Skip to main content

Autoscaling parameters

Feature Request

Hello, we find the Qovery autoscaling very limited. It only allows to change the CPU or memory (hpa.cpu.average_utilization_percent or hpa.memory.average_utilization_percent) to start scaling the instances/pods.

Qovery checks those parameters every 5 minutes.

We find it not enough. With just adding a couple more of parameters it would be competent autoscaler.

How difficult would be allowing to define the poll period and the number of instances we want to scale each time?

Thank you.

Posted by Juan José Cermeño Portet ·
4 comments

Log in to comment and vote

Comments4

  • Alessandro Carrano

    Team•

    Nov 6, 2024

    Hey @Juan José Cermeño Portet ,

    regarding the current behaviour: we use the default horizontal autoscaling behaviour, the metrics are checked by Kubernetes every 15 seconds but the behaviour is different for upscale and downscale:

    • upscale: it doubles the instances immediately

    • downscale: it goes down to the minimum if the average cpu/memory are below the threshold for 5 minutes

    can you describe a little bit more your use case and what you are trying to achieve?

    • Juan José Cermeño Portet

      •

      Nov 7, 2024

      Thank you, your response made me go to check the default behaviour and now I understand why we were having a different experience and we could change our settings to improve it.

      Thank you.

      • Alessandro Carrano

        Team•

        Nov 8, 2024

        Nice! just for my understanding, what kind of settings are you going to change?

        I’m going to keep this idea open in case other customers need something particular upscale/downscale.

        have a great day

        • Juan José Cermeño Portet

          •

          Nov 8, 2024

          we can just change memory percentage, which we don’t use, or the CPU percentage, that’s the one we are using. In this case, we lowered the CPU percentage so now it’s scaling much better. Thank you.