blob: 7359c2574da09370a2eb71baa5fa677bdb044197 [file] [log] [blame]
Willy Tarreaubaaee002006-06-26 02:48:02 +02001/*
2 * Health-checks functions.
3 *
Willy Tarreau26c25062009-03-08 09:38:41 +01004 * Copyright 2000-2009 Willy Tarreau <w@1wt.eu>
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +02005 * Copyright 2007-2009 Krzysztof Piotr Oledzki <ole@ans.pl>
Willy Tarreaubaaee002006-06-26 02:48:02 +02006 *
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
11 *
12 */
13
Willy Tarreaub8816082008-01-18 12:18:15 +010014#include <assert.h>
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +020015#include <ctype.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020016#include <errno.h>
Simon Horman0ba0e4a2015-01-30 11:23:00 +090017#include <stdarg.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020018#include <stdio.h>
Krzysztof Oledzkib304dc72007-10-14 23:40:01 +020019#include <stdlib.h>
Willy Tarreau2dd0d472006-06-29 17:53:05 +020020#include <string.h>
Krzysztof Oledzkib304dc72007-10-14 23:40:01 +020021#include <time.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020022#include <unistd.h>
Willy Tarreau9f6dc722019-03-01 11:15:10 +010023#include <sys/resource.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020024#include <sys/socket.h>
Dmitry Sivachenkocaf58982009-08-24 15:11:06 +040025#include <sys/types.h>
Simon Horman98637e52014-06-20 12:30:16 +090026#include <sys/wait.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020027#include <netinet/in.h>
Willy Tarreau1274bc42009-07-15 07:16:31 +020028#include <netinet/tcp.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020029#include <arpa/inet.h>
30
Willy Tarreau122eba92020-06-04 10:15:32 +020031#include <haproxy/action.h>
Willy Tarreau4c7e4b72020-05-27 12:58:42 +020032#include <haproxy/api.h>
Willy Tarreaub2551052020-06-09 09:07:15 +020033#include <haproxy/arg.h>
Willy Tarreau6be78492020-06-05 00:00:29 +020034#include <haproxy/cfgparse.h>
Willy Tarreau4aa573d2020-06-04 18:21:56 +020035#include <haproxy/check.h>
Willy Tarreauc13ed532020-06-02 10:22:45 +020036#include <haproxy/chunk.h>
Willy Tarreau7c18b542020-06-11 09:23:02 +020037#include <haproxy/dgram.h>
Christopher Faulet6b0a0fb2022-04-04 11:29:28 +020038#include <haproxy/dynbuf.h>
Willy Tarreaubcc67332020-06-05 15:31:31 +020039#include <haproxy/extcheck.h>
Willy Tarreaub2551052020-06-09 09:07:15 +020040#include <haproxy/fd.h>
41#include <haproxy/global.h>
42#include <haproxy/h1.h>
Willy Tarreaucd72d8c2020-06-02 19:11:26 +020043#include <haproxy/http.h>
Willy Tarreau87735332020-06-04 09:08:41 +020044#include <haproxy/http_htx.h>
Willy Tarreau16f958c2020-06-03 08:44:35 +020045#include <haproxy/htx.h>
Willy Tarreaub2551052020-06-09 09:07:15 +020046#include <haproxy/istbuf.h>
47#include <haproxy/list.h>
Willy Tarreauaeed4a82020-06-04 22:01:04 +020048#include <haproxy/log.h>
Willy Tarreaub2551052020-06-09 09:07:15 +020049#include <haproxy/mailers.h>
50#include <haproxy/port_range.h>
51#include <haproxy/proto_tcp.h>
Willy Tarreaub2551052020-06-09 09:07:15 +020052#include <haproxy/protocol.h>
Willy Tarreaua264d962020-06-04 22:29:18 +020053#include <haproxy/proxy.h>
Willy Tarreaub2551052020-06-09 09:07:15 +020054#include <haproxy/queue.h>
55#include <haproxy/regex.h>
Emeric Brunc9437992021-02-12 19:42:55 +010056#include <haproxy/resolvers.h>
Willy Tarreaub2551052020-06-09 09:07:15 +020057#include <haproxy/sample.h>
Willy Tarreau1e56f922020-06-04 23:20:13 +020058#include <haproxy/server.h>
Willy Tarreau209108d2020-06-04 20:30:20 +020059#include <haproxy/ssl_sock.h>
Willy Tarreau2eec9b52020-06-04 19:58:55 +020060#include <haproxy/stats-t.h>
Willy Tarreaucea0e1b2020-06-04 17:25:40 +020061#include <haproxy/task.h>
Willy Tarreau51cd5952020-06-05 12:25:38 +020062#include <haproxy/tcpcheck.h>
Willy Tarreaub2551052020-06-09 09:07:15 +020063#include <haproxy/thread.h>
64#include <haproxy/time.h>
65#include <haproxy/tools.h>
Christopher Faulet147b8c92021-04-10 09:00:38 +020066#include <haproxy/trace.h>
Willy Tarreaua1718922020-06-04 16:25:31 +020067#include <haproxy/vars.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020068
Christopher Faulet147b8c92021-04-10 09:00:38 +020069/* trace source and events */
70static void check_trace(enum trace_level level, uint64_t mask,
71 const struct trace_source *src,
72 const struct ist where, const struct ist func,
73 const void *a1, const void *a2, const void *a3, const void *a4);
74
75/* The event representation is split like this :
76 * check - check
77 *
78 * CHECK_EV_* macros are defined in <haproxy/check.h>
79 */
80static const struct trace_event check_trace_events[] = {
81 { .mask = CHK_EV_TASK_WAKE, .name = "task_wake", .desc = "Check task woken up" },
82 { .mask = CHK_EV_HCHK_START, .name = "hchck_start", .desc = "Health-check started" },
83 { .mask = CHK_EV_HCHK_WAKE, .name = "hchck_wake", .desc = "Health-check woken up" },
84 { .mask = CHK_EV_HCHK_RUN, .name = "hchck_run", .desc = "Health-check running" },
85 { .mask = CHK_EV_HCHK_END, .name = "hchck_end", .desc = "Health-check terminated" },
86 { .mask = CHK_EV_HCHK_SUCC, .name = "hchck_succ", .desc = "Health-check success" },
87 { .mask = CHK_EV_HCHK_ERR, .name = "hchck_err", .desc = "Health-check failure" },
88
89 { .mask = CHK_EV_TCPCHK_EVAL, .name = "tcp_check_eval", .desc = "tcp-check rules evaluation" },
90 { .mask = CHK_EV_TCPCHK_ERR, .name = "tcp_check_err", .desc = "tcp-check evaluation error" },
91 { .mask = CHK_EV_TCPCHK_CONN, .name = "tcp_check_conn", .desc = "tcp-check connection rule" },
92 { .mask = CHK_EV_TCPCHK_SND, .name = "tcp_check_send", .desc = "tcp-check send rule" },
93 { .mask = CHK_EV_TCPCHK_EXP, .name = "tcp_check_expect", .desc = "tcp-check expect rule" },
94 { .mask = CHK_EV_TCPCHK_ACT, .name = "tcp_check_action", .desc = "tcp-check action rule" },
95
96 { .mask = CHK_EV_RX_DATA, .name = "rx_data", .desc = "receipt of data" },
97 { .mask = CHK_EV_RX_BLK, .name = "rx_blk", .desc = "receipt blocked" },
98 { .mask = CHK_EV_RX_ERR, .name = "rx_err", .desc = "receipt error" },
99
100 { .mask = CHK_EV_TX_DATA, .name = "tx_data", .desc = "transmission of data" },
101 { .mask = CHK_EV_TX_BLK, .name = "tx_blk", .desc = "transmission blocked" },
102 { .mask = CHK_EV_TX_ERR, .name = "tx_err", .desc = "transmission error" },
103
104 {}
105};
106
107static const struct name_desc check_trace_lockon_args[4] = {
108 /* arg1 */ { /* already used by the check */ },
109 /* arg2 */ { },
110 /* arg3 */ { },
111 /* arg4 */ { }
112};
113
114static const struct name_desc check_trace_decoding[] = {
115#define CHK_VERB_CLEAN 1
116 { .name="clean", .desc="only user-friendly stuff, generally suitable for level \"user\"" },
117#define CHK_VERB_MINIMAL 2
Willy Tarreau4596fe22022-05-17 19:07:51 +0200118 { .name="minimal", .desc="report info on streams and connectors" },
Christopher Faulet147b8c92021-04-10 09:00:38 +0200119#define CHK_VERB_SIMPLE 3
120 { .name="simple", .desc="add info on request and response channels" },
121#define CHK_VERB_ADVANCED 4
122 { .name="advanced", .desc="add info on channel's buffer for data and developer levels only" },
123#define CHK_VERB_COMPLETE 5
124 { .name="complete", .desc="add info on channel's buffer" },
125 { /* end */ }
126};
127
128struct trace_source trace_check = {
129 .name = IST("check"),
130 .desc = "Health-check",
131 .arg_def = TRC_ARG1_CHK, // TRACE()'s first argument is always a stream
132 .default_cb = check_trace,
133 .known_events = check_trace_events,
134 .lockon_args = check_trace_lockon_args,
135 .decoding = check_trace_decoding,
136 .report_events = ~0, // report everything by default
137};
138
139#define TRACE_SOURCE &trace_check
140INITCALL1(STG_REGISTER, trace_register_source, TRACE_SOURCE);
141
Olivier Houchard9130a962017-10-17 17:33:43 +0200142
Gaetan Rivet05d692d2020-02-14 17:42:54 +0100143/* Dummy frontend used to create all checks sessions. */
Willy Tarreau51cd5952020-06-05 12:25:38 +0200144struct proxy checks_fe;
Christopher Faulet31dff9b2017-10-23 15:45:20 +0200145
Christopher Faulet147b8c92021-04-10 09:00:38 +0200146
147static inline void check_trace_buf(const struct buffer *buf, size_t ofs, size_t len)
148{
149 size_t block1, block2;
150 int line, ptr, newptr;
151
152 block1 = b_contig_data(buf, ofs);
153 block2 = 0;
154 if (block1 > len)
155 block1 = len;
156 block2 = len - block1;
157
158 ofs = b_peek_ofs(buf, ofs);
159
160 line = 0;
161 ptr = ofs;
162 while (ptr < ofs + block1) {
163 newptr = dump_text_line(&trace_buf, b_orig(buf), b_size(buf), ofs + block1, &line, ptr);
164 if (newptr == ptr)
165 break;
166 ptr = newptr;
167 }
168
169 line = ptr = 0;
170 while (ptr < block2) {
171 newptr = dump_text_line(&trace_buf, b_orig(buf), b_size(buf), block2, &line, ptr);
172 if (newptr == ptr)
173 break;
174 ptr = newptr;
175 }
176}
177
178/* trace source and events */
179static void check_trace(enum trace_level level, uint64_t mask,
180 const struct trace_source *src,
181 const struct ist where, const struct ist func,
182 const void *a1, const void *a2, const void *a3, const void *a4)
183{
184 const struct check *check = a1;
185 const struct server *srv = (check ? check->server : NULL);
186 const size_t *val = a4;
187 const char *res;
188
189 if (!check || src->verbosity < CHK_VERB_CLEAN)
190 return;
191
Christopher Faulet4f1825c2022-06-08 09:12:58 +0200192 if (srv) {
193 chunk_appendf(&trace_buf, " : [%c] SRV=%s",
194 ((check->type == PR_O2_EXT_CHK) ? 'E' : (check->state & CHK_ST_AGENT ? 'A' : 'H')),
195 srv->id);
Christopher Faulet147b8c92021-04-10 09:00:38 +0200196
Christopher Faulet4f1825c2022-06-08 09:12:58 +0200197 chunk_appendf(&trace_buf, " status=%d/%d %s",
198 (check->health >= check->rise) ? check->health - check->rise + 1 : check->health,
199 (check->health >= check->rise) ? check->fall : check->rise,
200 (check->health >= check->rise) ? (srv->uweight ? "UP" : "DRAIN") : "DOWN");
201 }
202 else
203 chunk_appendf(&trace_buf, " : [EMAIL]");
Christopher Faulet147b8c92021-04-10 09:00:38 +0200204
205 switch (check->result) {
206 case CHK_RES_NEUTRAL: res = "-"; break;
207 case CHK_RES_FAILED: res = "FAIL"; break;
208 case CHK_RES_PASSED: res = "PASS"; break;
209 case CHK_RES_CONDPASS: res = "COND"; break;
210 default: res = "UNK"; break;
211 }
212
213 if (src->verbosity == CHK_VERB_CLEAN)
214 return;
215
216 chunk_appendf(&trace_buf, " - last=%s(%d)/%s(%d)",
217 get_check_status_info(check->status), check->status,
218 res, check->result);
219
220 /* Display the value to the 4th argument (level > STATE) */
221 if (src->level > TRACE_LEVEL_STATE && val)
222 chunk_appendf(&trace_buf, " - VAL=%lu", (long)*val);
223
224 chunk_appendf(&trace_buf, " check=%p(0x%08x)", check, check->state);
225
226 if (src->verbosity == CHK_VERB_MINIMAL)
227 return;
228
229
