summaryrefslogtreecommitdiffstats
path: root/modules/stats
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-27 10:41:58 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-27 10:41:58 +0000
commit1852910ef0fd7393da62b88aee66ee092208748e (patch)
treead3b659dbbe622b58a5bda4fe0b5e1d80eee9277 /modules/stats
parentInitial commit. (diff)
downloadknot-resolver-1852910ef0fd7393da62b88aee66ee092208748e.tar.xz
knot-resolver-1852910ef0fd7393da62b88aee66ee092208748e.zip
Adding upstream version 5.3.1.upstream/5.3.1upstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'modules/stats')
-rw-r--r--modules/stats/.packaging/test.config4
-rw-r--r--modules/stats/README.rst211
-rw-r--r--modules/stats/meson.build24
-rw-r--r--modules/stats/stats.c510
-rw-r--r--modules/stats/test.integr/deckard.yaml12
-rw-r--r--modules/stats/test.integr/kresd_config.j2111
-rw-r--r--modules/stats/test.integr/stats.rpl194
7 files changed, 1066 insertions, 0 deletions
diff --git a/modules/stats/.packaging/test.config b/modules/stats/.packaging/test.config
new file mode 100644
index 0000000..fd25460
--- /dev/null
+++ b/modules/stats/.packaging/test.config
@@ -0,0 +1,4 @@
+-- SPDX-License-Identifier: GPL-3.0-or-later
+modules.load('stats')
+assert(stats)
+quit()
diff --git a/modules/stats/README.rst b/modules/stats/README.rst
new file mode 100644
index 0000000..27992aa
--- /dev/null
+++ b/modules/stats/README.rst
@@ -0,0 +1,211 @@
+.. SPDX-License-Identifier: GPL-3.0-or-later
+
+.. _mod-stats:
+
+Statistics collector
+====================
+
+Module ``stats`` gathers various counters from the query resolution
+and server internals, and offers them as a key-value storage.
+These metrics can be either exported to :ref:`mod-graphite`,
+exposed as :ref:`mod-http-prometheus`, or processed using user-provided script
+as described in chapter :ref:`async-events`.
+
+.. note:: Please remember that each Knot Resolver instance keeps its own
+ statistics, and instances can be started and stopped dynamically. This might
+ affect your data postprocessing procedures if you are using
+ :ref:`systemd-multiple-instances`.
+
+.. _mod-stats-list:
+
+Built-in statistics
+-------------------
+
+Built-in counters keep track of number of queries and answers matching specific criteria.
+
++-----------------------------------------------------------------+
+| **Global request counters** |
++------------------+----------------------------------------------+
+| request.total | total number of DNS requests |
+| | (including internal client requests) |
++------------------+----------------------------------------------+
+| request.internal | internal requests generated by Knot Resolver |
+| | (e.g. DNSSEC trust anchor updates) |
++------------------+----------------------------------------------+
+| request.udp | external requests received over plain UDP |
+| | (:rfc:`1035`) |
++------------------+----------------------------------------------+
+| request.tcp | external requests received over plain TCP |
+| | (:rfc:`1035`) |
++------------------+----------------------------------------------+
+| request.dot | external requests received over |
+| | DNS-over-TLS (:rfc:`7858`) |
++------------------+----------------------------------------------+
+| request.doh | external requests received over |
+| | DNS-over-HTTP (:rfc:`8484`) |
++------------------+----------------------------------------------+
+| request.xdp | external requests received over plain UDP |
+| | via an AF_XDP socket |
++------------------+----------------------------------------------+
+
++----------------------------------------------------+
+| **Global answer counters** |
++-----------------+----------------------------------+
+| answer.total | total number of answered queries |
++-----------------+----------------------------------+
+| answer.cached | queries answered from cache |
++-----------------+----------------------------------+
+
++-----------------+----------------------------------+
+| **Answers categorized by RCODE** |
++-----------------+----------------------------------+
+| answer.noerror | NOERROR answers |
++-----------------+----------------------------------+
+| answer.nodata | NOERROR, but empty answers |
++-----------------+----------------------------------+
+| answer.nxdomain | NXDOMAIN answers |
++-----------------+----------------------------------+
+| answer.servfail | SERVFAIL answers |
++-----------------+----------------------------------+
+
++-----------------+----------------------------------+
+| **Answer latency** |
++-----------------+----------------------------------+
+| answer.1ms | completed in 1ms |
++-----------------+----------------------------------+
+| answer.10ms | completed in 10ms |
++-----------------+----------------------------------+
+| answer.50ms | completed in 50ms |
++-----------------+----------------------------------+
+| answer.100ms | completed in 100ms |
++-----------------+----------------------------------+
