autoscaling

· One min read
Thibaut Raballand
Creator of Spider

Kubernetes brings autoscaling, and it's magic!

That's new​

On Swarm, you had to scale the service to what you estimate was needed.
And you let them bet.

In Kube, you have Horizontal Pod Autoscaling.
And that's magic:

  • Configure them on your Pod deployments
  • Set the metric to check (CPU / RAM or custom)
  • Set the threshold for scaling

And that's over!

It takes only a couple of hours to set the configuration.

Result​

Now, Spider autoscales all scalable services, and this is magic!

Check this:

Autoscaling.png

Everytime a layer changes color, Kube adjusted the number of replicas. We even had a peak at 20 replicas of Web-Write to absorb a performance test round 12:45!

Upload.png

Nice :-)