Willy Tarreaubde14ad2022-05-27 10:04:04 +0200230 if (check->sc) {
231 struct connection *conn = sc_conn(check->sc);
Christopher Faulet0256da12021-12-15 09:50:17 +0100232
233 chunk_appendf(&trace_buf, " - conn=%p(0x%08x)", conn, conn ? conn->flags : 0);
Willy Tarreaubde14ad2022-05-27 10:04:04 +0200234 chunk_appendf(&trace_buf, " sc=%p(0x%08x)", check->sc, check->sc->flags);
Christopher Faulet147b8c92021-04-10 09:00:38 +0200235 }
236
237 if (mask & CHK_EV_TCPCHK) {
238 const char *type;
239
240 switch (check->tcpcheck_rules->flags & TCPCHK_RULES_PROTO_CHK) {
241 case TCPCHK_RULES_PGSQL_CHK: type = "PGSQL"; break;
242 case TCPCHK_RULES_REDIS_CHK: type = "REDIS"; break;
243 case TCPCHK_RULES_SMTP_CHK: type = "SMTP"; break;
244 case TCPCHK_RULES_HTTP_CHK: type = "HTTP"; break;
245 case TCPCHK_RULES_MYSQL_CHK: type = "MYSQL"; break;
246 case TCPCHK_RULES_LDAP_CHK: type = "LDAP"; break;
247 case TCPCHK_RULES_SSL3_CHK: type = "SSL3"; break;
248 case TCPCHK_RULES_AGENT_CHK: type = "AGENT"; break;
249 case TCPCHK_RULES_SPOP_CHK: type = "SPOP"; break;
250 case TCPCHK_RULES_TCP_CHK: type = "TCP"; break;
251 default: type = "???"; break;
252 }
253 if (check->current_step)
254 chunk_appendf(&trace_buf, " - tcp-check=(%s,%d)", type, tcpcheck_get_step_id(check, NULL));
255 else
256 chunk_appendf(&trace_buf, " - tcp-check=(%s,-)", type);
257 }
258
259 /* Display bi and bo buffer info (level > USER & verbosity > SIMPLE) */
260 if (src->level > TRACE_LEVEL_USER) {
261 const struct buffer *buf = NULL;
262
263 chunk_appendf(&trace_buf, " bi=%u@%p+%u/%u",
264 (unsigned int)b_data(&check->bi), b_orig(&check->bi),
265 (unsigned int)b_head_ofs(&check->bi), (unsigned int)b_size(&check->bi));
266 chunk_appendf(&trace_buf, " bo=%u@%p+%u/%u",
267 (unsigned int)b_data(&check->bo), b_orig(&check->bo),
268 (unsigned int)b_head_ofs(&check->bo), (unsigned int)b_size(&check->bo));
269
270 if (src->verbosity >= CHK_VERB_ADVANCED && (mask & (CHK_EV_RX)))
271 buf = (b_is_null(&check->bi) ? NULL : &check->bi);
272 else if (src->verbosity >= CHK_VERB_ADVANCED && (mask & (CHK_EV_TX)))
273 buf = (b_is_null(&check->bo) ? NULL : &check->bo);
274
275 if (buf) {
276 if ((check->tcpcheck_rules->flags & TCPCHK_RULES_PROTO_CHK) == TCPCHK_RULES_HTTP_CHK) {
277 int full = (src->verbosity == CHK_VERB_COMPLETE);
278
279 chunk_memcat(&trace_buf, "\n\t", 2);
280 htx_dump(&trace_buf, htxbuf(buf), full);
281 }
282 else {
283 int max = ((src->verbosity == CHK_VERB_COMPLETE) ? 1024 : 256);
284
285 chunk_memcat(&trace_buf, "\n", 1);
286 if (b_data(buf) > max) {
287 check_trace_buf(buf, 0, max);
288 chunk_memcat(&trace_buf, " ...\n", 6);
289 }
290 else
291 check_trace_buf(buf, 0, b_data(buf));
292 }
293
294 }
295 }
296
297}
298
299
Christopher Faulet61cc8522020-04-20 14:54:42 +0200300/**************************************************************************/
301/************************ Handle check results ****************************/
302/**************************************************************************/
303struct check_status {
304 short result; /* one of SRV_CHK_* */
305 char *info; /* human readable short info */
306 char *desc; /* long description */
307};
308
309struct analyze_status {
310 char *desc; /* description */
311 unsigned char lr[HANA_OBS_SIZE]; /* result for l4/l7: 0 = ignore, 1 - error, 2 - OK */
312};
313
Simon Horman63a4a822012-03-19 07:24:41 +0900314static const struct check_status check_statuses[HCHK_STATUS_SIZE] = {
Willy Tarreau6aaa1b82013-12-11 17:09:34 +0100315 [HCHK_STATUS_UNKNOWN] = { CHK_RES_UNKNOWN, "UNK", "Unknown" },
316 [HCHK_STATUS_INI] = { CHK_RES_UNKNOWN, "INI", "Initializing" },
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200317 [HCHK_STATUS_START] = { /* SPECIAL STATUS*/ },
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +0200318
Willy Tarreau23964182014-05-20 20:56:30 +0200319 /* Below we have finished checks */
320 [HCHK_STATUS_CHECKED] = { CHK_RES_NEUTRAL, "CHECKED", "No status change" },
Willy Tarreau6aaa1b82013-12-11 17:09:34 +0100321 [HCHK_STATUS_HANA] = { CHK_RES_FAILED, "HANA", "Health analyze" },
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100322
Willy Tarreau6aaa1b82013-12-11 17:09:34 +0100323 [HCHK_STATUS_SOCKERR] = { CHK_RES_FAILED, "SOCKERR", "Socket error" },
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +0200324
Willy Tarreau6aaa1b82013-12-11 17:09:34 +0100325 [HCHK_STATUS_L4OK] = { CHK_RES_PASSED, "L4OK", "Layer4 check passed" },
326 [HCHK_STATUS_L4TOUT] = { CHK_RES_FAILED, "L4TOUT", "Layer4 timeout" },
327 [HCHK_STATUS_L4CON] = { CHK_RES_FAILED, "L4CON", "Layer4 connection problem" },
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +0200328
Willy Tarreau6aaa1b82013-12-11 17:09:34 +0100329 [HCHK_STATUS_L6OK] = { CHK_RES_PASSED, "L6OK", "Layer6 check passed" },
330 [HCHK_STATUS_L6TOUT] = { CHK_RES_FAILED, "L6TOUT", "Layer6 timeout" },
331 [HCHK_STATUS_L6RSP] = { CHK_RES_FAILED, "L6RSP", "Layer6 invalid response" },
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +0200332
Willy Tarreau6aaa1b82013-12-11 17:09:34 +0100333 [HCHK_STATUS_L7TOUT] = { CHK_RES_FAILED, "L7TOUT", "Layer7 timeout" },
334 [HCHK_STATUS_L7RSP] = { CHK_RES_FAILED, "L7RSP", "Layer7 invalid response" },
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +0200335
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200336 [HCHK_STATUS_L57DATA] = { /* DUMMY STATUS */ },
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +0200337
Willy Tarreau6aaa1b82013-12-11 17:09:34 +0100338 [HCHK_STATUS_L7OKD] = { CHK_RES_PASSED, "L7OK", "Layer7 check passed" },
339 [HCHK_STATUS_L7OKCD] = { CHK_RES_CONDPASS, "L7OKC", "Layer7 check conditionally passed" },
340 [HCHK_STATUS_L7STS] = { CHK_RES_FAILED, "L7STS", "Layer7 wrong status" },
Simon Horman98637e52014-06-20 12:30:16 +0900341
342 [HCHK_STATUS_PROCERR] = { CHK_RES_FAILED, "PROCERR", "External check error" },
343 [HCHK_STATUS_PROCTOUT] = { CHK_RES_FAILED, "PROCTOUT", "External check timeout" },
Cyril Bonté77010d82014-08-07 01:55:37 +0200344 [HCHK_STATUS_PROCOK] = { CHK_RES_PASSED, "PROCOK", "External check passed" },
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +0200345};
346
Simon Horman63a4a822012-03-19 07:24:41 +0900347static const struct analyze_status analyze_statuses[HANA_STATUS_SIZE] = { /* 0: ignore, 1: error, 2: OK */
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100348 [HANA_STATUS_UNKNOWN] = { "Unknown", { 0, 0 }},
349
350 [HANA_STATUS_L4_OK] = { "L4 successful connection", { 2, 0 }},
351 [HANA_STATUS_L4_ERR] = { "L4 unsuccessful connection", { 1, 1 }},
352
353 [HANA_STATUS_HTTP_OK] = { "Correct http response", { 0, 2 }},
354 [HANA_STATUS_HTTP_STS] = { "Wrong http response", { 0, 1 }},
355 [HANA_STATUS_HTTP_HDRRSP] = { "Invalid http response (headers)", { 0, 1 }},
356 [HANA_STATUS_HTTP_RSP] = { "Invalid http response", { 0, 1 }},
357
358 [HANA_STATUS_HTTP_READ_ERROR] = { "Read error (http)", { 0, 1 }},
359 [HANA_STATUS_HTTP_READ_TIMEOUT] = { "Read timeout (http)", { 0, 1 }},
360 [HANA_STATUS_HTTP_BROKEN_PIPE] = { "Close from server (http)", { 0, 1 }},
361};
362
Willy Tarreauc8dc20a2019-12-27 12:03:27 +0100363/* checks if <err> is a real error for errno or one that can be ignored, and
364 * return 0 for these ones or <err> for real ones.
365 */
366static inline int unclean_errno(int err)
367{
Willy Tarreauacef5e22022-04-25 20:32:15 +0200368 if (err == EAGAIN || err == EWOULDBLOCK || err == EINPROGRESS ||
Willy Tarreauc8dc20a2019-12-27 12:03:27 +0100369 err == EISCONN || err == EALREADY)
370 return 0;
371 return err;
372}
373
Christopher Faulet7aa32712021-02-01 13:11:50 +0100374/* Converts check_status code to result code */
375short get_check_status_result(short check_status)
376{
377 if (check_status < HCHK_STATUS_SIZE)
378 return check_statuses[check_status].result;
379 else
380 return check_statuses[HCHK_STATUS_UNKNOWN].result;
381}
382
Christopher Faulet61cc8522020-04-20 14:54:42 +0200383/* Converts check_status code to description */
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +0200384const char *get_check_status_description(short check_status) {
385
386 const char *desc;
387
388 if (check_status < HCHK_STATUS_SIZE)
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200389 desc = check_statuses[check_status].desc;
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +0200390 else
391 desc = NULL;
392
393 if (desc && *desc)
394 return desc;
395 else
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200396 return check_statuses[HCHK_STATUS_UNKNOWN].desc;
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +0200397}
398
Christopher Faulet61cc8522020-04-20 14:54:42 +0200399/* Converts check_status code to short info */
William Dauchyb26122b2021-02-14 22:26:23 +0100400const char *get_check_status_info(short check_status)
401{
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +0200402 const char *info;
403
404 if (check_status < HCHK_STATUS_SIZE)
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200405 info = check_statuses[check_status].info;
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +0200406 else
407 info = NULL;
408
409 if (info && *info)
410 return info;
411 else
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200412 return check_statuses[HCHK_STATUS_UNKNOWN].info;
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +0200413}
414
Christopher Faulet61cc8522020-04-20 14:54:42 +0200415/* Convert analyze_status to description */
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100416const char *get_analyze_status(short analyze_status) {
417
418 const char *desc;
419
420 if (analyze_status < HANA_STATUS_SIZE)
421 desc = analyze_statuses[analyze_status].desc;
422 else
423 desc = NULL;
424
425 if (desc && *desc)
426 return desc;
427 else
428 return analyze_statuses[HANA_STATUS_UNKNOWN].desc;
429}
430
Aurelien DARRAGONf3b48a82023-04-14 18:07:09 +0200431/* append check info to buffer msg */
432void check_append_info(struct buffer *msg, struct check *check)
433{
434 if (!check)
435 return;
436 chunk_appendf(msg, ", reason: %s", get_check_status_description(check->status));
437
438 if (check->status >= HCHK_STATUS_L57DATA)
439 chunk_appendf(msg, ", code: %d", check->code);
440
441 if (check->desc[0]) {
442 struct buffer src;
443
444 chunk_appendf(msg, ", info: \"");
445
446 chunk_initlen(&src, check->desc, 0, strlen(check->desc));
447 chunk_asciiencode(msg, &src, '"');
448
449 chunk_appendf(msg, "\"");
450 }
451
452 if (check->duration >= 0)
453 chunk_appendf(msg, ", check duration: %ldms", check->duration);
454}
455
Christopher Faulet61cc8522020-04-20 14:54:42 +0200456/* Sets check->status, update check->duration and fill check->result with an
457 * adequate CHK_RES_* value. The new check->health is computed based on the
458 * result.