+| answer.250ms | completed in 250ms |
++-----------------+----------------------------------+
+| answer.500ms | completed in 500ms |
++-----------------+----------------------------------+
+| answer.1000ms | completed in 1000ms |
++-----------------+----------------------------------+
+| answer.1500ms | completed in 1500ms |
++-----------------+----------------------------------+
+| answer.slow | completed in more than 1500ms |
++-----------------+----------------------------------+
+
++-----------------+----------------------------------+
+| **Answer flags** |
++-----------------+----------------------------------+
+| answer.aa | authoritative answer |
++-----------------+----------------------------------+
+| answer.tc | truncated answer |
++-----------------+----------------------------------+
+| answer.ra | recursion available |
++-----------------+----------------------------------+
+| answer.rd | recursion desired (in answer!) |
++-----------------+----------------------------------+
+| answer.ad | authentic data (DNSSEC) |
++-----------------+----------------------------------+
+| answer.cd | checking disabled (DNSSEC) |
++-----------------+----------------------------------+
+| answer.do | DNSSEC answer OK |
++-----------------+----------------------------------+
+| answer.edns0 | EDNS0 present |
++-----------------+----------------------------------+
+
++-----------------+----------------------------------+
+| **Query flags** |
++-----------------+----------------------------------+
+| query.edns | queries with EDNS present |
++-----------------+----------------------------------+
+| query.dnssec | queries with DNSSEC DO=1 |
++-----------------+----------------------------------+
+
+Example:
+
+.. code-block:: none
+
+ modules.load('stats')
+
+ -- Enumerate metrics
+ > stats.list()
+ [answer.cached] => 486178
+ [iterator.tcp] => 490
+ [answer.noerror] => 507367
+ [answer.total] => 618631
+ [iterator.udp] => 102408
+ [query.concurrent] => 149
+
+ -- Query metrics by prefix
+ > stats.list('iter')
+ [iterator.udp] => 105104
+ [iterator.tcp] => 490
+
+ -- Fetch most common queries
+ > stats.frequent()
+ [1] => {
+ [type] => 2
+ [count] => 4
+ [name] => cz.
+ }
+
+ -- Fetch most common queries (sorted by frequency)
+ > table.sort(stats.frequent(), function (a, b) return a.count > b.count end)
+
+ -- Show recently contacted authoritative servers
+ > stats.upstreams()
+ [2a01:618:404::1] => {
+ [1] => 26 -- RTT
+ }
+ [128.241.220.33] => {
+ [1] => 31 - RTT
+ }
+
+ -- Set custom metrics from modules
+ > stats['filter.match'] = 5
+ > stats['filter.match']
+ 5
+
+Module reference
+----------------
+
+.. function:: stats.get(key)
+
+ :param string key: i.e. ``"answer.total"``
+ :return: ``number``
+
+Return nominal value of given metric.
+
+.. function:: stats.set('key val')
+
+Set nominal value of given metric.
+
+Example:
+
+.. code-block:: lua
+
+ stats.set('answer.total 5')
+ -- or syntactic sugar
+ stats['answer.total'] = 5
+
+
+.. function:: stats.list([prefix])
+
+ :param string prefix: optional metric prefix, i.e. ``"answer"`` shows only metrics beginning with "answer"
+
+Outputs collected metrics as a JSON dictionary.
+
+.. function:: stats.upstreams()
+
+Outputs a list of recent upstreams and their RTT. It is sorted by time and stored in a ring buffer of
+a fixed size. This means it's not aggregated and readable by multiple consumers, but also that
+you may lose entries if you don't read quickly enough. The default ring size is 512 entries, and may be overriden on compile time by ``-DUPSTREAMS_COUNT=X``.
+
+.. function:: stats.frequent()
+
+Outputs list of most frequent iterative queries as a JSON array. The queries are sampled probabilistically,
+and include subrequests. The list maximum size is 5000 entries, make diffs if you want to track it over time.
+
+.. function:: stats.clear_frequent()
+
+Clear the list of most frequent iterative queries.
+
+.. include:: ../modules/graphite/README.rst
+.. include:: ../modules/http/prometheus.rst
diff --git a/modules/stats/meson.build b/modules/stats/meson.build
new file mode 100644
index 0000000..68dd411
--- /dev/null
+++ b/modules/stats/meson.build
@@ -0,0 +1,24 @@
+# SPDX-License-Identifier: GPL-3.0-or-later
+# C module: stats
+
+stats_src = files([
+ 'stats.c',
+])
+c_src_lint += stats_src
+
+integr_tests += [
+ ['stats', meson.current_source_dir() / 'test.integr'],
+]
+
+
+stats_mod = shared_module(
+ 'stats',
+ stats_src,
+ dependencies: [
+ luajit_inc,
+ ],
+ include_directories: mod_inc_dir,
+ name_prefix: '',
+ install: true,
+ install_dir: modules_dir,
+)
diff --git a/modules/stats/stats.c b/modules/stats/stats.c
new file mode 100644
index 0000000..54b7a87
--- /dev/null
+++ b/modules/stats/stats.c
@@ -0,0 +1,510 @@
+/* Copyright (C) 2014-2017 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz>
+ * SPDX-License-Identifier: GPL-3.0-or-later
+ */
+
+/**
+ * @file stats.c
+ * @brief Storage for various counters and metrics from query resolution.
