2 * daemon/stats.c - collect runtime performance indicators.
4 * Copyright (c) 2007, NLnet Labs. All rights reserved.
6 * This software is open source.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * Redistributions of source code must retain the above copyright notice,
13 * this list of conditions and the following disclaimer.
15 * Redistributions in binary form must reproduce the above copyright notice,
16 * this list of conditions and the following disclaimer in the documentation
17 * and/or other materials provided with the distribution.
19 * Neither the name of the NLNET LABS nor the names of its contributors may
20 * be used to endorse or promote products derived from this software without
21 * specific prior written permission.
23 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
24 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
25 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
26 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
27 * HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
28 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
29 * TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
30 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
39 * This file describes the data structure used to collect runtime performance
40 * numbers. These 'statistics' may be of interest to the operator.
47 #include <sys/types.h>
48 #include "daemon/stats.h"
49 #include "daemon/worker.h"
50 #include "daemon/daemon.h"
51 #include "services/mesh.h"
52 #include "services/outside_network.h"
53 #include "services/listen_dnsport.h"
54 #include "util/config_file.h"
55 #include "util/tube.h"
56 #include "util/timehist.h"
57 #include "util/net_help.h"
58 #include "validator/validator.h"
59 #include "iterator/iterator.h"
60 #include "sldns/sbuffer.h"
61 #include "services/cache/rrset.h"
62 #include "services/cache/infra.h"
63 #include "validator/val_kcache.h"
65 /** add timers and the values do not overflow or become negative */
67 stats_timeval_add(long long* d_sec, long long* d_usec, long long add_sec, long long add_usec)
71 (*d_usec) += add_usec;
72 if((*d_usec) > 1000000) {
79 void server_stats_init(struct ub_server_stats* stats, struct config_file* cfg)
81 memset(stats, 0, sizeof(*stats));
82 stats->extended = cfg->stat_extended;
85 void server_stats_querymiss(struct ub_server_stats* stats, struct worker* worker)
87 stats->num_queries_missed_cache++;
88 stats->sum_query_list_size += worker->env.mesh->all.count;
89 if((long long)worker->env.mesh->all.count > stats->max_query_list_size)
90 stats->max_query_list_size = (long long)worker->env.mesh->all.count;
93 void server_stats_prefetch(struct ub_server_stats* stats, struct worker* worker)
95 stats->num_queries_prefetch++;
96 /* changes the query list size so account that, like a querymiss */
97 stats->sum_query_list_size += worker->env.mesh->all.count;
98 if((long long)worker->env.mesh->all.count > stats->max_query_list_size)
99 stats->max_query_list_size = (long long)worker->env.mesh->all.count;
102 void server_stats_log(struct ub_server_stats* stats, struct worker* worker,
105 log_info("server stats for thread %d: %u queries, "
106 "%u answers from cache, %u recursions, %u prefetch, %u rejected by "
108 threadnum, (unsigned)stats->num_queries,
109 (unsigned)(stats->num_queries -
110 stats->num_queries_missed_cache),
111 (unsigned)stats->num_queries_missed_cache,
112 (unsigned)stats->num_queries_prefetch,
113 (unsigned)stats->num_queries_ip_ratelimited);
114 log_info("server stats for thread %d: requestlist max %u avg %g "
115 "exceeded %u jostled %u", threadnum,
116 (unsigned)stats->max_query_list_size,
117 (stats->num_queries_missed_cache+stats->num_queries_prefetch)?