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200459 *
Christopher Faulet61cc8522020-04-20 14:54:42 +0200460 * Shows information in logs about failed health check if server is UP or
461 * succeeded health checks if server is DOWN.
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +0200462 */
Willy Tarreau51cd5952020-06-05 12:25:38 +0200463void set_server_check_status(struct check *check, short status, const char *desc)
Willy Tarreau19d14ef2012-10-29 16:51:55 +0100464{
Simon Horman4a741432013-02-23 15:35:38 +0900465 struct server *s = check->server;
Willy Tarreaubef1b322014-05-13 21:01:39 +0200466 short prev_status = check->status;
Aurelien DARRAGONdcbc2d22023-03-30 10:19:08 +0200467 int report = (status != prev_status) ? 1 : 0;
Simon Horman4a741432013-02-23 15:35:38 +0900468
Christopher Faulet147b8c92021-04-10 09:00:38 +0200469 TRACE_POINT(CHK_EV_HCHK_RUN, check);
470
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200471 if (status == HCHK_STATUS_START) {
Willy Tarreau6aaa1b82013-12-11 17:09:34 +0100472 check->result = CHK_RES_UNKNOWN; /* no result yet */
Simon Horman4a741432013-02-23 15:35:38 +0900473 check->desc[0] = '\0';
Willy Tarreau69530f52023-04-28 09:16:15 +0200474 check->start = now_ns;
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200475 return;
476 }
477
Simon Horman4a741432013-02-23 15:35:38 +0900478 if (!check->status)
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200479 return;
480
Krzysztof Piotr Oledzkif7089f52009-10-10 21:06:49 +0200481 if (desc && *desc) {
Simon Horman4a741432013-02-23 15:35:38 +0900482 strncpy(check->desc, desc, HCHK_DESC_LEN-1);
483 check->desc[HCHK_DESC_LEN-1] = '\0';
Krzysztof Piotr Oledzkif7089f52009-10-10 21:06:49 +0200484 } else
Simon Horman4a741432013-02-23 15:35:38 +0900485 check->desc[0] = '\0';
Krzysztof Piotr Oledzkif7089f52009-10-10 21:06:49 +0200486
Simon Horman4a741432013-02-23 15:35:38 +0900487 check->status = status;
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200488 if (check_statuses[status].result)
Simon Horman4a741432013-02-23 15:35:38 +0900489 check->result = check_statuses[status].result;
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +0200490
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100491 if (status == HCHK_STATUS_HANA)
Simon Horman4a741432013-02-23 15:35:38 +0900492 check->duration = -1;
Willy Tarreaue8e47122023-04-28 14:39:50 +0200493 else if (check->start) {
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200494 /* set_server_check_status() may be called more than once */
Willy Tarreau69530f52023-04-28 09:16:15 +0200495 check->duration = ns_to_ms(now_ns - check->start);
Willy Tarreaue8e47122023-04-28 14:39:50 +0200496 check->start = 0;
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200497 }
498
Willy Tarreau23964182014-05-20 20:56:30 +0200499 /* no change is expected if no state change occurred */
500 if (check->result == CHK_RES_NEUTRAL)
501 return;
502
Olivier Houchard0923fa42019-01-11 18:43:04 +0100503 /* If the check was really just sending a mail, it won't have an
504 * associated server, so we're done now.
505 */
506 if (!s)
507 return;
Christopher Faulet147b8c92021-04-10 09:00:38 +0200508
Willy Tarreau7b1d47c2014-05-20 14:55:13 +0200509 switch (check->result) {
510 case CHK_RES_FAILED:
Willy Tarreau12634e12014-05-23 11:32:36 +0200511 /* Failure to connect to the agent as a secondary check should not
512 * cause the server to be marked down.
513 */
514 if ((!(check->state & CHK_ST_AGENT) ||
Simon Hormaneaabd522015-02-26 11:26:17 +0900515 (check->status >= HCHK_STATUS_L57DATA)) &&
Christopher Fauletb119a792018-05-02 12:12:45 +0200516 (check->health > 0)) {
Willy Tarreau4781b152021-04-06 13:53:36 +0200517 _HA_ATOMIC_INC(&s->counters.failed_checks);
Willy Tarreau7b1d47c2014-05-20 14:55:13 +0200518 report = 1;
519 check->health--;
520 if (check->health < check->rise)
521 check->health = 0;
522 }
523 break;
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200524
Willy Tarreau7b1d47c2014-05-20 14:55:13 +0200525 case CHK_RES_PASSED:
Christopher Faulet1e527cb2020-11-20 18:13:02 +0100526 case CHK_RES_CONDPASS:
527 if (check->health < check->rise + check->fall - 1) {
Willy Tarreau7b1d47c2014-05-20 14:55:13 +0200528 report = 1;
529 check->health++;
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200530
Willy Tarreau7b1d47c2014-05-20 14:55:13 +0200531 if (check->health >= check->rise)
532 check->health = check->rise + check->fall - 1; /* OK now */
533 }
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200534
Willy Tarreau7b1d47c2014-05-20 14:55:13 +0200535 /* clear consecutive_errors if observing is enabled */
536 if (s->onerror)
Aurelien DARRAGON81b7c952022-12-07 14:27:42 +0100537 HA_ATOMIC_STORE(&s->consecutive_errors, 0);
Willy Tarreau7b1d47c2014-05-20 14:55:13 +0200538 break;
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100539
Willy Tarreau7b1d47c2014-05-20 14:55:13 +0200540 default:
541 break;
542 }
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200543
Aurelien DARRAGONdcbc2d22023-03-30 10:19:08 +0200544 if (report)
545 srv_event_hdl_publish_check(s, check);
546
547 if (s->proxy->options2 & PR_O2_LOGHCHKS && report) {
Willy Tarreau7b1d47c2014-05-20 14:55:13 +0200548 chunk_printf(&trash,
Willy Tarreau12634e12014-05-23 11:32:36 +0200549 "%s check for %sserver %s/%s %s%s",
550 (check->state & CHK_ST_AGENT) ? "Agent" : "Health",
Willy Tarreauc93cd162014-05-13 15:54:22 +0200551 s->flags & SRV_F_BACKUP ? "backup " : "",
Willy Tarreau19d14ef2012-10-29 16:51:55 +0100552 s->proxy->id, s->id,
Willy Tarreau6aaa1b82013-12-11 17:09:34 +0100553 (check->result == CHK_RES_CONDPASS) ? "conditionally ":"",
Willy Tarreau7b1d47c2014-05-20 14:55:13 +0200554 (check->result >= CHK_RES_PASSED) ? "succeeded" : "failed");
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200555
Aurelien DARRAGONf3b48a82023-04-14 18:07:09 +0200556 check_append_info(&trash, check);
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200557
Willy Tarreau19d14ef2012-10-29 16:51:55 +0100558 chunk_appendf(&trash, ", status: %d/%d %s",
Willy Tarreau7b1d47c2014-05-20 14:55:13 +0200559 (check->health >= check->rise) ? check->health - check->rise + 1 : check->health,
560 (check->health >= check->rise) ? check->fall : check->rise,
561 (check->health >= check->rise) ? (s->uweight ? "UP" : "DRAIN") : "DOWN");
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200562
Willy Tarreau843b7cb2018-07-13 10:54:26 +0200563 ha_warning("%s.\n", trash.area);
564 send_log(s->proxy, LOG_NOTICE, "%s.\n", trash.area);
565 send_email_alert(s, LOG_INFO, "%s", trash.area);
Krzysztof Piotr Oledzki213014e2009-09-27 15:50:02 +0200566 }
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +0200567}
568
Aurelien DARRAGON1746b562023-04-04 10:17:40 +0200569static inline enum srv_op_st_chg_cause check_notify_cause(struct check *check)
570{
571 struct server *s = check->server;
572
573 /* We only report a cause for the check if we did not do so previously */
574 if (!s->track && !(s->proxy->options2 & PR_O2_LOGHCHKS))
575 return (check->state & CHK_ST_AGENT) ? SRV_OP_STCHGC_AGENT : SRV_OP_STCHGC_HEALTH;
576 return SRV_OP_STCHGC_NONE;
577}
578
Willy Tarreau4eec5472014-05-20 22:32:27 +0200579/* Marks the check <check>'s server down if the current check is already failed
580 * and the server is not down yet nor in maintenance.
Willy Tarreaubaaee002006-06-26 02:48:02 +0200581 */
Willy Tarreaubcc67332020-06-05 15:31:31 +0200582void check_notify_failure(struct check *check)
Willy Tarreaubaaee002006-06-26 02:48:02 +0200583{
Simon Horman4a741432013-02-23 15:35:38 +0900584 struct server *s = check->server;
Simon Hormane0d1bfb2011-06-21 14:34:58 +0900585
Willy Tarreau7b1d47c2014-05-20 14:55:13 +0200586 /* The agent secondary check should only cause a server to be marked
587 * as down if check->status is HCHK_STATUS_L7STS, which indicates
588 * that the agent returned "fail", "stopped" or "down".
589 * The implication here is that failure to connect to the agent
590 * as a secondary check should not cause the server to be marked
591 * down. */
592 if ((check->state & CHK_ST_AGENT) && check->status != HCHK_STATUS_L7STS)
593 return;
594
Willy Tarreau4eec5472014-05-20 22:32:27 +0200595 if (check->health > 0)
596 return;
Krzysztof Piotr Oledzkic8b16fc2008-02-18 01:26:35 +0100597
Christopher Faulet147b8c92021-04-10 09:00:38 +0200598 TRACE_STATE("health-check failed, set server DOWN", CHK_EV_HCHK_END|CHK_EV_HCHK_ERR, check);
Aurelien DARRAGON1746b562023-04-04 10:17:40 +0200599 srv_set_stopped(s, check_notify_cause(check));
Willy Tarreaubaaee002006-06-26 02:48:02 +0200600}
601
Willy Tarreauaf549582014-05-16 17:37:50 +0200602/* Marks the check <check> as valid and tries to set its server up, provided
Willy Tarreau3e048382014-05-21 10:30:54 +0200603 * it isn't in maintenance, it is not tracking a down server and other checks
604 * comply. The rule is simple : by default, a server is up, unless any of the
605 * following conditions is true :
606 * - health check failed (check->health < rise)
607 * - agent check failed (agent->health < rise)
608 * - the server tracks a down server (track && track->state == STOPPED)
609 * Note that if the server has a slowstart, it will switch to STARTING instead
610 * of RUNNING. Also, only the health checks support the nolb mode, so the
611 * agent's success may not take the server out of this mode.
