Reserved Nodes ​

Sometimes certain server instance types in public clouds make take time to acquire; creating a new compute cluster or altering an existing one to add nodes can take a long time, and ultimately fail if instances of the requested type are unavailable.

Reserved Nodes is functionality to allow Yellowbrick to acquire server instances in the background. Reserved Nodes are shared among all the compute clusters that are created for a single Yellowbrick instance, and they are immediately available for use when those clusters start running and doing work. You'll need to pay your cloud vendor for these reserved nodes, but Yellowbrick won't bill you for them unless they are part of a running compute cluster. This approach to reserving nodes is driven by Kubernetes auto-scaling, which results in the cloud vendor spinning up nodes even before any compute clusters have been created.

An alternate approach to solving capacity problems is to use on-demand capacity reservations (ODCRs) available from the cloud vendor. ODCRs typically belong to an account and have a global application, whereas Reserved Nodes are specific to the Yellowbrick platform. While Reserved Nodes don't require any extra IAM role permissions from the cloud vendor, ODCRs may be easier to automate using standard tooling, for example to acquire extra nodes ahead of an end-of-quarter reporting period.

To work with ODCRs, follow the cloud vendor's documentation.

To work with Reserved Nodes, in Yellowbrick Manager navigate to Configuration → Instance Management → Reservations.

The Reserved field specifies the number of reserved nodes requested. If empty, no background reservations are acquired. The Limit field sets a maximum number of compute cluster nodes the instance in question can consume in total. The limit per instance must be greater than or equal to the number of reserved nodes.