Sizing

The following sizes for Zenko instances have been tested on live systems using MetalK8s, which adds some overhead. If you are running a different Kubernetes engine, fewer resources may be required, but such configurations remain to be tested.

Reserve the following resources for each node.

  • Cores per server

    • 24 vCPU minimum
    • 32 vCPU medium load
    • 58 vCPU heavy load

  • RAM per server

    • 32 GB minimum
    • 64 GB medium load
    • 96 GB heavy load

  • Additional resources

    • 120 GB SSD (boot drive)
    • 80 GB SSD

  • Storage

    • 1 TB persistent volume per node minimum

      Note

      This requirement is for storage, not for the system device. Storage requirements depend on the sizing of different components and anticipated use. You may have to attach a separate storage volume to each cloud server instance.

All servers must run CentOS 7.4 or later, and must be ssh-accessible.

Custom Sizing

The default persistent volume sizing requirements are sufficient for a conventional deployment. Your requirements may vary based on total data managed and total number of objects managed.

Important

Persistent volume storage must match or exceed the maximum anticipated demand. Once set, the cluster cannot be resized without redefining new volumes.

To size your deployment, review the default values in Zenko/kubernetes/zenko/values.yaml. This file reserves space for each component in the build. This is the baseline build, which Helm will install unless instructed otherwise.

Next, review the values discussed in Zenko/kubernetes/zenko/storage.yaml. The storage.yaml file contains sizing instructions and settings that, when specified in a Helm installation, override the default values expressed in the values.yaml file. To override default values using storage.yaml, use the following addendum to the helm install invocation at Zenko deployment.

$ helm install [other options] -f storage.yaml

How much persistent volume space is required is calculable based on total data managed, total objects managed, and other factors. See storage.yaml for details.