Willy Tarreauaf549582014-05-16 17:37:50 +0200612 */
Willy Tarreaubcc67332020-06-05 15:31:31 +0200613void check_notify_success(struct check *check)
Willy Tarreauaf549582014-05-16 17:37:50 +0200614{
Simon Horman4a741432013-02-23 15:35:38 +0900615 struct server *s = check->server;
Krzysztof Piotr Oledzkic8b16fc2008-02-18 01:26:35 +0100616
Emeric Brun52a91d32017-08-31 14:41:55 +0200617 if (s->next_admin & SRV_ADMF_MAINT)
Willy Tarreauaf549582014-05-16 17:37:50 +0200618 return;
Cyril Bontécd19e512010-01-31 22:34:03 +0100619
Emeric Brun52a91d32017-08-31 14:41:55 +0200620 if (s->track && s->track->next_state == SRV_ST_STOPPED)
Willy Tarreauaf549582014-05-16 17:37:50 +0200621 return;
Krzysztof Piotr Oledzkic8b16fc2008-02-18 01:26:35 +0100622
Willy Tarreau3e048382014-05-21 10:30:54 +0200623 if ((s->check.state & CHK_ST_ENABLED) && (s->check.health < s->check.rise))
624 return;
Krzysztof Piotr Oledzkic8b16fc2008-02-18 01:26:35 +0100625
Willy Tarreau3e048382014-05-21 10:30:54 +0200626 if ((s->agent.state & CHK_ST_ENABLED) && (s->agent.health < s->agent.rise))
627 return;
Willy Tarreauaf549582014-05-16 17:37:50 +0200628
Emeric Brun52a91d32017-08-31 14:41:55 +0200629 if ((check->state & CHK_ST_AGENT) && s->next_state == SRV_ST_STOPPING)
Willy Tarreau3e048382014-05-21 10:30:54 +0200630 return;
Krzysztof Piotr Oledzkic8b16fc2008-02-18 01:26:35 +0100631
Christopher Faulet147b8c92021-04-10 09:00:38 +0200632 TRACE_STATE("health-check succeeded, set server RUNNING", CHK_EV_HCHK_END|CHK_EV_HCHK_SUCC, check);
Aurelien DARRAGON1746b562023-04-04 10:17:40 +0200633 srv_set_running(s, check_notify_cause(check));
Krzysztof Piotr Oledzkic8b16fc2008-02-18 01:26:35 +0100634}
635
Willy Tarreaudb58b792014-05-21 13:57:23 +0200636/* Marks the check <check> as valid and tries to set its server into stopping mode
637 * if it was running or starting, and provided it isn't in maintenance and other
638 * checks comply. The conditions for the server to be marked in stopping mode are
639 * the same as for it to be turned up. Also, only the health checks support the
640 * nolb mode.
Willy Tarreauaf549582014-05-16 17:37:50 +0200641 */
Willy Tarreaubcc67332020-06-05 15:31:31 +0200642void check_notify_stopping(struct check *check)
Willy Tarreauaf549582014-05-16 17:37:50 +0200643{
Simon Horman4a741432013-02-23 15:35:38 +0900644 struct server *s = check->server;
Krzysztof Piotr Oledzkic8b16fc2008-02-18 01:26:35 +0100645
Emeric Brun52a91d32017-08-31 14:41:55 +0200646 if (s->next_admin & SRV_ADMF_MAINT)
Willy Tarreauaf549582014-05-16 17:37:50 +0200647 return;
648
Willy Tarreaudb58b792014-05-21 13:57:23 +0200649 if (check->state & CHK_ST_AGENT)
650 return;
Krzysztof Piotr Oledzkic8b16fc2008-02-18 01:26:35 +0100651
Emeric Brun52a91d32017-08-31 14:41:55 +0200652 if (s->track && s->track->next_state == SRV_ST_STOPPED)
Willy Tarreaudb58b792014-05-21 13:57:23 +0200653 return;
Krzysztof Piotr Oledzkic8b16fc2008-02-18 01:26:35 +0100654
Willy Tarreaudb58b792014-05-21 13:57:23 +0200655 if ((s->check.state & CHK_ST_ENABLED) && (s->check.health < s->check.rise))
656 return;
Krzysztof Piotr Oledzkic8b16fc2008-02-18 01:26:35 +0100657
Willy Tarreaudb58b792014-05-21 13:57:23 +0200658 if ((s->agent.state & CHK_ST_ENABLED) && (s->agent.health < s->agent.rise))
659 return;
Krzysztof Piotr Oledzkic8b16fc2008-02-18 01:26:35 +0100660
Christopher Faulet147b8c92021-04-10 09:00:38 +0200661 TRACE_STATE("health-check condionnaly succeeded, set server STOPPING", CHK_EV_HCHK_END|CHK_EV_HCHK_SUCC, check);
Aurelien DARRAGON1746b562023-04-04 10:17:40 +0200662 srv_set_stopping(s, check_notify_cause(check));
Krzysztof Piotr Oledzkic8b16fc2008-02-18 01:26:35 +0100663}
Willy Tarreaubaaee002006-06-26 02:48:02 +0200664
Willy Tarreau9fe7aae2013-12-31 23:47:37 +0100665/* note: use health_adjust() only, which first checks that the observe mode is
Willy Tarreau4e9df272021-02-17 15:20:19 +0100666 * enabled. This will take the server lock if needed.
Willy Tarreau9fe7aae2013-12-31 23:47:37 +0100667 */
668void __health_adjust(struct server *s, short status)
Willy Tarreau19d14ef2012-10-29 16:51:55 +0100669{
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100670 int failed;
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100671
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100672 if (s->observe >= HANA_OBS_SIZE)
673 return;
674
Willy Tarreaubb956662013-01-24 00:37:39 +0100675 if (status >= HANA_STATUS_SIZE || !analyze_statuses[status].desc)
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100676 return;
677
678 switch (analyze_statuses[status].lr[s->observe - 1]) {
679 case 1:
680 failed = 1;
681 break;
682
683 case 2:
684 failed = 0;
685 break;
686
687 default:
688 return;
689 }
690
691 if (!failed) {
692 /* good: clear consecutive_errors */
Aurelien DARRAGON81b7c952022-12-07 14:27:42 +0100693 HA_ATOMIC_STORE(&s->consecutive_errors, 0);
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100694 return;
695 }
696
Aurelien DARRAGON81b7c952022-12-07 14:27:42 +0100697 if (HA_ATOMIC_ADD_FETCH(&s->consecutive_errors, 1) < s->consecutive_errors_limit)
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100698 return;
699
Willy Tarreau19d14ef2012-10-29 16:51:55 +0100700 chunk_printf(&trash, "Detected %d consecutive errors, last one was: %s",
Aurelien DARRAGON81b7c952022-12-07 14:27:42 +0100701 HA_ATOMIC_LOAD(&s->consecutive_errors), get_analyze_status(status));
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100702
Willy Tarreau4e9df272021-02-17 15:20:19 +0100703 HA_SPIN_LOCK(SERVER_LOCK, &s->lock);
704
Aurelien DARRAGON7d541a92022-12-07 12:17:24 +0100705 /* force fastinter for upcoming check
706 * (does nothing if fastinter is not enabled)
707 */
708 s->check.state |= CHK_ST_FASTINTER;
709
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100710 switch (s->onerror) {
711 case HANA_ONERR_FASTINTER:
712 /* force fastinter - nothing to do here as all modes force it */
713 break;
714
715 case HANA_ONERR_SUDDTH:
716 /* simulate a pre-fatal failed health check */
Simon Horman58c32972013-11-25 10:46:38 +0900717 if (s->check.health > s->check.rise)
718 s->check.health = s->check.rise + 1;
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100719
Willy Tarreau36a73432022-11-14 07:31:36 +0100720 __fallthrough;
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100721
722 case HANA_ONERR_FAILCHK:
723 /* simulate a failed health check */
Willy Tarreau843b7cb2018-07-13 10:54:26 +0200724 set_server_check_status(&s->check, HCHK_STATUS_HANA,
725 trash.area);
Willy Tarreau4eec5472014-05-20 22:32:27 +0200726 check_notify_failure(&s->check);
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100727 break;
728
729 case HANA_ONERR_MARKDWN:
730 /* mark server down */
Simon Horman58c32972013-11-25 10:46:38 +0900731 s->check.health = s->check.rise;
Willy Tarreau843b7cb2018-07-13 10:54:26 +0200732 set_server_check_status(&s->check, HCHK_STATUS_HANA,
733 trash.area);
Willy Tarreau4eec5472014-05-20 22:32:27 +0200734 check_notify_failure(&s->check);
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100735 break;
736
737 default:
738 /* write a warning? */
739 break;
740 }
741
Willy Tarreau4e9df272021-02-17 15:20:19 +0100742 HA_SPIN_UNLOCK(SERVER_LOCK, &s->lock);
743
Aurelien DARRAGON81b7c952022-12-07 14:27:42 +0100744 HA_ATOMIC_STORE(&s->consecutive_errors, 0);
Willy Tarreau4781b152021-04-06 13:53:36 +0200745 _HA_ATOMIC_INC(&s->counters.failed_hana);
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100746
Willy Tarreaua56798e2022-12-06 11:38:18 +0100747 if (s->check.fastinter) {
748 /* timer might need to be advanced, it might also already be
749 * running in another thread. Let's just wake the task up, it
750 * will automatically adjust its timer.
751 */
752 task_wakeup(s->check.task, TASK_WOKEN_MSG);
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +0100753 }
Willy Tarreauef781042010-01-27 11:53:01 +0100754}
755
Christopher Faulet61cc8522020-04-20 14:54:42 +0200756/* Checks the connection. If an error has already been reported or the socket is
Willy Tarreau20a18342013-12-05 00:31:46 +0100757 * closed, keep errno intact as it is supposed to contain the valid error code.
758 * If no error is reported, check the socket's error queue using getsockopt().
759 * Warning, this must be done only once when returning from poll, and never
760 * after an I/O error was attempted, otherwise the error queue might contain
761 * inconsistent errors. If an error is detected, the CO_FL_ERROR is set on the
762 * socket. Returns non-zero if an error was reported, zero if everything is
763 * clean (including a properly closed socket).
764 */
765static int retrieve_errno_from_socket(struct connection *conn)
766{
767 int skerr;
768 socklen_t lskerr = sizeof(skerr);
769
Willy Tarreauc8dc20a2019-12-27 12:03:27 +0100770 if (conn->flags & CO_FL_ERROR && (unclean_errno(errno) || !conn->ctrl))
Willy Tarreau20a18342013-12-05 00:31:46 +0100771 return 1;
772
Willy Tarreau3c728722014-01-23 13:50:42 +0100773 if (!conn_ctrl_ready(conn))
Willy Tarreau20a18342013-12-05 00:31:46 +0100774 return 0;
775
Willy Tarreau07ecfc52022-04-11 18:07:03 +0200776 BUG_ON(conn->flags & CO_FL_FDLESS);
777
Willy Tarreau585744b2017-08-24 14:31:19 +0200778 if (getsockopt(conn->handle.fd, SOL_SOCKET, SO_ERROR, &skerr, &lskerr) == 0)
Willy Tarreau20a18342013-12-05 00:31:46 +0100779 errno = skerr;
780
Willy Tarreauc8dc20a2019-12-27 12:03:27 +0100781 errno = unclean_errno(errno);
Willy Tarreau20a18342013-12-05 00:31:46 +0100782
783 if (!errno) {
784 /* we could not retrieve an error, that does not mean there is
785 * none. Just don't change anything and only report the prior
786 * error if any.
787 */
788 if (conn->flags & CO_FL_ERROR)
789 return 1;
790 else
791 return 0;
792 }
793
794 conn->flags |= CO_FL_ERROR | CO_FL_SOCK_WR_SH | CO_FL_SOCK_RD_SH;
795 return 1;
796}
797
Christopher Faulet61cc8522020-04-20 14:54:42 +0200798/* Tries to collect as much information as possible on the connection status,
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100799 * and adjust the server status accordingly. It may make use of <errno_bck>
800 * if non-null when the caller is absolutely certain of its validity (eg:
801 * checked just after a syscall). If the caller doesn't have a valid errno,
802 * it can pass zero, and retrieve_errno_from_socket() will be called to try
803 * to extract errno from the socket. If no error is reported, it will consider
804 * the <expired> flag. This is intended to be used when a connection error was
805 * reported in conn->flags or when a timeout was reported in <expired>. The
806 * function takes care of not updating a server status which was already set.
807 * All situations where at least one of <expired> or CO_FL_ERROR are set
808 * produce a status.