+ *
+ * You can either reuse this module to compute statistics or store custom metrics
+ * in it via the extensions.
+ */
+
+#include <libknot/packet/pkt.h>
+#include <libknot/packet/wire.h>
+#include <libknot/descriptor.h>
+#include <ccan/json/json.h>
+#include <contrib/cleanup.h>
+#include <arpa/inet.h>
+#include <lua.h>
+
+#include "lib/layer/iterate.h"
+#include "lib/rplan.h"
+#include "lib/module.h"
+#include "lib/layer.h"
+#include "lib/resolve.h"
+
+/* Defaults */
+#define VERBOSE_MSG(qry, ...) QRVERBOSE(qry, "stat", __VA_ARGS__)
+#define FREQUENT_PSAMPLE 10 /* Sampling rate, 1 in N */
+#ifdef LRU_REP_SIZE
+ #define FREQUENT_COUNT LRU_REP_SIZE /* Size of frequent tables */
+#else
+ #define FREQUENT_COUNT 5000 /* Size of frequent tables */
+#endif
+#ifndef UPSTREAMS_COUNT
+ #define UPSTREAMS_COUNT 512 /* Size of recent upstreams */
+#endif
+
+/** @cond internal Fixed-size map of predefined metrics. */
+#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,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) \
+ X(request,total) X(request,udp) X(request,tcp) X(request,xdp) \
+ X(request,dot) X(request,doh) X(request,internal) \
+ X(const,end)
+
+enum const_metric {
+ #define X(a,b) metric_ ## a ## _ ## b,
+ CONST_METRICS(X)
+ #undef X
+};
+struct const_metric_elm {
+ const char *key;
+ size_t val;
+};
+static struct const_metric_elm const_metrics[] = {
+ #define X(a,b) [metric_ ## a ## _ ## b] = { #a "." #b, 0 },
+ CONST_METRICS(X)
+ #undef X
+};
+/** @endcond */
+
+/** @internal LRU hash of most frequent names. */
+typedef lru_t(unsigned) namehash_t;
+typedef array_t(struct sockaddr_in6) addrlist_t;
+
+/** @internal Stats data structure. */
+struct stat_data {
+ map_t map;
+ struct {
+ namehash_t *frequent;
+ } queries;
+ struct {
+ addrlist_t q;
+ size_t head;
+ } upstreams;
+};
+
+/** @internal We don't store/publish port, repurpose it for RTT instead. */
+#define sin6_rtt sin6_port
+
+/** @internal Add to const map counter */
+static inline void stat_const_add(struct stat_data *data, enum const_metric key, ssize_t incr)
+{
+ const_metrics[key].val += incr;
+}
+
+static int collect_answer(struct stat_data *data, knot_pkt_t *pkt)
+{
+ stat_const_add(data, metric_answer_total, 1);
+ /* Count per-rcode */
+ switch(knot_wire_get_rcode(pkt->wire)) {
+ case KNOT_RCODE_NOERROR:
+ if (knot_wire_get_ancount(pkt->wire) > 0)
+ stat_const_add(data, metric_answer_noerror, 1);
+ else
+ stat_const_add(data, metric_answer_nodata, 1);
+ break;
+ case KNOT_RCODE_NXDOMAIN: stat_const_add(data, metric_answer_nxdomain, 1); break;
+ case KNOT_RCODE_SERVFAIL: stat_const_add(data, metric_answer_servfail, 1); break;
+ default: break;
+ }
+
+ return kr_ok();
+}
+
+static inline int collect_key(char *key, const knot_dname_t *name, uint16_t type)
+{
+ memcpy(key, &type, sizeof(type));
+ int key_len = knot_dname_to_wire((uint8_t *)key + sizeof(type), name, KNOT_DNAME_MAXLEN);
+ if (key_len < 0) {
+ return kr_error(key_len);
+ }
+ return key_len + sizeof(type);
+}
+
+static void collect_sample(struct stat_data *data, struct kr_rplan *rplan)
+{
+ /* Sample key = {[2] type, [1-255] owner} */
+ char key[sizeof(uint16_t) + KNOT_DNAME_MAXLEN];
+ for (size_t i = 0; i < rplan->resolved.len; ++i) {
+ /* Sample queries leading to iteration */
+ struct kr_query *qry = rplan->resolved.at[i];
+ if (qry->flags.CACHED) {
+ continue;
+ }
+ /* Consider 1 in N for frequent sampling.
