grave-v4.vcl

sub vcl_hit {
    if (obj.ttl >= 0s) {
        # normal hit
        return (deliver);
    }
    # We have no fresh fish. Lets look at the stale ones.
    if (std.healthy(req.backend_hint)) {
        # Backend is healthy. Limit age to 10s.
        if (obj.ttl + 10s > 0s) {
            set req.http.grace = "normal(limited)";
            return (deliver);
        } else {
            # No candidate for grace. Fetch a fresh object.
            return(fetch);
        }
    } else {
        # backend is sick - use full grace
        if (obj.ttl + obj.grace > 0s) {
            set req.http.grace = "full";
            return (deliver);
        } else {
            # no graced object.
            return (fetch);
        }
    }
}

sub vcl_backend_response {
    set beresp.ttl = 10s;
    set beresp.grace = 1h;
}

sub vcl_recv {
    # intial state
    set req.http.grace = "none";
}

sub vcl_deliver {
    # copy to resp so we can tell from the outside.
    set resp.http.grace = req.http.grace;
}

Does this configuration still honor TTLs sent from backend responses with a Cache-Control header for something other than 10 seconds?

@cletustboone I think you should remove the set beresp.ttl line to honor the origin's Cache-Control max-age/s-maxage headers.

Disclaimer: Newbie to Varnish.
Question: Shouldn't vcl_recv finish with a return(hash);?

return (fetch); should be replaced with return (miss); if using Varnish 5.0

@arikogan return(hash) is called at the end of the default vcl_recv function and is not needed in this vcl, unless you want to prevent the default vcl_recv behavior for some reason

@dancriel I'm running Varnish 5.0.0 and without return (hash) at the end of my altered vcl_recv, Varnish won't return cached results. When calling /etc/init.d/varnish configtest varnish returns the built-in configuration as well which includes this note:

NB! You do NOT need to copy & paste all of these functions into your"
" * own vcl code, if you do not provide a definition of one of these"
" * functions, the compiler will automatically fall back to the default"
" * code from this file."

This sounds like "as long as you don't define vcl_recv the built-in version will be used but if you redefine it, you should copy 'n paste it and add your custom code".

@ureimers My understanding is that the default behavior will run after your custom behavior, unless you return somewhere in your vcl_recv function, which will prevent the default behavior (not recommended). Hopefully someone can correct me if this is wrong.

Is this working on Varnish 6? I never hit the req.http.grace = "full"; line. After 10 seconds of the backend going offline, I get a 503 Backend fetch failed