Ocean by Spot continuously makes sure that all pods’ requirements are met so they can be scheduled by Kubernetes on the right nodes, with intelligent bin packing for optimized resource usage.
In some use cases there are instance level dependencies, such as:
- A pool of IP addresses available for use
- Relatively expensive instances that are able to facilitate a variety of pods, but must facilitate specific pods first.
- Misbehaving applications requesting new instances by respawning on failure and increasing infrastructure cost
To ensure that these instance level dependencies are met, we are pleased to share that Ocean launch specifications now supports a maximum number of instances allowed to run concurrently.
Now Ocean can operate within a framework to properly optimize cluster allocation with the cluster admin having a governance mechanism for better control over any unintended expansion of cluster infrastructure.
This all can be addressed by the new ability of configuring
resourceLimits.maxInstanceCount within a launch specification.
For more details about the API to activate this feature, please see the relevant Spot documentation.