Backing up and recovering your environment

This topic describes how the SC//Reliant platform backs up workloads and data, and how a node is recovered after a hardware failure.

Backup and recovery model

The SC//Reliant platform's backup and recovery model is built around the same declarative state that drives the rest of the platform: workloads and configuration are reproducible from source, full backups handle the application data, and bare-metal recovery is largely a matter of plugging in replacement hardware. The combination of declarative configuration, scheduled backups, and zero-touch reprovisioning is designed to minimize on-site IT effort when a node fails.

Scheduled backups run at the site-node level. Container backups are handled through the image registry plus persistent-volume snapshots; only changed layers are transmitted on update. Backup orchestration runs through the SC//Reliant Platform Manager, with schedules configurable per workload, per site, and per region.

Backup target topology

The SC//Reliant platform supports the following backup target patterns:

Local. On-node second drive or on-site NAS, for fast restore.
Cloud. S3, Azure Blob, or Google Cloud Storage in any region, including regions that meet specific data-residency requirements.
Tiered. Both local and cloud, combining fast on-site restore with offsite durability.

For sites with stricter operational standards, a customer-deployed backup container (for example, Veeam, Commvault, or Rubrik) can run as a workload on the SC//Reliant platform in place of the native backup orchestration.

Data deduplication

Backup deduplication is supported and depends on the chosen backup target. Container images naturally deduplicate via the layered registry; only changed layers move on update. VM image deduplication is supported when a deduplication-capable backup target is used: Veeam, Commvault, Rubrik, or similar deployed as a containerized workload on the SC//Reliant platform, or a deduplication-aware cloud backup target (AWS Backup, Azure Backup with dedup).

Bare-metal recovery

Bare-metal recovery is the workflow used when a node fails and is replaced. Replacement hardware is provisioned with the SC//Reliant platform base image (USB or network boot), registered in the SC//Reliant Platform Manager with the failed node's site identifier, and on first boot pulls the full declarative configuration. Workloads are restored from the latest backups.

Break-fix replacement can be handled two ways:

A pre-configured system is shipped by Scale Computing.
A bare-metal system is provisioned from the existing node in a two-node cluster.

Expected recovery time is minutes for software restoration, plus hardware-swap logistics. Because the software setup is zero-touch, the on-site visit is a short hardware swap and does not require an IT specialist for software configuration.

Data preserved on recovery includes all configuration and workloads declared in Puppet, plus any application data covered by the most recent backup.

Backup posture across site sizes

Where high availability and fault tolerance are required, a two-node HA cluster is the preferred deployment pattern. Because the SC//Reliant platform's hardware requirements are minimal, low-cost or composable hardware can be used for both nodes. The total cost of a two-node cluster is typically less than building high availability through a single system with redundant disks and power supplies.

Where a single node is the only option, the SC//Reliant platform is hardware-agnostic, so deployments can use systems that are:

fanless and entirely solid-state,
equipped with redundant power supply units (PSUs),
using error-correcting code (ECC) memory, and
configured with RAID for the storage subsystem.

In addition, the system can leverage external storage, for example, a low-cost NAS, for local, on-site, frequent backup and critical-file replication.

Critical resilience patterns stay intact even in single-node mode:

Payment store-and-forward continuity. No transaction loss during an outage.
Local workload autonomy. No Wide Area Network (WAN) dependency at runtime.
Rapid bare-metal recovery. Replacement hardware re-provisions in minutes via the SC//Reliant Platform Manager zero-touch workflow.