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 "sldns/sbuffer.h"
60 #include "services/cache/rrset.h"
61 #include "services/cache/infra.h"
62 #include "validator/val_kcache.h"
64 /** add timers and the values do not overflow or become negative */
66 timeval_add(struct timeval* d, const struct timeval* add)
69 d->tv_sec += add->tv_sec;
70 d->tv_usec += add->tv_usec;
71 if(d->tv_usec > 1000000) {
72 d->tv_usec -= 1000000;
78 void server_stats_init(struct server_stats* stats, struct config_file* cfg)
80 memset(stats, 0, sizeof(*stats));
81 stats->extended = cfg->stat_extended;
84 void server_stats_querymiss(struct server_stats* stats, struct worker* worker)
86 stats->num_queries_missed_cache++;
87 stats->sum_query_list_size += worker->env.mesh->all.count;
88 if(worker->env.mesh->all.count > stats->max_query_list_size)
89 stats->max_query_list_size = worker->env.mesh->all.count;
92 void server_stats_prefetch(struct server_stats* stats, struct worker* worker)
94 stats->num_queries_prefetch++;
95 /* changes the query list size so account that, like a querymiss */
96 stats->sum_query_list_size += worker->env.mesh->all.count;
97 if(worker->env.mesh->all.count > stats->max_query_list_size)
98 stats->max_query_list_size = worker->env.mesh->all.count;
101 void server_stats_log(struct server_stats* stats, struct worker* worker,
104 log_info("server stats for thread %d: %u queries, "
105 "%u answers from cache, %u recursions, %u prefetch, %u rejected by "
107 threadnum, (unsigned)stats->num_queries,
108 (unsigned)(stats->num_queries -
109 stats->num_queries_missed_cache),
110 (unsigned)stats->num_queries_missed_cache,
111 (unsigned)stats->num_queries_prefetch,
112 (unsigned)stats->num_queries_ip_ratelimited);
113 log_info("server stats for thread %d: requestlist max %u avg %g "
114 "exceeded %u jostled %u", threadnum,
115 (unsigned)stats->max_query_list_size,
116 (stats->num_queries_missed_cache+stats->num_queries_prefetch)?
117 (double)stats->sum_query_list_size/
118 (stats->num_queries_missed_cache+
119 stats->num_queries_prefetch) : 0.0,
120 (unsigned)worker->env.mesh->stats_dropped,
121 (unsigned)worker->env.mesh->stats_jostled);
124 /** get rrsets bogus number from validator */
126 get_rrset_bogus(struct worker* worker)
128 int m = modstack_find(&worker->env.mesh->mods, "validator");
133 ve = (struct val_env*)worker->env.modinfo[m];
134 lock_basic_lock(&ve->bogus_lock);
135 r = ve->num_rrset_bogus;
136 if(!worker->env.cfg->stat_cumulative)
137 ve->num_rrset_bogus = 0;
138 lock_basic_unlock(&ve->bogus_lock);
143 server_stats_compile(struct worker* worker, struct stats_info* s, int reset)
146 struct listen_list* lp;
148 s->svr = worker->stats;
149 s->mesh_num_states = worker->env.mesh->all.count;
150 s->mesh_num_reply_states = worker->env.mesh->num_reply_states;
151 s->mesh_jostled = worker->env.mesh->stats_jostled;
152 s->mesh_dropped = worker->env.mesh->stats_dropped;
153 s->mesh_replies_sent = worker->env.mesh->replies_sent;
154 s->mesh_replies_sum_wait = worker->env.mesh->replies_sum_wait;
155 s->mesh_time_median = timehist_quartile(worker->env.mesh->histogram,
158 /* add in the values from the mesh */
159 s->svr.ans_secure += worker->env.mesh->ans_secure;
160 s->svr.ans_bogus += worker->env.mesh->ans_bogus;
161 s->svr.ans_rcode_nodata += worker->env.mesh->ans_nodata;
163 s->svr.ans_rcode[i] += worker->env.mesh->ans_rcode[i];
164 timehist_export(worker->env.mesh->histogram, s->svr.hist,
166 /* values from outside network */
167 s->svr.unwanted_replies = worker->back->unwanted_replies;
168 s->svr.qtcp_outgoing = worker->back->num_tcp_outgoing;
170 /* get and reset validator rrset bogus number */
171 s->svr.rrset_bogus = get_rrset_bogus(worker);
173 /* get cache sizes */
174 s->svr.msg_cache_count = count_slabhash_entries(worker->env.msg_cache);
175 s->svr.rrset_cache_count = count_slabhash_entries(&worker->env.rrset_cache->table);
176 s->svr.infra_cache_count = count_slabhash_entries(worker->env.infra_cache->hosts);
177 if(worker->env.key_cache)
178 s->svr.key_cache_count = count_slabhash_entries(worker->env.key_cache->slab);
179 else s->svr.key_cache_count = 0;
181 /* get tcp accept usage */
182 s->svr.tcp_accept_usage = 0;
183 for(lp = worker->front->cps; lp; lp = lp->next) {
184 if(lp->com->type == comm_tcp_accept)
185 s->svr.tcp_accept_usage += lp->com->cur_tcp_count;
188 if(reset && !worker->env.cfg->stat_cumulative) {
189 worker_stats_clear(worker);
193 void server_stats_obtain(struct worker* worker, struct worker* who,
194 struct stats_info* s, int reset)