+ * TODO: redesign the sampling approach. */
+ if (kr_rand_coin(1, FREQUENT_PSAMPLE)) {
+ int key_len = collect_key(key, qry->sname, qry->stype);
+ if (key_len < 0) {
+ assert(false);
+ continue;
+ }
+ unsigned *count = lru_get_new(data->queries.frequent, key, key_len, NULL);
+ if (count)
+ *count += 1;
+ }
+ }
+}
+
+static int collect_rtt(kr_layer_t *ctx, knot_pkt_t *pkt)
+{
+ struct kr_request *req = ctx->req;
+ struct kr_query *qry = req->current_query;
+ if (qry->flags.CACHED || !req->upstream.transport) {
+ return ctx->state;
+ }
+
+ /* Push address and RTT to the ring buffer head */
+ struct kr_module *module = ctx->api->data;
+ struct stat_data *data = module->data;
+
+ /* Socket address is encoded into sockaddr_in6 struct that
+ * unions with sockaddr_in and differ in sa_family */
+ struct sockaddr_in6 *e = &data->upstreams.q.at[data->upstreams.head];
+ const union inaddr *src = &req->upstream.transport->address;
+ switch (src->ip.sa_family) {
+ case AF_INET: memcpy(e, &src->ip4, sizeof(src->ip4)); break;
+ case AF_INET6: memcpy(e, &src->ip6, sizeof(src->ip6)); break;
+ default: return ctx->state;
+ }
+ /* Replace port number with the RTT information (cap is UINT16_MAX milliseconds) */
+ e->sin6_rtt = req->upstream.rtt;
+
+ /* Advance ring buffer head */
+ data->upstreams.head = (data->upstreams.head + 1) % UPSTREAMS_COUNT;
+ return ctx->state;
+}
+
+static int collect_transport(kr_layer_t *ctx)
+{
+ struct kr_request *req = ctx->req;
+ struct kr_module *module = ctx->api->data;
+ struct stat_data *data = module->data;
+
+ stat_const_add(data, metric_request_total, 1);
+ if (req->qsource.dst_addr == NULL) {
+ stat_const_add(data, metric_request_internal, 1);
+ return ctx->state;
+ }
+
+ /**
+ * Count each transport only once,
+ * i.e. DoT does not count as TCP and XDP does not count as UDP.
+ */
+ if (req->qsource.flags.http)
+ stat_const_add(data, metric_request_doh, 1);
+ else if (req->qsource.flags.tls)
+ stat_const_add(data, metric_request_dot, 1);
+ else if (req->qsource.flags.tcp)
+ stat_const_add(data, metric_request_tcp, 1);
+ else if (req->qsource.flags.xdp)
+ stat_const_add(data, metric_request_xdp, 1);
+ else
+ stat_const_add(data, metric_request_udp, 1);
+ return ctx->state;
+}
+
+static int collect(kr_layer_t *ctx)
+{
+ struct kr_request *param = ctx->req;
+ struct kr_module *module = ctx->api->data;
+ struct kr_rplan *rplan = &param->rplan;
+ struct stat_data *data = module->data;
+
+ collect_sample(data, rplan);
+ if (!param->answer) {
+ /* The answer is being dropped. TODO: perhaps add some stat for this? */
+ return ctx->state;
+ }
+
+ /* Collect data on final answer */
+ collect_answer(data, param->answer);
+ /* Count cached and unresolved */
+ if (rplan->resolved.len > 0) {
+ /* Histogram of answer latency. */
+ struct kr_query *first = rplan->resolved.at[0];
+ uint64_t elapsed = kr_now() - first->timestamp_mono;
+ if (elapsed <= 1) {
+ stat_const_add(data, metric_answer_1ms, 1);
+ } else if (elapsed <= 10) {
+ stat_const_add(data, metric_answer_10ms, 1);
+ } else if (elapsed <= 50) {
+ stat_const_add(data, metric_answer_50ms, 1);
+ } else if (elapsed <= 100) {
+ stat_const_add(data, metric_answer_100ms, 1);
+ } else if (elapsed <= 250) {
+ stat_const_add(data, metric_answer_250ms, 1);
+ } else if (elapsed <= 500) {
+ stat_const_add(data, metric_answer_500ms, 1);
+ } else if (elapsed <= 1000) {
+ stat_const_add(data, metric_answer_1000ms, 1);
+ } else if (elapsed <= 1500) {
+ stat_const_add(data, metric_answer_1500ms, 1);
+ } else {
+ stat_const_add(data, metric_answer_slow, 1);
+ }
+ /* Observe the final query. */
+ struct kr_query *last = kr_rplan_last(rplan);
+ stat_const_add(data, metric_answer_cached, last->flags.CACHED);
+ }
+
+ /* Keep stats of all response header flags;
+ * these don't return bool, so that's why we use !! */
+ stat_const_add(data, metric_answer_aa, !!knot_wire_get_aa(param->answer->wire));
+ stat_const_add(data, metric_answer_tc, !!knot_wire_get_tc(param->answer->wire));
+ stat_const_add(data, metric_answer_rd, !!knot_wire_get_rd(param->answer->wire));
+ stat_const_add(data, metric_answer_ra, !!knot_wire_get_ra(param->answer->wire));
+ stat_const_add(data, metric_answer_ad, !!knot_wire_get_ad(param->answer->wire));
+ stat_const_add(data, metric_answer_cd, !!knot_wire_get_cd(param->answer->wire));
+
+ /* EDNS0 stats */
+ stat_const_add(data, metric_answer_edns0, knot_pkt_has_edns(param->answer));
+ stat_const_add(data, metric_answer_do, knot_pkt_has_dnssec(param->answer));
+
+ /* Query parameters and transport mode */
+ /*
+ DEPRECATED
+ use new names metric_answer_edns0 and metric_answer_do
+ */
+ 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));
+
+ return ctx->state;
+}
+
+/**
+ * Set nominal value of a key.
