Equinix Metal™ offers single-tenant, dedicated servers in its cloud, offering a curated set of server configurations that are perfect for common use-cases.
If hardware fails with any of these standard public cloud configurations, our SLA requires us to give you another machine. Equinix Metal does not keep an agent on a machine once it is provisioned, so customers are responsible for monitoring the health of their hardware, such as disk drives. If hardware fails or becomes problematic, customers should notify email@example.com and request a new machine. Equinix Metal does not provide a migration service for your data.
Equinix Metal™ also does not provide load-balancing or high-availability configurations, relying on customers provision and configure their servers in a way that provides the right level of redundancy, capacity, backups, and disaster recovery planning for their use-case.
When planning for resilience, Equinix Metal deploys a pair of switches per rack so no one rack has a single point of failure. Depending on a variety of factors — including the size of each of our global facilities — servers may be available in multiple racks. For further redundancy within a data center, you may want to spread your servers over multiple racks.
Although users are not able to query rack information in advance of a deploy, we do provide a Switch ID hash (in the Portal or via our API) as part of the details for any already-deployed instance. Each Switch ID hash represents a switch pair, which allows you to understand the diversity of your infrastructure.
If you are looking to achieve a certain diversity (or concentration) of infrastructure, these are your options:
- Provision on demand servers. You can destroy and then provision another available server to achieve a particular setup. We cannot always provide the diversity.
- Reserve hardware in advance. Equinix Metal can work with you to provide the needed setup, pending availability.
For specialized workloads, Equinix Metal can provide customized server configurations on a contract basis. Interacting with custom configurations is similar to our public cloud, but inventory is more limited. If there is a problem with custom hardware, users can either leverage on-demand servers or tap into dedicated “spares” — extra machines that are reserved for just this purpose.
No matter the use case, we highly encourage a thoughtful approach to high-availability. Sparing is an excellent tool and we would be happy to discuss it with you.