809 */
Willy Tarreau51cd5952020-06-05 12:25:38 +0200810void chk_report_conn_err(struct check *check, int errno_bck, int expired)
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100811{
Willy Tarreaubde14ad2022-05-27 10:04:04 +0200812 struct stconn *sc = check->sc;
813 struct connection *conn = sc_conn(sc);
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100814 const char *err_msg;
Willy Tarreau83061a82018-07-13 11:56:34 +0200815 struct buffer *chk;
Willy Tarreau213c6782014-10-02 14:51:02 +0200816 int step;
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100817
Christopher Faulet147b8c92021-04-10 09:00:38 +0200818 if (check->result != CHK_RES_UNKNOWN) {
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100819 return;
Christopher Faulet147b8c92021-04-10 09:00:38 +0200820 }
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100821
Willy Tarreauc8dc20a2019-12-27 12:03:27 +0100822 errno = unclean_errno(errno_bck);
823 if (conn && errno)
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100824 retrieve_errno_from_socket(conn);
825
Christopher Faulete4b40192022-06-13 07:59:46 +0200826 if (conn && !(conn->flags & CO_FL_ERROR) && !sc_ep_test(sc, SE_FL_ERROR) && !expired)
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100827 return;
828
Christopher Faulet147b8c92021-04-10 09:00:38 +0200829 TRACE_ENTER(CHK_EV_HCHK_END|CHK_EV_HCHK_ERR, check, 0, 0, (size_t[]){expired});
830
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100831 /* we'll try to build a meaningful error message depending on the
832 * context of the error possibly present in conn->err_code, and the
833 * socket error possibly collected above. This is useful to know the
834 * exact step of the L6 layer (eg: SSL handshake).
835 */
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +0200836 chk = get_trash_chunk();
837
Christopher Faulet799f3a42020-04-07 12:06:14 +0200838 if (check->type == PR_O2_TCPCHK_CHK &&
Christopher Fauletd7e63962020-04-17 20:15:59 +0200839 (check->tcpcheck_rules->flags & TCPCHK_RULES_PROTO_CHK) == TCPCHK_RULES_TCP_CHK) {
Christopher Fauletb2c2e0f2020-03-30 11:05:10 +0200840 step = tcpcheck_get_step_id(check, NULL);
Christopher Faulet147b8c92021-04-10 09:00:38 +0200841 if (!step) {
842 TRACE_DEVEL("initial connection failure", CHK_EV_HCHK_END|CHK_EV_HCHK_ERR, check);
Willy Tarreau213c6782014-10-02 14:51:02 +0200843 chunk_printf(chk, " at initial connection step of tcp-check");
Christopher Faulet147b8c92021-04-10 09:00:38 +0200844 }
Willy Tarreau213c6782014-10-02 14:51:02 +0200845 else {
846 chunk_printf(chk, " at step %d of tcp-check", step);
847 /* we were looking for a string */
Christopher Fauletb2c2e0f2020-03-30 11:05:10 +0200848 if (check->current_step && check->current_step->action == TCPCHK_ACT_CONNECT) {
849 if (check->current_step->connect.port)
850 chunk_appendf(chk, " (connect port %d)" ,check->current_step->connect.port);
Willy Tarreau213c6782014-10-02 14:51:02 +0200851 else
852 chunk_appendf(chk, " (connect)");
Christopher Faulet147b8c92021-04-10 09:00:38 +0200853 TRACE_DEVEL("connection failure", CHK_EV_HCHK_END|CHK_EV_HCHK_ERR, check);
Willy Tarreau213c6782014-10-02 14:51:02 +0200854 }
Christopher Fauletb2c2e0f2020-03-30 11:05:10 +0200855 else if (check->current_step && check->current_step->action == TCPCHK_ACT_EXPECT) {
856 struct tcpcheck_expect *expect = &check->current_step->expect;
Gaetan Rivetb616add2020-02-07 15:37:17 +0100857
858 switch (expect->type) {
859 case TCPCHK_EXPECT_STRING:
Christopher Fauletb61caf42020-04-21 10:57:42 +0200860 chunk_appendf(chk, " (expect string '%.*s')", (unsigned int)istlen(expect->data), istptr(expect->data));
Gaetan Rivetb616add2020-02-07 15:37:17 +0100861 break;
862 case TCPCHK_EXPECT_BINARY:
Christopher Fauletb61caf42020-04-21 10:57:42 +0200863 chunk_appendf(chk, " (expect binary '%.*s')", (unsigned int)istlen(expect->data), istptr(expect->data));
Gaetan Rivetb616add2020-02-07 15:37:17 +0100864 break;
Christopher Faulet67a23452020-05-05 18:10:01 +0200865 case TCPCHK_EXPECT_STRING_REGEX:
Willy Tarreau213c6782014-10-02 14:51:02 +0200866 chunk_appendf(chk, " (expect regex)");
Gaetan Rivetb616add2020-02-07 15:37:17 +0100867 break;
Christopher Faulet67a23452020-05-05 18:10:01 +0200868 case TCPCHK_EXPECT_BINARY_REGEX:
Gaetan Rivetefab6c62020-02-07 15:37:17 +0100869 chunk_appendf(chk, " (expect binary regex)");
870 break;
Christopher Fauletaaab0832020-05-05 15:54:22 +0200871 case TCPCHK_EXPECT_STRING_LF:
872 chunk_appendf(chk, " (expect log-format string)");
873 break;
874 case TCPCHK_EXPECT_BINARY_LF:
875 chunk_appendf(chk, " (expect log-format binary)");
876 break;
Christopher Faulete5870d82020-04-15 11:32:03 +0200877 case TCPCHK_EXPECT_HTTP_STATUS:
Christopher Faulet8021a5f2020-04-24 13:53:12 +0200878 chunk_appendf(chk, " (expect HTTP status codes)");
Christopher Faulete5870d82020-04-15 11:32:03 +0200879 break;
Christopher Faulet67a23452020-05-05 18:10:01 +0200880 case TCPCHK_EXPECT_HTTP_STATUS_REGEX:
Christopher Faulete5870d82020-04-15 11:32:03 +0200881 chunk_appendf(chk, " (expect HTTP status regex)");
882 break;
Christopher Faulet39708192020-05-05 10:47:36 +0200883 case TCPCHK_EXPECT_HTTP_HEADER:
884 chunk_appendf(chk, " (expect HTTP header pattern)");
885 break;
Christopher Faulete5870d82020-04-15 11:32:03 +0200886 case TCPCHK_EXPECT_HTTP_BODY:
Christopher Fauletb61caf42020-04-21 10:57:42 +0200887 chunk_appendf(chk, " (expect HTTP body content '%.*s')", (unsigned int)istlen(expect->data), istptr(expect->data));
Christopher Faulete5870d82020-04-15 11:32:03 +0200888 break;
Christopher Faulet67a23452020-05-05 18:10:01 +0200889 case TCPCHK_EXPECT_HTTP_BODY_REGEX:
Christopher Faulete5870d82020-04-15 11:32:03 +0200890 chunk_appendf(chk, " (expect HTTP body regex)");
891 break;
Christopher Fauletaaab0832020-05-05 15:54:22 +0200892 case TCPCHK_EXPECT_HTTP_BODY_LF:
893 chunk_appendf(chk, " (expect log-format HTTP body)");
894 break;
Christopher Faulet9e6ed152020-04-03 15:24:06 +0200895 case TCPCHK_EXPECT_CUSTOM:
896 chunk_appendf(chk, " (expect custom function)");
897 break;
Gaetan Rivetb616add2020-02-07 15:37:17 +0100898 case TCPCHK_EXPECT_UNDEF:
899 chunk_appendf(chk, " (undefined expect!)");
900 break;
901 }
Christopher Faulet147b8c92021-04-10 09:00:38 +0200902 TRACE_DEVEL("expect rule failed", CHK_EV_HCHK_END|CHK_EV_HCHK_ERR, check);
Willy Tarreau213c6782014-10-02 14:51:02 +0200903 }
Christopher Fauletb2c2e0f2020-03-30 11:05:10 +0200904 else if (check->current_step && check->current_step->action == TCPCHK_ACT_SEND) {
Willy Tarreau213c6782014-10-02 14:51:02 +0200905 chunk_appendf(chk, " (send)");
Christopher Faulet147b8c92021-04-10 09:00:38 +0200906 TRACE_DEVEL("send rule failed", CHK_EV_HCHK_END|CHK_EV_HCHK_ERR, check);
Willy Tarreau213c6782014-10-02 14:51:02 +0200907 }
Baptiste Assmann22b09d22015-05-01 08:03:04 +0200908
Christopher Faulet6f2a5e42020-04-01 13:11:41 +0200909 if (check->current_step && check->current_step->comment)
910 chunk_appendf(chk, " comment: '%s'", check->current_step->comment);
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +0200911 }
912 }
913
Willy Tarreau00149122017-10-04 18:05:01 +0200914 if (conn && conn->err_code) {
Willy Tarreauc8dc20a2019-12-27 12:03:27 +0100915 if (unclean_errno(errno))
Willy Tarreau843b7cb2018-07-13 10:54:26 +0200916 chunk_printf(&trash, "%s (%s)%s", conn_err_code_str(conn), strerror(errno),
917 chk->area);
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100918 else
Willy Tarreau843b7cb2018-07-13 10:54:26 +0200919 chunk_printf(&trash, "%s%s", conn_err_code_str(conn),
920 chk->area);
921 err_msg = trash.area;
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100922 }
923 else {
Willy Tarreauc8dc20a2019-12-27 12:03:27 +0100924 if (unclean_errno(errno)) {
Willy Tarreau843b7cb2018-07-13 10:54:26 +0200925 chunk_printf(&trash, "%s%s", strerror(errno),
926 chk->area);
927 err_msg = trash.area;
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100928 }
929 else {
Willy Tarreau843b7cb2018-07-13 10:54:26 +0200930 err_msg = chk->area;
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100931 }
932 }
933
Willy Tarreau00149122017-10-04 18:05:01 +0200934 if (check->state & CHK_ST_PORT_MISS) {
Baptiste Assmann95db2bc2016-06-13 14:15:41 +0200935 /* NOTE: this is reported after <fall> tries */
Baptiste Assmann95db2bc2016-06-13 14:15:41 +0200936 set_server_check_status(check, HCHK_STATUS_SOCKERR, err_msg);
937 }
938
Christopher Faulete4b40192022-06-13 07:59:46 +0200939 if (!conn || !conn->ctrl) {
Christopher Faulet5e293762020-10-26 11:10:49 +0100940 /* error before any connection attempt (connection allocation error or no control layer) */
Willy Tarreau00149122017-10-04 18:05:01 +0200941 set_server_check_status(check, HCHK_STATUS_SOCKERR, err_msg);
942 }
Willy Tarreauc192b0a2020-01-23 09:11:58 +0100943 else if (conn->flags & CO_FL_WAIT_L4_CONN) {
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100944 /* L4 not established (yet) */
Willy Tarreaubde14ad2022-05-27 10:04:04 +0200945 if (conn->flags & CO_FL_ERROR || sc_ep_test(sc, SE_FL_ERROR))
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100946 set_server_check_status(check, HCHK_STATUS_L4CON, err_msg);
947 else if (expired)
948 set_server_check_status(check, HCHK_STATUS_L4TOUT, err_msg);
Baptiste Assmanna68ca962015-04-14 01:15:08 +0200949
950 /*
951 * might be due to a server IP change.
952 * Let's trigger a DNS resolution if none are currently running.