+ *
+ * Input: { key, val }
+ *
+ */
+static char* stats_set(void *env, struct kr_module *module, const char *args)
+{
+ if (args == NULL)
+ return NULL;
+
+ struct stat_data *data = module->data;
+
+ auto_free char *pair = strdup(args);
+ char *val = strchr(pair, ' ');
+ if (val) {
+ *val = '\0';
+ size_t number = strtoul(val + 1, NULL, 10);
+ for (unsigned i = 0; i < metric_const_end; ++i) {
+ if (strcmp(const_metrics[i].key, pair) == 0) {
+ const_metrics[i].val = number;
+ return NULL;
+ }
+ }
+ map_set(&data->map, pair, (void *)number);
+ }
+
+ return NULL;
+}
+
+/**
+ * Retrieve metrics by key.
+ *
+ * Input: string key
+ * Output: number value
+ */
+static char* stats_get(void *env, struct kr_module *module, const char *args)
+{
+ if (args == NULL)
+ return NULL;
+
+ struct stat_data *data = module->data;
+
+ /* Expecting CHAR_BIT to be 8, this is a safe bet */
+ char *ret = malloc(3 * sizeof(size_t) + 2);
+ if (!ret) {
+ return NULL;
+ }
+
+ /* Check if it exists in const map. */
+ for (unsigned i = 0; i < metric_const_end; ++i) {
+ if (strcmp(const_metrics[i].key, args) == 0) {
+ sprintf(ret, "%zu", const_metrics[i].val);
+ return ret;
+ }
+ }
+ /* Check in variable map */
+ if (!map_contains(&data->map, args)) {
+ free(ret);
+ return NULL;
+ }
+ void *val = map_get(&data->map, args);
+ sprintf(ret, "%zu", (size_t) val);
+ return ret;
+}
+
+static int list_entry(const char *key, void *val, void *baton)
+{
+ JsonNode *root = baton;
+ size_t number = (size_t) val;
+ json_append_member(root, key, json_mknumber(number));
+ return 0;
+}
+
+/**
+ * List observed metrics.
+ *
+ * Output: { key: val, ... }
+ */
+static char* stats_list(void *env, struct kr_module *module, const char *args)
+{
+ struct stat_data *data = module->data;
+ JsonNode *root = json_mkobject();
+ /* Walk const metrics map */
+ size_t args_len = args ? strlen(args) : 0;
+ for (unsigned i = 0; i < metric_const_end; ++i) {
+ struct const_metric_elm *elm = &const_metrics[i];
+ if (!args || strncmp(elm->key, args, args_len) == 0) {
+ json_append_member(root, elm->key, json_mknumber(elm->val));
+ }
+ }
+ map_walk_prefixed(&data->map, (args_len > 0) ? args : "", list_entry, root);
+ char *ret = json_encode(root);
+ json_delete(root);
+ return ret;
+}
+
+/** @internal Helper for dump_list: add a single namehash_t item to JSON. */
+static enum lru_apply_do dump_value(const char *key, uint len, unsigned *val, void *baton)
+{
+ uint16_t key_type = 0;
+ /* Extract query name, type and counter */
+ memcpy(&key_type, key, sizeof(key_type));
+ KR_DNAME_GET_STR(key_name, (uint8_t *)key + sizeof(key_type));
+ KR_RRTYPE_GET_STR(type_str, key_type);
+
+ /* Convert to JSON object */
+ JsonNode *json_val = json_mkobject();
+ json_append_member(json_val, "count", json_mknumber(*val));
+ json_append_member(json_val, "name", json_mkstring(key_name));
+ json_append_member(json_val, "type", json_mkstring(type_str));
+ json_append_element((JsonNode *)baton, json_val);
+ return LRU_APPLY_DO_NOTHING; // keep the item
+}
+/**
+ * List frequent names.