118 (double)stats->sum_query_list_size/
119 (double)(stats->num_queries_missed_cache+
120 stats->num_queries_prefetch) : 0.0,
121 (unsigned)worker->env.mesh->stats_dropped,
122 (unsigned)worker->env.mesh->stats_jostled);
125 /** get rrsets bogus number from validator */
127 get_rrset_bogus(struct worker* worker, int reset)
129 int m = modstack_find(&worker->env.mesh->mods, "validator");
134 ve = (struct val_env*)worker->env.modinfo[m];
135 lock_basic_lock(&ve->bogus_lock);
136 r = ve->num_rrset_bogus;
137 if(reset && !worker->env.cfg->stat_cumulative)
138 ve->num_rrset_bogus = 0;
139 lock_basic_unlock(&ve->bogus_lock);
143 /** get number of ratelimited queries from iterator */
145 get_queries_ratelimit(struct worker* worker, int reset)
147 int m = modstack_find(&worker->env.mesh->mods, "iterator");
152 ie = (struct iter_env*)worker->env.modinfo[m];
153 lock_basic_lock(&ie->queries_ratelimit_lock);
154 r = ie->num_queries_ratelimited;
155 if(reset && !worker->env.cfg->stat_cumulative)
156 ie->num_queries_ratelimited = 0;
157 lock_basic_unlock(&ie->queries_ratelimit_lock);
162 /** get the number of shared secret cache miss */
164 get_dnscrypt_cache_miss(struct worker* worker, int reset)
167 struct dnsc_env* de = worker->daemon->dnscenv;
170 lock_basic_lock(&de->shared_secrets_cache_lock);
171 r = de->num_query_dnscrypt_secret_missed_cache;
172 if(reset && !worker->env.cfg->stat_cumulative)
173 de->num_query_dnscrypt_secret_missed_cache = 0;
174 lock_basic_unlock(&de->shared_secrets_cache_lock);
177 #endif /* USE_DNSCRYPT */
180 server_stats_compile(struct worker* worker, struct ub_stats_info* s, int reset)
183 struct listen_list* lp;
185 s->svr = worker->stats;
186 s->mesh_num_states = (long long)worker->env.mesh->all.count;
187 s->mesh_num_reply_states = (long long)worker->env.mesh->num_reply_states;
188 s->mesh_jostled = (long long)worker->env.mesh->stats_jostled;
189 s->mesh_dropped = (long long)worker->env.mesh->stats_dropped;
190 s->mesh_replies_sent = (long long)worker->env.mesh->replies_sent;
191 s->mesh_replies_sum_wait_sec = (long long)worker->env.mesh->replies_sum_wait.tv_sec;
192 s->mesh_replies_sum_wait_usec = (long long)worker->env.mesh->replies_sum_wait.tv_usec;
193 s->mesh_time_median = timehist_quartile(worker->env.mesh->histogram,
196 /* add in the values from the mesh */
197 s->svr.ans_secure += (long long)worker->env.mesh->ans_secure;
198 s->svr.ans_bogus += (long long)worker->env.mesh->ans_bogus;
199 s->svr.ans_rcode_nodata += (long long)worker->env.mesh->ans_nodata;
201 s->svr.ans_rcode[i] += (long long)worker->env.mesh->ans_rcode[i];
202 timehist_export(worker->env.mesh->histogram, s->svr.hist,
204 /* values from outside network */
205 s->svr.unwanted_replies = (long long)worker->back->unwanted_replies;
206 s->svr.qtcp_outgoing = (long long)worker->back->num_tcp_outgoing;
208 /* get and reset validator rrset bogus number */
209 s->svr.rrset_bogus = (long long)get_rrset_bogus(worker, reset);
211 /* get and reset iterator query ratelimit number */
212 s->svr.queries_ratelimited = (long long)get_queries_ratelimit(worker, reset);
214 /* get cache sizes */
215 s->svr.msg_cache_count = (long long)count_slabhash_entries(worker->env.msg_cache);
