Merge !1591: modules/stats add answer.stale

NEWS

@@ -12,6 +12,7 @@ Improvements
 
 - answer NOTIMPL for meta-types and non-IN RR classes (!1589)
 - views: improve interaction with old-style policies (!1576)
+- stats: add stale answer counter 'answer.stale' (!1591)
 
 Bugfixes
 --------

daemon/lua/kres-gen-33.lua

@@ -248,6 +248,7 @@ struct kr_request {
 	ranked_rr_array_t add_selected;
 	_Bool answ_validated;
 	_Bool auth_validated;
+	_Bool stale_accounted;
 	uint8_t rank;
 	struct kr_rplan rplan;
 	trace_log_f trace_log;

lib/resolve.h

@@ -260,6 +260,7 @@ struct kr_request {
 	ranked_rr_array_t add_selected;
 	bool answ_validated; /**< internal to validator; beware of caching, etc. */
 	bool auth_validated; /**< see answ_validated ^^ ; TODO */
+	bool stale_accounted;
 
 	/** Overall rank for the request.
 	 *

manager/knot_resolver_manager/statistics.py

@@ -119,6 +119,12 @@ def _parse_resolver_metrics(instance_id: "KresID", metrics: Any) -> Generator[Me
             label=("instance_id", sid),
             value=metrics["answer"]["cached"],
         )
+        yield _counter(
+            "resolver_answer_stale",
+            "number of queries that utilized stale data",
+            label=("instance_id", sid),
+            value=metrics["answer"]["stale"],
+        )
         yield _counter(
             "resolver_answer_rcode_noerror",
             "number of NOERROR answers",

modules/serve_stale/serve_stale.lua

@@ -8,9 +8,10 @@ local ffi = require('ffi')
 M.timeout = 3*sec
 
 M.callback = ffi.cast("kr_stale_cb",
-	function (ttl) --, name, type, qry)
+	function (ttl, _, _, qry)
 		--log_debug(ffi.C.SRVSTALE, '   => called back with TTL: ' .. tostring(ttl))
 		if ttl + 3600 * 24 > 0 then -- at most one day stale
+			qry.request.stale_accounted = true
 			return 1
 		else
 			return -1
@@ -27,7 +28,9 @@ M.layer = {
 		local now = ffi.C.kr_now()
 		local deadline = qry.creation_time_mono + M.timeout
 		if now > deadline or qry.flags.NO_NS_FOUND then
-			log_debug(ffi.C.LOG_GRP_SRVSTALE, '   => no reachable NS, using stale data')
+			log_qry(qry, ffi.C.LOG_GRP_SRVSTALE,
+				'   => no reachable NS, using stale data "%s"',
+				kres.dname2str(qry:name()))
 			qry.stale_cb = M.callback
 			-- TODO: probably start the same request that doesn't stale-serve,
 			-- but first we need some detection of non-interactive / internal requests.

modules/stats/README.rst

@@ -55,6 +55,8 @@ Built-in counters keep track of number of queries and answers matching specific
 +-----------------+----------------------------------+
 | answer.cached   | queries answered from cache      |
 +-----------------+----------------------------------+
+| answer.stale    | queries that utilized stale data |
++-----------------+----------------------------------+
 
 +-----------------+----------------------------------+
 | **Answers categorized by RCODE**                   |

modules/stats/stats.c

@@ -37,12 +37,17 @@
  #define UPSTREAMS_COUNT  512 /* Size of recent upstreams */
 #endif
 
-/** @cond internal Fixed-size map of predefined metrics. */
+/** @cond internal Fixed-size map of predefined metrics.
+ *
+ * When changing the list, don't forget _parse_resolver_metrics()
+ * in ../../manager/knot_resolver_manager/statistics.py
+ */
 #define CONST_METRICS(X) \
 	X(answer,total) X(answer,noerror) X(answer,nodata) X(answer,nxdomain) X(answer,servfail) \
 	X(answer,cached) X(answer,1ms) X(answer,10ms) X(answer,50ms) X(answer,100ms) \
 	X(answer,250ms) X(answer,500ms) X(answer,1000ms) X(answer,1500ms) X(answer,slow) \
 	X(answer,sum_ms) \
+	X(answer,stale) \
 	X(answer,aa) X(answer,tc) X(answer,rd) X(answer,ra) X(answer, ad) X(answer,cd) \
 	X(answer,edns0) X(answer,do) \
 	X(query,edns) X(query,dnssec) \
@@ -303,6 +308,7 @@ static int collect(kr_layer_t *ctx)
 		DEPRECATED
 		use new names metric_answer_edns0 and metric_answer_do
 	*/
+	stat_const_add(data, metric_answer_stale, param->stale_accounted);
 	stat_const_add(data, metric_query_edns, knot_pkt_has_edns(param->answer));
 	stat_const_add(data, metric_query_dnssec, knot_pkt_has_dnssec(param->answer));