+ *
+ * Output: [{ count: <counter>, name: <qname>, type: <qtype>}, ... ]
+ */
+static char* dump_list(void *env, struct kr_module *module, const char *args, namehash_t *table)
+{
+ if (!table) {
+ return NULL;
+ }
+ JsonNode *root = json_mkarray();
+ lru_apply(table, dump_value, root);
+ char *ret = json_encode(root);
+ json_delete(root);
+ return ret;
+}
+
+static char* dump_frequent(void *env, struct kr_module *module, const char *args)
+{
+ struct stat_data *data = module->data;
+ return dump_list(env, module, args, data->queries.frequent);
+}
+
+static char* clear_frequent(void *env, struct kr_module *module, const char *args)
+{
+ struct stat_data *data = module->data;
+ lru_reset(data->queries.frequent);
+ return NULL;
+}
+
+static char* dump_upstreams(void *env, struct kr_module *module, const char *args)
+{
+ struct stat_data *data = module->data;
+ if (!data) {
+ return NULL;
+ }
+
+ /* Walk the ring backwards until AF_UNSPEC or we hit head. */
+ JsonNode *root = json_mkobject();
+ size_t head = data->upstreams.head;
+ for (size_t i = 1; i < UPSTREAMS_COUNT; ++i) {
+ size_t h = (UPSTREAMS_COUNT + head - i) % UPSTREAMS_COUNT;
+ struct sockaddr_in6 *e = &data->upstreams.q.at[h];
+ if (e->sin6_family == AF_UNSPEC) {
+ break;
+ }
+ /* Convert address to string */
+ char addr_str[INET6_ADDRSTRLEN];
+ const char *ret = inet_ntop(e->sin6_family, kr_inaddr((const struct sockaddr *)e), addr_str, sizeof(addr_str));
+ if (!ret) {
+ break;
+ }
+ /* Append to map with an array encoding RTTs */
+ JsonNode *json_val = json_find_member(root, addr_str);
+ if (!json_val) {
+ json_val = json_mkarray();
+ json_append_member(root, addr_str, json_val);
+ }
+ json_append_element(json_val, json_mknumber(e->sin6_rtt));
+ }
+
+ /* Encode and return */
+ char *ret = json_encode(root);
+ json_delete(root);
+ return ret;
+}
+
+KR_EXPORT
+int stats_init(struct kr_module *module)
+{
+ static kr_layer_api_t layer = {
+ .consume = &collect_rtt,
+ .finish = &collect,
+ .begin = &collect_transport,
+ };
+ /* Store module reference */
+ layer.data = module;
+ module->layer = &layer;
+
+ static const struct kr_prop props[] = {
+ { &stats_set, "set", "Set {key, val} metrics.", },
+ { &stats_get, "get", "Get metrics for given key.", },
+ { &stats_list, "list", "List observed metrics.", },
+ { &dump_frequent, "frequent", "List most frequent queries.", },
+ { &clear_frequent,"clear_frequent", "Clear frequent queries log.", },
+ { &dump_upstreams, "upstreams", "List recently seen authoritatives.", },
+ { NULL, NULL, NULL }
+ };
+ module->props = props;
+
+ struct stat_data *data = calloc(1, sizeof(*data));
+ if (!data) {
+ return kr_error(ENOMEM);
+ }
+ data->map = map_make(NULL);
+ module->data = data;
+ lru_create(&data->queries.frequent, FREQUENT_COUNT, NULL, NULL);
+ /* Initialize ring buffer of recently visited upstreams */
+ array_init(data->upstreams.q);
+ if (array_reserve(data->upstreams.q, UPSTREAMS_COUNT) != 0) {
+ return kr_error(ENOMEM);
+ }
+ data->upstreams.q.len = UPSTREAMS_COUNT; /* signify we use the entries */
+ for (size_t i = 0; i < UPSTREAMS_COUNT; ++i) {
+ data->upstreams.q.at[i].sin6_family = AF_UNSPEC;
+ }
+ return kr_ok();
+}
+
+KR_EXPORT
+int stats_deinit(struct kr_module *module)
+{
+ struct stat_data *data = module->data;
+ if (data) {
+ map_clear(&data->map);
+ lru_free(data->queries.frequent);
+ array_clear(data->upstreams.q);
+ free(data);
+ }
+ return kr_ok();
+}
+
+KR_MODULE_EXPORT(stats)
+
+#undef VERBOSE_MSG
diff --git a/modules/stats/test.integr/deckard.yaml b/modules/stats/test.integr/deckard.yaml
new file mode 100644
index 0000000..6dd0c22
--- /dev/null
+++ b/modules/stats/test.integr/deckard.yaml
@@ -0,0 +1,12 @@
+# SPDX-License-Identifier: GPL-3.0-or-later
+programs:
+- name: kresd
+ binary: kresd
+ additional:
+ - --noninteractive
+ templates:
+ - modules/stats/test.integr/kresd_config.j2
+ - tests/integration/hints_zone.j2
+ configs:
+ - config
+ - hints
diff --git a/modules/stats/test.integr/kresd_config.j2 b/modules/stats/test.integr/kresd_config.j2
new file mode 100644
index 0000000..6c83e54
