backends.http
http
The HTTP backend enables Radiator to make HTTP requests to external web services for authentication and authorization purposes.
HTTP operations
Define named get, post, put, patch, or delete operations with a path
relative to the backend url. Use bindings to send request values. Use
mapping to assign response fields to execution context variables. Both blocks
are optional. mapping head reads the HTTP status code or response headers.
http "USERS" {
url "https://api.example.com";
get "FIND_USER" {
path "/users/%{aaa.identity}";
mapping json {
user.username = "$.username";
}
}
}
Multiple servers
Define named server blocks instead of
one backend-level url to share queries and authentication across endpoints.
Use server_selection to
choose how requests are distributed across servers.
Health Monitoring
HTTP backends have no SLO by default. Add a service_level_objective block to
enable automatic circuit breaking and exponential backoff.
When an explicit service_level_objective block is present, omitted fields use the
standard defaults (failure-rate 3/5, initial_backoff_period 3s,
max_backoff_period 30s, recovery_probe_count 2). See
Service Level Objective for details.
Only infrastructure-level HTTP responses count as failures toward the SLO:
- 5xx - server-side errors (the backend is crashing or misconfigured)
- 429 Too Many Requests - the backend is overloaded or rate-limiting
Application-level responses such as 401 Unauthorized, 403 Forbidden, and
404 Not Found are valid AAA outcomes and do not affect the failure rate.
Request retries
Generic HTTP backends do not retry by default. Use the
retry block to configure additional
attempts, their delay, and the timeout, request error, or HTTP status conditions
that warrant a retry.
Persistent connections
HTTP backends reuse persistent connections automatically. Set
idle_timeout to control how
long unused connections stay open for reuse. The default is 90s. Use the
keepalive block to tune the TCP
keepalive probes that detect broken connections.