Skip to content

Single Point of Failure

A single point of failure is one person, account, system, or component whose loss or failure can prevent an important service from continuing.

A single point of failure is one system, account, person, or component whose failure can prevent an important service from continuing.

The concept applies to people and operational knowledge as well as technical infrastructure.

Technical Examples

A website might have a single point of failure when:

  • One server carries all critical traffic
  • One database has no recovery path
  • One DNS account controls everything
  • One external API is required for checkout
  • One integration holds the only copy of important data

Cloudflare describes the infrastructure problem directly: if all traffic depends on a single origin server, failure or overload of that origin can take the application offline. Cloudflare: Balancing traffic across origins

Human Single Points of Failure

The same idea applies to knowledge and access.

Examples include:

Only Alex knows how deployments work.
Only the former developer has registrar access.
Only one employee has the two-factor-authentication device.
The integration is undocumented and only its original author understands it.

These may be more common on small-business websites than redundant-server failures.

Why It Matters

A single point of failure concentrates risk.

If the critical dependency disappears, the organization may not have another path forward.

The impact might be:

  • Website outage
  • Lost access
  • Inability to deploy
  • Payment failure
  • Lost data
  • Expensive emergency redevelopment

Not Every Single Dependency Needs Redundancy

Redundancy has its own cost.

A small brochure website does not necessarily need:

  • Multiple origin servers
  • Automated geographic failover
  • Several DNS providers

The risk should justify the complexity.

But one organization-controlled account plus documented recovery information is usually inexpensive.

Common Problems

Common problems include:

How to Improve

Identify the things whose loss would seriously interrupt operations.

For each one, ask:

What happens if this disappears tomorrow?

Possible responses include:

  • Add another authorized owner
  • Create Backup & Restore
  • Document the system
  • Export important data
  • Add monitoring
  • Replace unsupported Dependency
  • Establish a recovery procedure