Scale a Cluster

Warning

SingleStore 9.0 gives you the opportunity to preview, evaluate, and provide feedback on new and upcoming features prior to their general availability. In the interim, SingleStore 8.9 is recommended for production workloads, which can later be upgraded to SingleStore 9.0.

Horizontal Scaling

Scaling a cluster up or down horizontally means increasing or decreasing the number of nodes available. The number of child aggregator and leaf nodes can be adjusted independently.

To adjust the number (count) of aggregators, edit the spec.aggregatorSpec.count field in the sdb-cluster.yaml file.

The value of this field is the sum of the number of Master Aggregators and child aggregators. There is only one Master Aggregator, so the number of child aggregators are always spec.aggregatorSpec.count - 1. Only the Master Aggregator will be deployed when this field is set to 1.

To adjust the number of leaf nodes, edit the spec.leafSpec.count field in the sdb-cluster.yaml file. This value will be the number of leaf nodes in one availability group (AG).

Vertical Scaling

Scaling a cluster up or down vertically means increasing or decreasing the resource capacity in each node. The following resources can be scaled up or down.

CPU

The CPU can be scaled up or scaled down. CPU request and CPU limit are determined by the following formulas:

cpu_limit = spec.[aggregatorSpec/leafSpec].height * 8
- Where spec.[aggregatorSpec/leafSpec].height is a float that can be specified in the sdb-cluster.yaml file. Refer to the sdb-cluster.yaml file for an explanation of height.
cpu_request = cpu_limit

Both the cpu_limit and cpu_request will be passed into the node pod’s spec.

spec.containers[“node”].resources.limits.cpu
spec.containers[“node”].resources.requests.cpu

Memory

Memory can be scaled up or scaled down. Memory request and memory limit are determined by the following formulas:

memory_limit = spec.[aggregatorSpec/leafSpec].height * 32
- Where spec.[aggregatorSpec/leafSpec].height is the same height as for CPU.
memory_request = memory_limit

Both memory_limit and memory_request will be passed into the node pod’s spec.

spec.containers[“node”].resources.limits.memory
spec.containers[“node”].resources.requests.memory

Storage

Storage can only be scaled up, and scaling up is only possible if the defined StorageClass has allowVolumeExpansion set to true.

The following fields in sdb-cluster.yaml control the persisted volume size for each pod, which must be expressed as integer values:

spec.aggregatorSpec.storageGB
spec.leafSpec.storageGB

To scale up storage, ensure that there is available storage space and increase these values to the desired amount.
Apply the new values.
Shell
```
kubectl apply -f sdb-cluster.yaml
```