953 */
Olivier Houchard0923fa42019-01-11 18:43:04 +0100954 if (check->server)
Emeric Brund30e9a12020-12-23 18:49:16 +0100955 resolv_trigger_resolution(check->server->resolv_requester);
Baptiste Assmanna68ca962015-04-14 01:15:08 +0200956
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100957 }
Willy Tarreauc192b0a2020-01-23 09:11:58 +0100958 else if (conn->flags & CO_FL_WAIT_L6_CONN) {
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100959 /* L6 not established (yet) */
Willy Tarreaubde14ad2022-05-27 10:04:04 +0200960 if (conn->flags & CO_FL_ERROR || sc_ep_test(sc, SE_FL_ERROR))
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100961 set_server_check_status(check, HCHK_STATUS_L6RSP, err_msg);
962 else if (expired)
963 set_server_check_status(check, HCHK_STATUS_L6TOUT, err_msg);
964 }
Willy Tarreaubde14ad2022-05-27 10:04:04 +0200965 else if (conn->flags & CO_FL_ERROR || sc_ep_test(sc, SE_FL_ERROR)) {
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100966 /* I/O error after connection was established and before we could diagnose */
967 set_server_check_status(check, HCHK_STATUS_SOCKERR, err_msg);
968 }
969 else if (expired) {
Christopher Fauletcf80f2f2020-04-01 11:04:52 +0200970 enum healthcheck_status tout = HCHK_STATUS_L7TOUT;
971
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100972 /* connection established but expired check */
Christopher Faulet1941bab2020-05-05 07:55:50 +0200973 if (check->current_step && check->current_step->action == TCPCHK_ACT_EXPECT &&
974 check->current_step->expect.tout_status != HCHK_STATUS_UNKNOWN)
Christopher Faulet811f78c2020-04-01 11:10:27 +0200975 tout = check->current_step->expect.tout_status;
976 set_server_check_status(check, tout, err_msg);
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100977 }
978
Christopher Faulet147b8c92021-04-10 09:00:38 +0200979 TRACE_LEAVE(CHK_EV_HCHK_END|CHK_EV_HCHK_ERR, check);
Willy Tarreau25e2ab52013-12-04 11:17:05 +0100980 return;
981}
982
Simon Horman98637e52014-06-20 12:30:16 +0900983
Christopher Faulet61cc8522020-04-20 14:54:42 +0200984/* Builds the server state header used by HTTP health-checks */
Willy Tarreau51cd5952020-06-05 12:25:38 +0200985int httpchk_build_status_header(struct server *s, struct buffer *buf)
Simon Horman98637e52014-06-20 12:30:16 +0900986{
Christopher Faulet61cc8522020-04-20 14:54:42 +0200987 int sv_state;
988 int ratio;
989 char addr[46];
990 char port[6];
991 const char *srv_hlt_st[7] = { "DOWN", "DOWN %d/%d",
992 "UP %d/%d", "UP",
993 "NOLB %d/%d", "NOLB",
994 "no check" };
Simon Horman98637e52014-06-20 12:30:16 +0900995
Christopher Faulet61cc8522020-04-20 14:54:42 +0200996 if (!(s->check.state & CHK_ST_ENABLED))
997 sv_state = 6;
998 else if (s->cur_state != SRV_ST_STOPPED) {
999 if (s->check.health == s->check.rise + s->check.fall - 1)
1000 sv_state = 3; /* UP */
1001 else
1002 sv_state = 2; /* going down */
Simon Horman98637e52014-06-20 12:30:16 +09001003
Christopher Faulet61cc8522020-04-20 14:54:42 +02001004 if (s->cur_state == SRV_ST_STOPPING)
1005 sv_state += 2;
1006 } else {
1007 if (s->check.health)
1008 sv_state = 1; /* going up */
1009 else
1010 sv_state = 0; /* DOWN */
Simon Horman98637e52014-06-20 12:30:16 +09001011 }
Willy Tarreaub7b24782016-06-21 15:32:29 +02001012
Christopher Faulet61cc8522020-04-20 14:54:42 +02001013 chunk_appendf(buf, srv_hlt_st[sv_state],
1014 (s->cur_state != SRV_ST_STOPPED) ? (s->check.health - s->check.rise + 1) : (s->check.health),
1015 (s->cur_state != SRV_ST_STOPPED) ? (s->check.fall) : (s->check.rise));
Willy Tarreaub7b24782016-06-21 15:32:29 +02001016
Christopher Faulet61cc8522020-04-20 14:54:42 +02001017 addr_to_str(&s->addr, addr, sizeof(addr));
1018 if (s->addr.ss_family == AF_INET || s->addr.ss_family == AF_INET6)
1019 snprintf(port, sizeof(port), "%u", s->svc_port);
1020 else
1021 *port = 0;
Willy Tarreaub7b24782016-06-21 15:32:29 +02001022
Christopher Faulet61cc8522020-04-20 14:54:42 +02001023 chunk_appendf(buf, "; address=%s; port=%s; name=%s/%s; node=%s; weight=%d/%d; scur=%d/%d; qcur=%d",
1024 addr, port, s->proxy->id, s->id,
1025 global.node,
1026 (s->cur_eweight * s->proxy->lbprm.wmult + s->proxy->lbprm.wdiv - 1) / s->proxy->lbprm.wdiv,
1027 (s->proxy->lbprm.tot_weight * s->proxy->lbprm.wmult + s->proxy->lbprm.wdiv - 1) / s->proxy->lbprm.wdiv,
Willy Tarreau7f3c1df2021-06-18 09:22:21 +02001028 s->cur_sess, s->proxy->beconn - s->proxy->queue.length,
Willy Tarreaua0570452021-06-18 09:30:30 +02001029 s->queue.length);
Willy Tarreau9f6dc722019-03-01 11:15:10 +01001030
Christopher Faulet61cc8522020-04-20 14:54:42 +02001031 if ((s->cur_state == SRV_ST_STARTING) &&
Willy Tarreau69530f52023-04-28 09:16:15 +02001032 ns_to_sec(now_ns) < s->last_change + s->slowstart &&
1033 ns_to_sec(now_ns) >= s->last_change) {
1034 ratio = MAX(1, 100 * (ns_to_sec(now_ns) - s->last_change) / s->slowstart);
Christopher Faulet61cc8522020-04-20 14:54:42 +02001035 chunk_appendf(buf, "; throttle=%d%%", ratio);
1036 }
Christopher Fauletaaae9a02020-04-26 09:50:31 +02001037
Christopher Faulet61cc8522020-04-20 14:54:42 +02001038 return b_data(buf);
1039}
Christopher Fauletaaae9a02020-04-26 09:50:31 +02001040
Willy Tarreau51cd5952020-06-05 12:25:38 +02001041/**************************************************************************/
Willy Tarreau51cd5952020-06-05 12:25:38 +02001042/***************** Health-checks based on connections *********************/
1043/**************************************************************************/
1044/* This function is used only for server health-checks. It handles connection
1045 * status updates including errors. If necessary, it wakes the check task up.
1046 * It returns 0 on normal cases, <0 if at least one close() has happened on the
1047 * connection (eg: reconnect). It relies on tcpcheck_main().
Christopher Faulet61cc8522020-04-20 14:54:42 +02001048 */
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001049int wake_srv_chk(struct stconn *sc)
Christopher Faulet61cc8522020-04-20 14:54:42 +02001050{
Christopher Faulet0256da12021-12-15 09:50:17 +01001051 struct connection *conn;
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001052 struct check *check = __sc_check(sc);
Willy Tarreau51cd5952020-06-05 12:25:38 +02001053 struct email_alertq *q = container_of(check, typeof(*q), check);
1054 int ret = 0;
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001055
Christopher Faulet147b8c92021-04-10 09:00:38 +02001056 TRACE_ENTER(CHK_EV_HCHK_WAKE, check);
Christopher Faulet08c8f8e2022-05-18 14:35:49 +02001057 if (check->result != CHK_RES_UNKNOWN)
1058 goto end;
1059
Willy Tarreau51cd5952020-06-05 12:25:38 +02001060 if (check->server)
1061 HA_SPIN_LOCK(SERVER_LOCK, &check->server->lock);
1062 else
1063 HA_SPIN_LOCK(EMAIL_ALERTS_LOCK, &q->lock);
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001064
Willy Tarreau51cd5952020-06-05 12:25:38 +02001065 /* we may have to make progress on the TCP checks */
1066 ret = tcpcheck_main(check);
Christopher Fauletaaab0832020-05-05 15:54:22 +02001067
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001068 sc = check->sc;
1069 conn = sc_conn(sc);
Christopher Fauletaaab0832020-05-05 15:54:22 +02001070
Christopher Faulete4b40192022-06-13 07:59:46 +02001071 if (unlikely(!conn || conn->flags & CO_FL_ERROR || sc_ep_test(sc, SE_FL_ERROR))) {
Willy Tarreau51cd5952020-06-05 12:25:38 +02001072 /* We may get error reports bypassing the I/O handlers, typically
1073 * the case when sending a pure TCP check which fails, then the I/O
1074 * handlers above are not called. This is completely handled by the
1075 * main processing task so let's simply wake it up. If we get here,
1076 * we expect errno to still be valid.
1077 */
Christopher Faulet147b8c92021-04-10 09:00:38 +02001078 TRACE_ERROR("report connection error", CHK_EV_HCHK_WAKE|CHK_EV_HCHK_END|CHK_EV_HCHK_ERR, check);
Willy Tarreau51cd5952020-06-05 12:25:38 +02001079 chk_report_conn_err(check, errno, 0);
1080 task_wakeup(check->task, TASK_WOKEN_IO);
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001081 }
1082
Christopher Faulet8f100422021-01-18 15:47:03 +01001083 if (check->result != CHK_RES_UNKNOWN || ret == -1) {
Christopher Faulet08c8f8e2022-05-18 14:35:49 +02001084 /* Check complete or aborted. Wake the check task up to be sure
1085 * the result is handled ASAP. */
Willy Tarreau51cd5952020-06-05 12:25:38 +02001086 ret = -1;
Willy Tarreau51cd5952020-06-05 12:25:38 +02001087 task_wakeup(check->task, TASK_WOKEN_IO);
Christopher Faulet61cc8522020-04-20 14:54:42 +02001088 }
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001089
Willy Tarreau51cd5952020-06-05 12:25:38 +02001090 if (check->server)
1091 HA_SPIN_UNLOCK(SERVER_LOCK, &check->server->lock);
1092 else
1093 HA_SPIN_UNLOCK(EMAIL_ALERTS_LOCK, &q->lock);
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001094
Christopher Faulet08c8f8e2022-05-18 14:35:49 +02001095 end:
Christopher Faulet147b8c92021-04-10 09:00:38 +02001096 TRACE_LEAVE(CHK_EV_HCHK_WAKE, check);
Christopher Faulet61cc8522020-04-20 14:54:42 +02001097 return ret;
1098}
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001099
Willy Tarreau51cd5952020-06-05 12:25:38 +02001100/* This function checks if any I/O is wanted, and if so, attempts to do so */
Christopher Faulet361417f2022-05-18 14:50:30 +02001101struct task *srv_chk_io_cb(struct task *t, void *ctx, unsigned int state)
Christopher Faulet61cc8522020-04-20 14:54:42 +02001102{
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001103 struct stconn *sc = ctx;
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001104
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001105 wake_srv_chk(sc);
Willy Tarreau51cd5952020-06-05 12:25:38 +02001106 return NULL;
Christopher Faulet61cc8522020-04-20 14:54:42 +02001107}
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001108
Willy Tarreau51cd5952020-06-05 12:25:38 +02001109/* manages a server health-check that uses a connection. Returns
1110 * the time the task accepts to wait, or TIME_ETERNITY for infinity.
Christopher Faulet61cc8522020-04-20 14:54:42 +02001111 *
1112 * Please do NOT place any return statement in this function and only leave
Willy Tarreau51cd5952020-06-05 12:25:38 +02001113 * via the out_unlock label.
Christopher Faulet61cc8522020-04-20 14:54:42 +02001114 */
Willy Tarreau144f84a2021-03-02 16:09:26 +01001115struct task *process_chk_conn(struct task *t, void *context, unsigned int state)
Christopher Faulet61cc8522020-04-20 14:54:42 +02001116{
Willy Tarreau51cd5952020-06-05 12:25:38 +02001117 struct check *check = context;
1118 struct proxy *proxy = check->proxy;
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001119 struct stconn *sc;
Christopher Faulet92017a32021-05-06 16:01:18 +02001120 struct connection *conn;
Willy Tarreau51cd5952020-06-05 12:25:38 +02001121 int rv;
1122 int expired = tick_is_expired(t->expire, now_ms);
Willy Tarreaudeccd112018-06-14 18:38:55 +02001123
Christopher Faulet147b8c92021-04-10 09:00:38 +02001124 TRACE_ENTER(CHK_EV_TASK_WAKE, check);
1125
Willy Tarreaud114f4a2022-10-12 20:58:18 +02001126 if (check->state & CHK_ST_SLEEPING) {
1127 /* This check just restarted. It's still time to verify if
1128 * we're on an overloaded thread or if a more suitable one is
1129 * available. This helps spread the load over the available
1130 * threads, without migrating too often. For this we'll check
1131 * our load, and pick a random thread, check if it has less
1132 * than half of the current thread's load, and if so we'll
1133 * bounce the task there. It's possible because it's not yet
1134 * tied to the current thread. The other thread will not bounce
1135 * the task again because we're removing CHK_ST_SLEEPING.