--- /dev/null
+++ b/modules/stats/test.integr/kresd_config.j2
@@ -0,0 +1,111 @@
+-- SPDX-License-Identifier: GPL-3.0-or-later
+{% raw %}
+modules.load('stats')
+
+FWD_TARGET = policy.FORWARD('192.0.2.1')
+
+function check_stats(got)
+ log('checking if stat values match expected values:')
+ local expected = {
+ ['answer.cd'] = 2,
+ ['answer.cached'] = 1,
+ ['answer.nodata'] = 1,
+ ['answer.noerror'] = 2,
+ ['answer.nxdomain'] = 1,
+ ['answer.servfail'] = 2,
+ ['answer.edns0'] = 6,
+ ['answer.ra'] = 6,
+ ['answer.rd'] = 5,
+ ['answer.do'] = 1,
+ ['answer.ad'] = 0,
+ ['answer.tc'] = 0,
+ ['answer.aa'] = 0,
+ ['answer.total'] = 6
+ }
+ print(table_print(expected))
+
+ local ok = true
+ for key, expval in pairs(expected) do
+ if got[key] ~= expval then
+ log('ERROR: stats key ' .. key
+ .. ' has unexpected value'
+ .. ' (expected ' .. tostring(expval)
+ .. ' got ' .. tostring(got[key] .. ')'))
+ ok = false
+ end
+ end
+ if ok then
+ log('no problem found')
+ return FWD_TARGET
+ else
+ return policy.DENY_MSG('Stats test failure')
+ end
+end
+
+function reply_result(state, req)
+ local got = stats.list()
+ log('current stats.list() values:')
+ print(table_print(got))
+ local result = check_stats(got)
+ return result(state, req)
+end
+policy.add(policy.pattern(reply_result, 'stats.test.'))
+policy.add(policy.all(FWD_TARGET)) -- avoid iteration
+
+-- make sure DNSSEC is turned off for tests
+trust_anchors.remove('.')
+
+-- Enable queries without RD bit
+pcall(modules.unload, 'refuse_nord')
+
+-- Disable RFC5011 TA update
+if ta_update then
+ modules.unload('ta_update')
+end
+
+-- Disable RFC8145 signaling, scenario doesn't provide expected answers
+if ta_signal_query then
+ modules.unload('ta_signal_query')
+end
+
+-- Disable RFC8109 priming, scenario doesn't provide expected answers
+if priming then
+ modules.unload('priming')
+end
+
+-- Disable this module because it make one priming query
+if detect_time_skew then
+ modules.unload('detect_time_skew')
+end
+
+_hint_root_file('hints')
+cache.size = 2*MB
+verbose(true)
+{% endraw %}
+
+net = { '{{SELF_ADDR}}' }
+
+
+{% if QMIN == "false" %}
+option('NO_MINIMIZE', true)
+{% else %}
+option('NO_MINIMIZE', false)
+{% endif %}
+
+
+-- Self-checks on globals
+assert(help() ~= nil)
+assert(worker.id ~= nil)
+-- Self-checks on facilities
+assert(cache.count() == 0)
+assert(cache.stats() ~= nil)
+assert(cache.backends() ~= nil)
+assert(worker.stats() ~= nil)
+assert(net.interfaces() ~= nil)
+-- Self-checks on loaded stuff
+assert(net.list()[1].transport.ip == '{{SELF_ADDR}}')
+assert(#modules.list() > 0)
+-- Self-check timers
+ev = event.recurrent(1 * sec, function (ev) return 1 end)
+event.cancel(ev)
+ev = event.after(0, function (ev) return 1 end)
diff --git a/modules/stats/test.integr/stats.rpl b/modules/stats/test.integr/stats.rpl
new file mode 100644
index 0000000..ecab062
--- /dev/null
+++ b/modules/stats/test.integr/stats.rpl
@@ -0,0 +1,194 @@
+; SPDX-License-Identifier: GPL-3.0-or-later
+ trust-anchor: "example. DNSKEY 257 3 7 AwEAAcUlFV1vhmqx6NSOUOq2R/dsR7Xm3upJ ( j7IommWSpJABVfW8Q0rOvXdM6kzt+TAu92L9 AbsUdblMFin8CVF3n4s= )"
+CONFIG_END
+
+SCENARIO_BEGIN Test stats module
+
+RANGE_BEGIN 0 100
+ ADDRESS 192.0.2.1
+
+ENTRY_BEGIN
+REPLY QR RA RD CD NOERROR
+MATCH opcode question rcode
+ADJUST copy_id
+SECTION QUESTION
+cd.test. IN TXT
+SECTION ANSWER
+cd.test. IN TXT "CD is set"
+ENTRY_END
+
+ENTRY_BEGIN
+REPLY QR RA RD CD NOERROR
+MATCH opcode question rcode
+ADJUST copy_id
+SECTION QUESTION
+nodata.test. IN TXT
+ENTRY_END
+
+ENTRY_BEGIN
+REPLY QR RA RD CD NXDOMAIN
+MATCH opcode question
+ADJUST copy_id
+SECTION QUESTION
+nxdomain.test. IN TXT
+ENTRY_END
+
+; failing DNSSEC-signed subdomain
+ENTRY_BEGIN
+REPLY QR RA RD CD SERVFAIL
+MATCH opcode subdomain
+ADJUST copy_id copy_query
+SECTION QUESTION
+bogus.test. IN TXT
+ENTRY_END
+
+; query for this name triggers check in Lua config
+ENTRY_BEGIN
+REPLY QR RA RD CD NOERROR
+MATCH opcode question rcode
+ADJUST copy_id
+SECTION QUESTION
+stats.test. IN TXT
+SECTION ANSWER
+stats.test. IN TXT "Ok, trigger query was not intercepted!"
