What is hardware failure domain in OCI?
In Oracle Cloud Infrastructure (OCI), a Fault Domain is the fundamental unit of hardware failure isolation. While a Data Center (Availability Domain) is a massive building, OCI segments its internal guts into three distinct "logical data centers" called Fault Domains (FDs) to ensure that a hardware failure in one doesn't cascade to another.
Think of it as anti-affinity at the rack level.
A Fault Domain is a grouping of physical infrastructure that does not share single points of failure. Each FD is physically separated and includes its own:
Physical Server Racks: Instances in FD1 are on physically different racks than those in FD2.
Power Distribution Units (PDUs): Each FD has independent and redundant power paths. A "blown fuse" or PDU failure in FD1 will not cut power to FD2.
Top-of-Rack (ToR) Switches: Each FD uses its own set of network switches. This prevents a network card failure or switch crash in one domain from silencing your entire application.
The architecture is designed to protect your application from two specific types of events:
Hardware eventually dies. Whether it's a motherboard short-circuit, a power supply failure, or a localized cooling issue, FDs ensure the "blast radius" is limited. If you have a three-node cluster (one node in each FD), a catastrophic rack failure only takes out 33% of your capacity.
Oracle must periodically patch the underlying "off-box" software, update firmware on network switches, or service power systems.
The Rule: Oracle never performs maintenance on more than one Fault Domain in an Availability Domain at a time.
The Benefit: If your app is distributed across FDs, your services remain online even while Oracle is actively working on the underlying hardware of one domain.
To leverage this hardware isolation, you must design for it. In the OCI Console, when you launch a Bare Metal or VM instance, you have the option to manually select the Fault Domain.
| Scenario | Distribution Strategy |
| Web Tier | Deploy 3+ VMs, choosing a different FD for each. |
| Database (RAC) | Place Node 1 in FD1 and Node 2 in FD2. |
| HPC Clusters | Group instances in a single FD if ultra-low latency is required, but replicate the cluster in another FD for redundancy. |
Pro-Tip: If you don't select a Fault Domain during launch, OCI uses a "best-effort" algorithm to automatically spread your instances across the three domains to prevent them from landing on the same physical rack.
| Level | Scope of Protection | Protection Against... |
| Fault Domain | Rack / Row | Server crash, ToR switch failure, PDU failure. |
| Availability Domain | Data Center | Fire, flood, total building power loss. |
| Region | Geographic Area | Earthquake, large-scale power grid failure, regional conflict. |
"Fault Domains are OCI's way of giving you the benefits of three data centers for the price of one. By ensuring your instances 'don't live in the same neighborhood' within the data center, you eliminate the risk of a single physical failure taking down your entire stack."