The specification for fault injection introduced into traffic to test the
resiliency of clients to backend service failure. As part of fault
injection, when clients send requests to a backend service, delays can be
introduced by Loadbalancer on a percentage of requests before sending
those request to the backend service. Similarly requests from clients can
be aborted by the Loadbalancer for a percentage of requests.
timeout and retry_policy will be ignored by clients that are configured
with a fault_injection_policy.
Specifies the policy on how requests intended for the route's backends are
shadowed to a separate mirrored backend service. Loadbalancer does not
wait for responses from the shadow service. Prior to sending traffic to
the shadow service, the host / authority header is suffixed with -shadow.
Specifies the timeout for the selected route. Timeout is computed from the
time the request has been fully processed (i.e. end-of-stream) up until
the response has been completely processed. Timeout includes all retries.
If not specified, will use the largest timeout among all backend services
associated with the route.
A list of weighted backend services to send traffic to when a route match
occurs. The weights determine the fraction of traffic that flows to their
corresponding backend service. If all traffic needs to go to a single
backend service, there must be one weightedBackendService with weight set
to a non 0 number.
Once a backendService is identified and before forwarding the request to
the backend service, advanced routing actions like Url rewrites and header
transformations are applied depending on additional settings specified in