Fastly Compute Backend Production Setup

Problem

Backends defined in fastly.toml work during local development with fastly compute serve but return "Requested backend named 'X' does not exist" errors in production after deployment with fastly compute publish.

Context / Trigger Conditions

• Error: Requested backend named 'funnelcake' does not exist
• Error: TLS alert received (alert_id=0) when connecting to HTTPS backends
• Backend works with fastly compute serve locally
• fastly.toml has [[backends]] section properly configured
• Service was created before the backend was added to fastly.toml

Root Cause

The [[backends]] section in fastly.toml is only processed during initial service creation via the [setup] configuration. Once a service exists, changes to backends in fastly.toml are ignored - you must add/modify backends through the Fastly CLI or API.

From the Fastly CLI output:

INFO: Processing of the fastly.toml [setup] configuration happens only for a new service. Once a service is created, any further changes to the service or its resources must be made manually.

Solution

Step 1: Create the backend via CLI

fastly backend create \
  --service-id YOUR_SERVICE_ID \
  --version active \
  --autoclone \
  --name backend-name \
  --address api.example.com \
  --port 443 \
  --override-host api.example.com \
  --use-ssl \
  --ssl-cert-hostname api.example.com \
  --ssl-sni-hostname api.example.com

Key flags:

• --autoclone: Creates a new version automatically
• --ssl-cert-hostname: Required for SSL certificate validation
• --ssl-sni-hostname: Required for TLS SNI (fixes alert_id=0 errors)

Step 2: Activate the new version

fastly service-version activate --service-id YOUR_SERVICE_ID --version VERSION_NUMBER

Step 3: Verify backend exists

fastly backend list --service-id YOUR_SERVICE_ID --version active

Step 4: Redeploy your code

After adding the backend, redeploy to ensure the code and backend are in the same version:

fastly compute publish

Then verify backends still exist in the new version:

fastly backend list --service-id YOUR_SERVICE_ID --version active

Verification

1. Check logs with fastly log-tail --service-id YOUR_SERVICE_ID
2. Make a request that uses the backend
3. Confirm no "backend does not exist" or TLS errors in logs

Example

Given this fastly.toml (which works locally but not in production):

[[backends]]
  name = "funnelcake"
  address = "relay.dvines.org"
  port = 443
  override_host = "relay.dvines.org"
  use_ssl = true

Add the backend to an existing service:

# Create backend with SSL properly configured
fastly backend create \
  --service-id WfOrPTFYmwwxRvqrfralLA \
  --version active \
  --autoclone \
  --name funnelcake \
  --address relay.dvines.org \
  --port 443 \
  --override-host relay.dvines.org \
  --use-ssl \
  --ssl-cert-hostname relay.dvines.org \
  --ssl-sni-hostname relay.dvines.org

# Activate the new version
fastly service-version activate --service-id WfOrPTFYmwwxRvqrfralLA --version 49

# Redeploy code
fastly compute publish

Notes

• Each fastly compute publish creates a new version that inherits backends from the previous active version
• If you need different backends for different environments, consider using environment variables or separate services
• The --autoclone flag is essential - you cannot modify an active version directly
• Backend names in your code must exactly match the --name parameter

Related Errors

• Requested backend named 'X' does not exist - Backend not created in service
• TLS alert received (alert_id=0) - Missing SNI hostname configuration
• Mandatory SSL cert checks require specifying cert hostname - Missing ssl-cert-hostname

References

• Fastly Compute backends documentation
• Fastly CLI backend commands