+ENTRY_END
+
+ENTRY_BEGIN
+REPLY QR RD RA CD TC NOERROR
+MATCH opcode question rcode
+ADJUST copy_id
+SECTION QUESTION
+tc.test. IN URI
+ENTRY_END
+
+RANGE_END
+
+
+; +cd +rd
+STEP 10 QUERY
+ENTRY_BEGIN
+REPLY RD CD NOERROR
+SECTION QUESTION
+cd.test. IN TXT
+ENTRY_END
+
+STEP 11 CHECK_ANSWER
+ENTRY_BEGIN
+MATCH all
+REPLY QR RD RA CD NOERROR
+SECTION QUESTION
+cd.test. IN TXT
+SECTION ANSWER
+cd.test. IN TXT "CD is set"
+ENTRY_END
+
+; +cd +cached +rd
+STEP 12 QUERY
+ENTRY_BEGIN
+REPLY RD CD NOERROR
+SECTION QUESTION
+cd.test. IN TXT
+ENTRY_END
+
+STEP 13 CHECK_ANSWER
+ENTRY_BEGIN
+MATCH all
+REPLY QR RD RA CD NOERROR
+SECTION QUESTION
+cd.test. IN TXT
+SECTION ANSWER
+cd.test. IN TXT "CD is set"
+ENTRY_END
+
+; +nodata +rd
+STEP 20 QUERY
+ENTRY_BEGIN
+REPLY RD NOERROR
+SECTION QUESTION
+nodata.test. IN TXT
+SECTION ADDITIONAL
+ENTRY_END
+
+STEP 21 CHECK_ANSWER
+ENTRY_BEGIN
+REPLY QR RD RA NOERROR
+MATCH all
+SECTION QUESTION
+nodata.test. IN TXT
+ENTRY_END
+
+; +nxdomain +rd
+STEP 30 QUERY
+ENTRY_BEGIN
+REPLY RD NOERROR
+SECTION QUESTION
+nxdomain.test. IN TXT
+SECTION ADDITIONAL
+ENTRY_END
+
+STEP 31 CHECK_ANSWER
+ENTRY_BEGIN
+REPLY QR RD RA NXDOMAIN
+MATCH all
+SECTION QUESTION
+nxdomain.test. IN TXT
+ENTRY_END
+
+; +servfail +do +rd
+STEP 40 QUERY
+ENTRY_BEGIN
+REPLY RD DO NOERROR
+SECTION QUESTION
+bogus.test. IN TXT
+SECTION ADDITIONAL
+ENTRY_END
+
+STEP 41 CHECK_ANSWER
+ENTRY_BEGIN
+REPLY QR RD RA DO SERVFAIL
+MATCH all
+SECTION QUESTION
+bogus.test. IN TXT
+ENTRY_END
+
+; no rd
+STEP 50 QUERY
+ENTRY_BEGIN
+REPLY NOERROR
+SECTION QUESTION
+bogus.test. IN TXT
+SECTION ADDITIONAL
+ENTRY_END
+
+STEP 51 CHECK_ANSWER
+ENTRY_BEGIN
+REPLY QR RA SERVFAIL
+MATCH all
+SECTION QUESTION
+bogus.test. IN TXT
+ENTRY_END
+
+
+
+
+STEP 100 QUERY
+ENTRY_BEGIN
+REPLY RD NOERROR
+SECTION QUESTION
+stats.test. IN TXT
+SECTION ADDITIONAL
+ENTRY_END
+
+STEP 101 CHECK_ANSWER
+ENTRY_BEGIN
+REPLY NOERROR
+MATCH opcode question additional rcode answer
+; AD must not be set in the answer
+SECTION QUESTION
+stats.test. IN TXT
+SECTION ANSWER
+stats.test. IN TXT "Ok, trigger query was not intercepted!"
+ENTRY_END
+
+
+SCENARIO_END