216 s->svr.rrset_cache_count = (long long)count_slabhash_entries(&worker->env.rrset_cache->table);
217 s->svr.infra_cache_count = (long long)count_slabhash_entries(worker->env.infra_cache->hosts);
218 if(worker->env.key_cache)
219 s->svr.key_cache_count = (long long)count_slabhash_entries(worker->env.key_cache->slab);
220 else s->svr.key_cache_count = 0;
223 if(worker->daemon->dnscenv) {
224 s->svr.num_query_dnscrypt_secret_missed_cache =
225 (long long)get_dnscrypt_cache_miss(worker, reset);
226 s->svr.shared_secret_cache_count = (long long)count_slabhash_entries(
227 worker->daemon->dnscenv->shared_secrets_cache);
229 s->svr.num_query_dnscrypt_secret_missed_cache = 0;
230 s->svr.shared_secret_cache_count = 0;
233 s->svr.num_query_dnscrypt_secret_missed_cache = 0;
234 s->svr.shared_secret_cache_count = 0;
235 #endif /* USE_DNSCRYPT */
237 /* get tcp accept usage */
238 s->svr.tcp_accept_usage = 0;
239 for(lp = worker->front->cps; lp; lp = lp->next) {
240 if(lp->com->type == comm_tcp_accept)
241 s->svr.tcp_accept_usage += (long long)lp->com->cur_tcp_count;
244 if(reset && !worker->env.cfg->stat_cumulative) {
245 worker_stats_clear(worker);
249 void server_stats_obtain(struct worker* worker, struct worker* who,
250 struct ub_stats_info* s, int reset)
252 uint8_t *reply = NULL;
255 /* just fill it in */
256 server_stats_compile(worker, s, reset);
259 /* communicate over tube */
260 verbose(VERB_ALGO, "write stats cmd");
262 worker_send_cmd(who, worker_cmd_stats);
263 else worker_send_cmd(who, worker_cmd_stats_noreset);
264 verbose(VERB_ALGO, "wait for stats reply");
265 if(!tube_read_msg(worker->cmd, &reply, &len, 0))
266 fatal_exit("failed to read stats over cmd channel");
267 if(len != (uint32_t)sizeof(*s))
268 fatal_exit("stats on cmd channel wrong length %d %d",
269 (int)len, (int)sizeof(*s));
270 memcpy(s, reply, (size_t)len);
274 void server_stats_reply(struct worker* worker, int reset)
276 struct ub_stats_info s;
277 server_stats_compile(worker, &s, reset);
278 verbose(VERB_ALGO, "write stats replymsg");
279 if(!tube_write_msg(worker->daemon->workers[0]->cmd,
280 (uint8_t*)&s, sizeof(s), 0))
281 fatal_exit("could not write stat values over cmd channel");
284 void server_stats_add(struct ub_stats_info* total, struct ub_stats_info* a)
286 total->svr.num_queries += a->svr.num_queries;
287 total->svr.num_queries_ip_ratelimited += a->svr.num_queries_ip_ratelimited;
288 total->svr.num_queries_missed_cache += a->svr.num_queries_missed_cache;
289 total->svr.num_queries_prefetch += a->svr.num_queries_prefetch;
290 total->svr.sum_query_list_size += a->svr.sum_query_list_size;
292 total->svr.num_query_dnscrypt_crypted += a->svr.num_query_dnscrypt_crypted;
293 total->svr.num_query_dnscrypt_cert += a->svr.num_query_dnscrypt_cert;
294 total->svr.num_query_dnscrypt_cleartext += \
295 a->svr.num_query_dnscrypt_cleartext;
296 total->svr.num_query_dnscrypt_crypted_malformed += \
297 a->svr.num_query_dnscrypt_crypted_malformed;
298 #endif /* USE_DNSCRYPT */
299 /* the max size reached is upped to higher of both */
300 if(a->svr.max_query_list_size > total->svr.max_query_list_size)