1136 */
1137 uint my_load = HA_ATOMIC_LOAD(&th_ctx->rq_total);
1138
1139 check->state &= ~CHK_ST_SLEEPING;
1140
1141 if (my_load >= 2) {
1142 uint new_tid = statistical_prng_range(global.nbthread);
1143 uint new_load = HA_ATOMIC_LOAD(&ha_thread_ctx[new_tid].rq_total);
1144
1145 if (new_load <= my_load / 2) {
1146 /* Found one. Let's migrate the task over there. We have to
1147 * remove it from the WQ first and kill its expire time
1148 * otherwise the scheduler will reinsert it and trigger a
1149 * BUG_ON() as we're not allowed to call task_queue() for a
1150 * foreign thread. The recipient will restore the expiration.
1151 */
1152 task_unlink_wq(t);
1153 t->expire = TICK_ETERNITY;
1154 task_set_thread(t, new_tid);
1155 task_wakeup(t, TASK_WOKEN_MSG);
1156 TRACE_LEAVE(CHK_EV_TASK_WAKE, check);
1157 return t;
1158 }
1159 }
1160 }
1161
Willy Tarreau51cd5952020-06-05 12:25:38 +02001162 if (check->server)
1163 HA_SPIN_LOCK(SERVER_LOCK, &check->server->lock);
Christopher Faulet92017a32021-05-06 16:01:18 +02001164
Willy Tarreaud114f4a2022-10-12 20:58:18 +02001165 if (!(check->state & (CHK_ST_INPROGRESS|CHK_ST_IN_ALLOC|CHK_ST_OUT_ALLOC))) {
1166 /* This task might have bounced from another overloaded thread, it
1167 * needs an expiration timer that was supposed to be now, but that
1168 * was erased during the bounce.
1169 */
Willy Tarreauacbd22f2023-09-01 07:41:46 +02001170 if (!tick_isset(t->expire)) {
Willy Tarreaud114f4a2022-10-12 20:58:18 +02001171 t->expire = now_ms;
Willy Tarreauacbd22f2023-09-01 07:41:46 +02001172 expired = 0;
1173 }
Willy Tarreaud114f4a2022-10-12 20:58:18 +02001174 }
1175
Amaury Denoyelleb33a0ab2021-07-29 15:51:45 +02001176 if (unlikely(check->state & CHK_ST_PURGE)) {
1177 TRACE_STATE("health-check state to purge", CHK_EV_TASK_WAKE, check);
1178 }
1179 else if (!(check->state & (CHK_ST_INPROGRESS))) {
Willy Tarreau9e080bf2022-12-06 18:20:56 +01001180 /* no check currently running, but we might have been woken up
1181 * before the timer's expiration to update it according to a
1182 * new state (e.g. fastinter), in which case we'll reprogram
1183 * the new timer.
1184 */
Willy Tarreauacbd22f2023-09-01 07:41:46 +02001185 if (!tick_is_expired(t->expire, now_ms)) { /* woke up too early */
Willy Tarreau9e080bf2022-12-06 18:20:56 +01001186 if (check->server) {
1187 int new_exp = tick_add(now_ms, MS_TO_TICKS(srv_getinter(check)));
1188
1189 if (tick_is_expired(new_exp, t->expire)) {
1190 TRACE_STATE("health-check was advanced", CHK_EV_TASK_WAKE, check);
1191 goto update_timer;
1192 }
1193 }
1194
Christopher Faulet147b8c92021-04-10 09:00:38 +02001195 TRACE_STATE("health-check wake up too early", CHK_EV_TASK_WAKE, check);
Willy Tarreau51cd5952020-06-05 12:25:38 +02001196 goto out_unlock;
Christopher Faulet147b8c92021-04-10 09:00:38 +02001197 }
Willy Tarreauabca5b62013-12-06 14:19:25 +01001198
Willy Tarreau51cd5952020-06-05 12:25:38 +02001199 /* we don't send any health-checks when the proxy is
1200 * stopped, the server should not be checked or the check
1201 * is disabled.
1202 */
1203 if (((check->state & (CHK_ST_ENABLED | CHK_ST_PAUSED)) != CHK_ST_ENABLED) ||
Christopher Fauletdfd10ab2021-10-06 14:24:19 +02001204 (proxy->flags & (PR_FL_DISABLED|PR_FL_STOPPED))) {
Christopher Faulet147b8c92021-04-10 09:00:38 +02001205 TRACE_STATE("health-check paused or disabled", CHK_EV_TASK_WAKE, check);
Willy Tarreau51cd5952020-06-05 12:25:38 +02001206 goto reschedule;
Christopher Faulet147b8c92021-04-10 09:00:38 +02001207 }
Christopher Faulet404f9192020-04-09 23:13:54 +02001208
Willy Tarreau51cd5952020-06-05 12:25:38 +02001209 /* we'll initiate a new check */
1210 set_server_check_status(check, HCHK_STATUS_START, NULL);
Christopher Faulet404f9192020-04-09 23:13:54 +02001211
Willy Tarreau51cd5952020-06-05 12:25:38 +02001212 check->state |= CHK_ST_INPROGRESS;
Christopher Faulet147b8c92021-04-10 09:00:38 +02001213 TRACE_STATE("init new health-check", CHK_EV_TASK_WAKE|CHK_EV_HCHK_START, check);
Christopher Faulet61cc8522020-04-20 14:54:42 +02001214
Willy Tarreaueed39112022-06-15 17:20:16 +02001215 task_set_thread(t, tid);
Willy Tarreau51cd5952020-06-05 12:25:38 +02001216
1217 check->current_step = NULL;
Christopher Faulet9ec2f4d2022-03-23 15:15:29 +01001218
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001219 check->sc = sc_new_from_check(check, SC_FL_NONE);
1220 if (!check->sc) {
Christopher Faulet177a0e62022-04-12 17:47:07 +02001221 set_server_check_status(check, HCHK_STATUS_SOCKERR, NULL);
1222 goto end;
Christopher Faulet9ec2f4d2022-03-23 15:15:29 +01001223 }
Willy Tarreau51cd5952020-06-05 12:25:38 +02001224 tcpcheck_main(check);
Christopher Faulet92017a32021-05-06 16:01:18 +02001225 expired = 0;
Christopher Faulet61cc8522020-04-20 14:54:42 +02001226 }
Christopher Faulet92017a32021-05-06 16:01:18 +02001227
Christopher Faulet92017a32021-05-06 16:01:18 +02001228 /* there was a test running.
1229 * First, let's check whether there was an uncaught error,
1230 * which can happen on connect timeout or error.
1231 */
Amaury Denoyelleb33a0ab2021-07-29 15:51:45 +02001232 if (check->result == CHK_RES_UNKNOWN && likely(!(check->state & CHK_ST_PURGE))) {
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001233 sc = check->sc;
Christopher Faulete4b40192022-06-13 07:59:46 +02001234 conn = sc_conn(sc);
Christopher Faulet177a0e62022-04-12 17:47:07 +02001235
Christopher Faulet92017a32021-05-06 16:01:18 +02001236 /* Here the connection must be defined. Otherwise the
1237 * error would have already been detected
Willy Tarreau51cd5952020-06-05 12:25:38 +02001238 */
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001239 if ((conn && ((conn->flags & CO_FL_ERROR) || sc_ep_test(sc, SE_FL_ERROR))) || expired) {
Christopher Faulet92017a32021-05-06 16:01:18 +02001240 TRACE_ERROR("report connection error", CHK_EV_TASK_WAKE|CHK_EV_HCHK_END|CHK_EV_HCHK_ERR, check);
1241 chk_report_conn_err(check, 0, expired);
1242 }
1243 else {
1244 if (check->state & CHK_ST_CLOSE_CONN) {
1245 TRACE_DEVEL("closing current connection", CHK_EV_TASK_WAKE|CHK_EV_HCHK_RUN, check);
Christopher Faulet92017a32021-05-06 16:01:18 +02001246 check->state &= ~CHK_ST_CLOSE_CONN;
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001247 if (!sc_reset_endp(check->sc)) {
Christopher Fauleta6c4a482022-04-28 18:25:24 +02001248 /* error will be handled by tcpcheck_main().
Willy Tarreaub605c422022-05-17 17:04:55 +02001249 * On success, remove all flags except SE_FL_DETACHED
Christopher Fauleta6c4a482022-04-28 18:25:24 +02001250 */
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001251 sc_ep_clr(check->sc, ~SE_FL_DETACHED);
Christopher Fauleta6c4a482022-04-28 18:25:24 +02001252 }
Christopher Faulet92017a32021-05-06 16:01:18 +02001253 tcpcheck_main(check);
Willy Tarreau51cd5952020-06-05 12:25:38 +02001254 }
Christopher Faulet92017a32021-05-06 16:01:18 +02001255 if (check->result == CHK_RES_UNKNOWN) {
1256 TRACE_DEVEL("health-check not expired", CHK_EV_TASK_WAKE|CHK_EV_HCHK_RUN, check);
1257 goto out_unlock; /* timeout not reached, wait again */
Christopher Faulet8f100422021-01-18 15:47:03 +01001258 }
Christopher Faulet61cc8522020-04-20 14:54:42 +02001259 }
Christopher Faulet92017a32021-05-06 16:01:18 +02001260 }
Christopher Faulet404f9192020-04-09 23:13:54 +02001261
Christopher Faulet92017a32021-05-06 16:01:18 +02001262 /* check complete or aborted */
1263 TRACE_STATE("health-check complete or aborted", CHK_EV_TASK_WAKE|CHK_EV_HCHK_END, check);
Christopher Fauletba3c68f2020-04-01 16:27:05 +02001264
Christopher Faulete4b40192022-06-13 07:59:46 +02001265 /* check->sc may be NULL when the healthcheck is purged */
Christopher Faulet92017a32021-05-06 16:01:18 +02001266 check->current_step = NULL;
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001267 sc = check->sc;
1268 conn = (sc ? sc_conn(sc) : NULL);
Christopher Fauletba3c68f2020-04-01 16:27:05 +02001269
Christopher Faulet92017a32021-05-06 16:01:18 +02001270 if (conn && conn->xprt) {
1271 /* The check was aborted and the connection was not yet closed.
1272 * This can happen upon timeout, or when an external event such
1273 * as a failed response coupled with "observe layer7" caused the
1274 * server state to be suddenly changed.
