Whether request handling went through
Request handling will go through
vcl_error and a response will be generated
locally when the VCL executes the
error action or when Varnish encounters a
problem while handling the request, such as a communication failure with the
backend. This VCL variable allows you to distinguish locally generated error
responses from error responses received from upstream (from the origin or
from a shield pop).
All synthetic responses are generated in
vcl_error, not only error responses.
If you use
vcl_error to generate synthetic responses in non-error cases,
you might need to check
resp.status in combination with
resp.is_locally_generated to determine that a response is a locally
generated error response.