196 uint8_t *reply = NULL;
199 /* just fill it in */
200 server_stats_compile(worker, s, reset);
203 /* communicate over tube */
204 verbose(VERB_ALGO, "write stats cmd");
206 worker_send_cmd(who, worker_cmd_stats);
207 else worker_send_cmd(who, worker_cmd_stats_noreset);
208 verbose(VERB_ALGO, "wait for stats reply");
209 if(!tube_read_msg(worker->cmd, &reply, &len, 0))
210 fatal_exit("failed to read stats over cmd channel");
211 if(len != (uint32_t)sizeof(*s))
212 fatal_exit("stats on cmd channel wrong length %d %d",
213 (int)len, (int)sizeof(*s));
214 memcpy(s, reply, (size_t)len);
218 void server_stats_reply(struct worker* worker, int reset)
221 server_stats_compile(worker, &s, reset);
222 verbose(VERB_ALGO, "write stats replymsg");
223 if(!tube_write_msg(worker->daemon->workers[0]->cmd,
224 (uint8_t*)&s, sizeof(s), 0))
225 fatal_exit("could not write stat values over cmd channel");
228 void server_stats_add(struct stats_info* total, struct stats_info* a)
230 total->svr.num_queries += a->svr.num_queries;
231 total->svr.num_queries_ip_ratelimited += a->svr.num_queries_ip_ratelimited;
232 total->svr.num_queries_missed_cache += a->svr.num_queries_missed_cache;
233 total->svr.num_queries_prefetch += a->svr.num_queries_prefetch;
234 total->svr.sum_query_list_size += a->svr.sum_query_list_size;
235 /* the max size reached is upped to higher of both */
236 if(a->svr.max_query_list_size > total->svr.max_query_list_size)
237 total->svr.max_query_list_size = a->svr.max_query_list_size;
239 if(a->svr.extended) {
241 total->svr.qtype_big += a->svr.qtype_big;
242 total->svr.qclass_big += a->svr.qclass_big;
243 total->svr.qtcp += a->svr.qtcp;
244 total->svr.qtcp_outgoing += a->svr.qtcp_outgoing;
245 total->svr.qipv6 += a->svr.qipv6;
246 total->svr.qbit_QR += a->svr.qbit_QR;
247 total->svr.qbit_AA += a->svr.qbit_AA;
248 total->svr.qbit_TC += a->svr.qbit_TC;
249 total->svr.qbit_RD += a->svr.qbit_RD;
250 total->svr.qbit_RA += a->svr.qbit_RA;
251 total->svr.qbit_Z += a->svr.qbit_Z;
252 total->svr.qbit_AD += a->svr.qbit_AD;
253 total->svr.qbit_CD += a->svr.qbit_CD;
254 total->svr.qEDNS += a->svr.qEDNS;
255 total->svr.qEDNS_DO += a->svr.qEDNS_DO;
256 total->svr.ans_rcode_nodata += a->svr.ans_rcode_nodata;
257 total->svr.zero_ttl_responses += a->svr.zero_ttl_responses;
258 total->svr.ans_secure += a->svr.ans_secure;
259 total->svr.ans_bogus += a->svr.ans_bogus;
260 total->svr.rrset_bogus += a->svr.rrset_bogus;
261 total->svr.unwanted_replies += a->svr.unwanted_replies;
262 total->svr.unwanted_queries += a->svr.unwanted_queries;
263 total->svr.tcp_accept_usage += a->svr.tcp_accept_usage;
264 for(i=0; i<STATS_QTYPE_NUM; i++)
265 total->svr.qtype[i] += a->svr.qtype[i];
266 for(i=0; i<STATS_QCLASS_NUM; i++)
267 total->svr.qclass[i] += a->svr.qclass[i];
268 for(i=0; i<STATS_OPCODE_NUM; i++)
269 total->svr.qopcode[i] += a->svr.qopcode[i];
270 for(i=0; i<STATS_RCODE_NUM; i++)
271 total->svr.ans_rcode[i] += a->svr.ans_rcode[i];
272 for(i=0; i<NUM_BUCKETS_HIST; i++)
273 total->svr.hist[i] += a->svr.hist[i];
276 total->mesh_num_states += a->mesh_num_states;
277 total->mesh_num_reply_states += a->mesh_num_reply_states;
278 total->mesh_jostled += a->mesh_jostled;
279 total->mesh_dropped += a->mesh_dropped;
280 total->mesh_replies_sent += a->mesh_replies_sent;
281 timeval_add(&total->mesh_replies_sum_wait, &a->mesh_replies_sum_wait);
282 /* the medians are averaged together, this is not as accurate as
283 * taking the median over all of the data, but is good and fast
284 * added up here, division later*/
285 total->mesh_time_median += a->mesh_time_median;
288 void server_stats_insquery(struct server_stats* stats, struct comm_point* c,
289 uint16_t qtype, uint16_t qclass, struct edns_data* edns,
290 struct comm_reply* repinfo)
292 uint16_t flags = sldns_buffer_read_u16_at(c->buffer, 2);
293 if(qtype < STATS_QTYPE_NUM)
294 stats->qtype[qtype]++;
295 else stats->qtype_big++;
296 if(qclass < STATS_QCLASS_NUM)
297 stats->qclass[qclass]++;
298 else stats->qclass_big++;
299 stats->qopcode[ LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) ]++;
300 if(c->type != comm_udp)
302 if(repinfo && addr_is_ip6(&repinfo->addr, repinfo->addrlen))
320 if(edns->edns_present) {
322 if( (edns->bits & EDNS_DO) )
327 void server_stats_insrcode(struct server_stats* stats, sldns_buffer* buf)
329 if(stats->extended && sldns_buffer_limit(buf) != 0) {
330 int r = (int)LDNS_RCODE_WIRE( sldns_buffer_begin(buf) );
331 stats->ans_rcode[r] ++;
332 if(r == 0 && LDNS_ANCOUNT( sldns_buffer_begin(buf) ) == 0)
333 stats->ans_rcode_nodata ++;