Applies to: DynamoDb, Amazon RDS MSSQL, Amazon RDS PostgreSQL
Auto-scaling enhances efficiency and performance, especially for large-scale operations, by dynamically deploying access nodes in a region only when required, and then promptly powering off and decommissioning the access nodes after they are used.
Supported AWS regions
Auto-scaling is supported for all commercial AWS Regions. Edge locations (such as GovCloud, AWS Outposts, Local Zones, and Wavelength Zones) are not supported. Temporary access nodes are provisioned within the AWS VPC configured in the scale configuration.
How auto-scaling works with Amazon Workloads
Auto-scaling is managed by a scale configuration that is associated with a Server group. The Server group is associated with a workload resource pool in a specific AWS region. The Server group can be associated with a workload instance to enable auto-scaling in the AWS account specified in the scale configuration.
Based on workload demand, Commvault automatically provisions the required access nodes from this resource pool to efficiently support backup and restore operations.
It uses the auto-scaling framework to provision temporary access nodes on demand for export-based backup and restore operations. Browse and detect operations continue to use a permanent access node.
Where auto-scaled access nodes are created
Auto-scaled access nodes are created in the AWS regions and availability zones that you specify in the scale configuration. The Scale Manager provisions temporary EC2 instances from Commvault-provided AMIs; these instances register with CommServe and are used for backup and restore operations.
The access nodes are deployed in the same region where your S3 buckets reside to optimize data transfer performance and minimize cross-region data transfer costs.
Decommissioning of access nodes
Auto-scaled access nodes terminate after they are idle for a short configurable period (5 minutes are considered as the idle timeout). If no active access node is available, auto-scale automatically creates a new access node on demand.
The temporary access nodes are automatically deprovisioned after the idle period regardless of whether the job succeeded or failed; if no activity occurs on an access node after the job completes, the access node is deleted.