301 total->svr.max_query_list_size = a->svr.max_query_list_size;
303 if(a->svr.extended) {
305 total->svr.qtype_big += a->svr.qtype_big;
306 total->svr.qclass_big += a->svr.qclass_big;
307 total->svr.qtcp += a->svr.qtcp;
308 total->svr.qtcp_outgoing += a->svr.qtcp_outgoing;
309 total->svr.qipv6 += a->svr.qipv6;
310 total->svr.qbit_QR += a->svr.qbit_QR;
311 total->svr.qbit_AA += a->svr.qbit_AA;
312 total->svr.qbit_TC += a->svr.qbit_TC;
313 total->svr.qbit_RD += a->svr.qbit_RD;
314 total->svr.qbit_RA += a->svr.qbit_RA;
315 total->svr.qbit_Z += a->svr.qbit_Z;
316 total->svr.qbit_AD += a->svr.qbit_AD;
317 total->svr.qbit_CD += a->svr.qbit_CD;
318 total->svr.qEDNS += a->svr.qEDNS;
319 total->svr.qEDNS_DO += a->svr.qEDNS_DO;
320 total->svr.ans_rcode_nodata += a->svr.ans_rcode_nodata;
321 total->svr.zero_ttl_responses += a->svr.zero_ttl_responses;
322 total->svr.ans_secure += a->svr.ans_secure;
323 total->svr.ans_bogus += a->svr.ans_bogus;
324 total->svr.unwanted_replies += a->svr.unwanted_replies;
325 total->svr.unwanted_queries += a->svr.unwanted_queries;
326 total->svr.tcp_accept_usage += a->svr.tcp_accept_usage;
327 for(i=0; i<UB_STATS_QTYPE_NUM; i++)
328 total->svr.qtype[i] += a->svr.qtype[i];
329 for(i=0; i<UB_STATS_QCLASS_NUM; i++)
330 total->svr.qclass[i] += a->svr.qclass[i];
331 for(i=0; i<UB_STATS_OPCODE_NUM; i++)
332 total->svr.qopcode[i] += a->svr.qopcode[i];
333 for(i=0; i<UB_STATS_RCODE_NUM; i++)
334 total->svr.ans_rcode[i] += a->svr.ans_rcode[i];
335 for(i=0; i<NUM_BUCKETS_HIST; i++)
336 total->svr.hist[i] += a->svr.hist[i];
339 total->mesh_num_states += a->mesh_num_states;
340 total->mesh_num_reply_states += a->mesh_num_reply_states;
341 total->mesh_jostled += a->mesh_jostled;
342 total->mesh_dropped += a->mesh_dropped;
343 total->mesh_replies_sent += a->mesh_replies_sent;
344 stats_timeval_add(&total->mesh_replies_sum_wait_sec, &total->mesh_replies_sum_wait_usec, a->mesh_replies_sum_wait_sec, a->mesh_replies_sum_wait_usec);
345 /* the medians are averaged together, this is not as accurate as
346 * taking the median over all of the data, but is good and fast
347 * added up here, division later*/
348 total->mesh_time_median += a->mesh_time_median;
351 void server_stats_insquery(struct ub_server_stats* stats, struct comm_point* c,
352 uint16_t qtype, uint16_t qclass, struct edns_data* edns,
353 struct comm_reply* repinfo)
355 uint16_t flags = sldns_buffer_read_u16_at(c->buffer, 2);
356 if(qtype < UB_STATS_QTYPE_NUM)
357 stats->qtype[qtype]++;
358 else stats->qtype_big++;
359 if(qclass < UB_STATS_QCLASS_NUM)
360 stats->qclass[qclass]++;
361 else stats->qclass_big++;
362 stats->qopcode[ LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) ]++;
363 if(c->type != comm_udp)
365 if(repinfo && addr_is_ip6(&repinfo->addr, repinfo->addrlen))
383 if(edns->edns_present) {
385 if( (edns->bits & EDNS_DO) )
390 void server_stats_insrcode(struct ub_server_stats* stats, sldns_buffer* buf)
392 if(stats->extended && sldns_buffer_limit(buf) != 0) {
393 int r = (int)LDNS_RCODE_WIRE( sldns_buffer_begin(buf) );
394 stats->ans_rcode[r] ++;
395 if(r == 0 && LDNS_ANCOUNT( sldns_buffer_begin(buf) ) == 0)
396 stats->ans_rcode_nodata ++;