1275 */
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001276 sc_conn_drain_and_shut(sc);
Christopher Faulet92017a32021-05-06 16:01:18 +02001277 }
Christopher Fauletba3c68f2020-04-01 16:27:05 +02001278
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001279 if (sc) {
1280 sc_destroy(sc);
Christopher Fauletcb760302023-05-17 09:25:32 +02001281 check->sc = NULL;
Christopher Faulet177a0e62022-04-12 17:47:07 +02001282 }
Willy Tarreau51cd5952020-06-05 12:25:38 +02001283
Christopher Faulet92017a32021-05-06 16:01:18 +02001284 if (check->sess != NULL) {
1285 vars_prune(&check->vars, check->sess, NULL);
1286 session_free(check->sess);
1287 check->sess = NULL;
1288 }
Willy Tarreau51cd5952020-06-05 12:25:38 +02001289
Christopher Fauletb041b232022-03-24 10:27:02 +01001290 end:
Amaury Denoyelleb33a0ab2021-07-29 15:51:45 +02001291 if (check->server && likely(!(check->state & CHK_ST_PURGE))) {
Christopher Faulet92017a32021-05-06 16:01:18 +02001292 if (check->result == CHK_RES_FAILED) {
1293 /* a failure or timeout detected */
1294 TRACE_DEVEL("report failure", CHK_EV_TASK_WAKE|CHK_EV_HCHK_END|CHK_EV_HCHK_ERR, check);
1295 check_notify_failure(check);
1296 }
1297 else if (check->result == CHK_RES_CONDPASS) {
1298 /* check is OK but asks for stopping mode */
1299 TRACE_DEVEL("report conditional success", CHK_EV_TASK_WAKE|CHK_EV_HCHK_END|CHK_EV_HCHK_SUCC, check);
1300 check_notify_stopping(check);
Christopher Faulet61cc8522020-04-20 14:54:42 +02001301 }
Christopher Faulet92017a32021-05-06 16:01:18 +02001302 else if (check->result == CHK_RES_PASSED) {
1303 /* a success was detected */
1304 TRACE_DEVEL("report success", CHK_EV_TASK_WAKE|CHK_EV_HCHK_END|CHK_EV_HCHK_SUCC, check);
1305 check_notify_success(check);
1306 }
1307 }
Christopher Faulet6d781f62022-05-18 14:24:43 +02001308
1309 if (LIST_INLIST(&check->buf_wait.list))
1310 LIST_DEL_INIT(&check->buf_wait.list);
1311
Christopher Faulet92017a32021-05-06 16:01:18 +02001312 check_release_buf(check, &check->bi);
1313 check_release_buf(check, &check->bo);
1314 check->state &= ~(CHK_ST_INPROGRESS|CHK_ST_IN_ALLOC|CHK_ST_OUT_ALLOC);
Willy Tarreaud114f4a2022-10-12 20:58:18 +02001315 check->state |= CHK_ST_SLEEPING;
Christopher Fauletba3c68f2020-04-01 16:27:05 +02001316
Willy Tarreau9e080bf2022-12-06 18:20:56 +01001317 update_timer:
Christopher Faulet92017a32021-05-06 16:01:18 +02001318 if (check->server) {
1319 rv = 0;
1320 if (global.spread_checks > 0) {
1321 rv = srv_getinter(check) * global.spread_checks / 100;
Willy Tarreaua840b4a2022-10-12 21:48:17 +02001322 rv -= (int) (2 * rv * (statistical_prng() / 4294967295.0));
Christopher Faulet61cc8522020-04-20 14:54:42 +02001323 }
Christopher Faulet92017a32021-05-06 16:01:18 +02001324 t->expire = tick_add(now_ms, MS_TO_TICKS(srv_getinter(check) + rv));
Aurelien DARRAGON7d541a92022-12-07 12:17:24 +01001325 /* reset fastinter flag (if set) so that srv_getinter()
1326 * only returns fastinter if server health is degraded
1327 */
1328 check->state &= ~CHK_ST_FASTINTER;
Christopher Faulet61cc8522020-04-20 14:54:42 +02001329 }
Willy Tarreau51cd5952020-06-05 12:25:38 +02001330
1331 reschedule:
Christopher Faulet8bca3cc2023-05-16 18:07:51 +02001332 if (proxy->flags & (PR_FL_DISABLED|PR_FL_STOPPED))
1333 t->expire = TICK_ETERNITY;
1334 else {
1335 while (tick_is_expired(t->expire, now_ms))
1336 t->expire = tick_add(t->expire, MS_TO_TICKS(check->inter));
1337 }
1338
Willy Tarreau51cd5952020-06-05 12:25:38 +02001339 out_unlock:
1340 if (check->server)
1341 HA_SPIN_UNLOCK(SERVER_LOCK, &check->server->lock);
Christopher Faulet147b8c92021-04-10 09:00:38 +02001342
1343 TRACE_LEAVE(CHK_EV_TASK_WAKE, check);
Amaury Denoyelleb33a0ab2021-07-29 15:51:45 +02001344
1345 /* Free the check if set to PURGE. After this, the check instance may be
Amaury Denoyellebc2ebfa2021-08-25 15:34:53 +02001346 * freed via the srv_drop invocation, so it must not be accessed after
1347 * this point.
Amaury Denoyelleb33a0ab2021-07-29 15:51:45 +02001348 */
1349 if (unlikely(check->state & CHK_ST_PURGE)) {
Amaury Denoyelle26cb8342021-08-10 16:23:49 +02001350 free_check(check);
Amaury Denoyelle9ba34ae2021-08-09 15:09:17 +02001351 if (check->server)
Amaury Denoyellebc2ebfa2021-08-25 15:34:53 +02001352 srv_drop(check->server);
Amaury Denoyelle26cb8342021-08-10 16:23:49 +02001353
1354 t = NULL;
Amaury Denoyelleb33a0ab2021-07-29 15:51:45 +02001355 }
1356
Willy Tarreau51cd5952020-06-05 12:25:38 +02001357 return t;
Christopher Fauletba3c68f2020-04-01 16:27:05 +02001358}
1359
Willy Tarreau51cd5952020-06-05 12:25:38 +02001360
Christopher Faulet61cc8522020-04-20 14:54:42 +02001361/**************************************************************************/
1362/************************** Init/deinit checks ****************************/
1363/**************************************************************************/
Christopher Fauletb381a502020-11-25 13:47:00 +01001364/*
1365 * Tries to grab a buffer and to re-enables processing on check <target>. The
1366 * check flags are used to figure what buffer was requested. It returns 1 if the
1367 * allocation succeeds, in which case the I/O tasklet is woken up, or 0 if it's
1368 * impossible to wake up and we prefer to be woken up later.
1369 */
1370int check_buf_available(void *target)
Christopher Faulet61cc8522020-04-20 14:54:42 +02001371{
Christopher Fauletb381a502020-11-25 13:47:00 +01001372 struct check *check = target;
1373
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001374 BUG_ON(!check->sc);
Christopher Fauletc95eaef2022-05-18 15:57:15 +02001375
Willy Tarreaud68d4f12021-03-22 14:44:31 +01001376 if ((check->state & CHK_ST_IN_ALLOC) && b_alloc(&check->bi)) {
Christopher Faulet147b8c92021-04-10 09:00:38 +02001377 TRACE_STATE("unblocking check, input buffer allocated", CHK_EV_TCPCHK_EXP|CHK_EV_RX_BLK, check);
Christopher Fauletb381a502020-11-25 13:47:00 +01001378 check->state &= ~CHK_ST_IN_ALLOC;
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001379 tasklet_wakeup(check->sc->wait_event.tasklet);
Christopher Fauletb381a502020-11-25 13:47:00 +01001380 return 1;
1381 }
Willy Tarreaud68d4f12021-03-22 14:44:31 +01001382 if ((check->state & CHK_ST_OUT_ALLOC) && b_alloc(&check->bo)) {
Christopher Faulet147b8c92021-04-10 09:00:38 +02001383 TRACE_STATE("unblocking check, output buffer allocated", CHK_EV_TCPCHK_SND|CHK_EV_TX_BLK, check);
Christopher Fauletb381a502020-11-25 13:47:00 +01001384 check->state &= ~CHK_ST_OUT_ALLOC;
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001385 tasklet_wakeup(check->sc->wait_event.tasklet);
Christopher Fauletb381a502020-11-25 13:47:00 +01001386 return 1;
1387 }
1388
1389 return 0;
1390}
Christopher Fauletba3c68f2020-04-01 16:27:05 +02001391
Christopher Fauletb381a502020-11-25 13:47:00 +01001392/*
William Dauchyf4300902021-02-06 20:47:50 +01001393 * Allocate a buffer. If it fails, it adds the check in buffer wait queue.
Christopher Fauletb381a502020-11-25 13:47:00 +01001394 */
1395struct buffer *check_get_buf(struct check *check, struct buffer *bptr)
1396{
1397 struct buffer *buf = NULL;
Christopher Fauletba3c68f2020-04-01 16:27:05 +02001398
Willy Tarreau2b718102021-04-21 07:32:39 +02001399 if (likely(!LIST_INLIST(&check->buf_wait.list)) &&
Willy Tarreaud68d4f12021-03-22 14:44:31 +01001400 unlikely((buf = b_alloc(bptr)) == NULL)) {
Christopher Fauletb381a502020-11-25 13:47:00 +01001401 check->buf_wait.target = check;
1402 check->buf_wait.wakeup_cb = check_buf_available;
Willy Tarreaub4e34762021-09-30 19:02:18 +02001403 LIST_APPEND(&th_ctx->buffer_wq, &check->buf_wait.list);
Christopher Fauletb381a502020-11-25 13:47:00 +01001404 }
1405 return buf;
1406}
1407
1408/*
1409 * Release a buffer, if any, and try to wake up entities waiting in the buffer
1410 * wait queue.
1411 */
1412void check_release_buf(struct check *check, struct buffer *bptr)
1413{
1414 if (bptr->size) {
1415 b_free(bptr);
Willy Tarreau4d77bbf2021-02-20 12:02:46 +01001416 offer_buffers(check->buf_wait.target, 1);
Christopher Fauletb381a502020-11-25 13:47:00 +01001417 }
1418}
1419
1420const char *init_check(struct check *check, int type)
1421{
1422 check->type = type;
Christopher Fauletba3c68f2020-04-01 16:27:05 +02001423
Christopher Fauletb381a502020-11-25 13:47:00 +01001424 check->bi = BUF_NULL;
1425 check->bo = BUF_NULL;
Willy Tarreau90f366b2021-02-20 11:49:49 +01001426 LIST_INIT(&check->buf_wait.list);
Christopher Faulet61cc8522020-04-20 14:54:42 +02001427 return NULL;
1428}
1429
Amaury Denoyelleb33a0ab2021-07-29 15:51:45 +02001430/* Liberates the resources allocated for a check.
1431 *
Amaury Denoyelle6d7fc442021-08-10 16:22:51 +02001432 * This function must only be run by the thread owning the check.
Amaury Denoyelleb33a0ab2021-07-29 15:51:45 +02001433 */
Christopher Faulet61cc8522020-04-20 14:54:42 +02001434void free_check(struct check *check)
Gaetan Rivet707b52f2020-02-21 18:14:59 +01001435{
Amaury Denoyelle6d7fc442021-08-10 16:22:51 +02001436 /* For agent-check, free the rules / vars from the server. This is not
1437 * done for health-check : the proxy is the owner of the rules / vars
1438 * in this case.
1439 */
1440 if (check->state & CHK_ST_AGENT) {
1441 free_tcpcheck_vars(&check->tcpcheck_rules->preset_vars);
1442 ha_free(&check->tcpcheck_rules);
1443 }
1444
Christopher Faulet61cc8522020-04-20 14:54:42 +02001445 task_destroy(check->task);
Christopher Faulet61cc8522020-04-20 14:54:42 +02001446
Christopher Fauletb381a502020-11-25 13:47:00 +01001447 check_release_buf(check, &check->bi);
1448 check_release_buf(check, &check->bo);
Willy Tarreaubde14ad2022-05-27 10:04:04 +02001449 if (check->sc) {
1450 sc_destroy(check->sc);
1451 check->sc = NULL;
Christopher Faulet61cc8522020-04-20 14:54:42 +02001452 }
Gaetan Rivet707b52f2020-02-21 18:14:59 +01001453}
1454
Amaury Denoyelleb33a0ab2021-07-29 15:51:45 +02001455/* This function must be used in order to free a started check. The check will
1456 * be scheduled for a next execution in order to properly close and free all
1457 * check elements.
1458 *
1459 * Non thread-safe.
1460 */
1461void check_purge(struct check *check)
1462{
Amaury Denoyelle25fe1032021-08-10 16:21:55 +02001463 check->state |= CHK_ST_PURGE;
Amaury Denoyelleb33a0ab2021-07-29 15:51:45 +02001464 task_wakeup(check->task, TASK_WOKEN_OTHER);
1465}
1466
Christopher Faulet61cc8522020-04-20 14:54:42 +02001467/* manages a server health-check. Returns the time the task accepts to wait, or
1468 * TIME_ETERNITY for infinity.
1469 */
Willy Tarreau144f84a2021-03-02 16:09:26 +01001470struct task *process_chk(struct task *t, void *context, unsigned int state)
Gaetan Rivet707b52f2020-02-21 18:14:59 +01001471{
Christopher Faulet61cc8522020-04-20 14:54:42 +02001472 struct check *check = context;
1473
1474 if (check->type == PR_O2_EXT_CHK)
1475 return process_chk_proc(t, context, state);
1476 return process_chk_conn(t, context, state);
1477
Gaetan Rivet707b52f2020-02-21 18:14:59 +01001478}
1479
Christopher Faulet61cc8522020-04-20 14:54:42 +02001480
Amaury Denoyelle3c2ab1a2021-07-22 16:04:40 +02001481int start_check_task(struct check *check, int mininter,
Christopher Faulet61cc8522020-04-20 14:54:42 +02001482 int nbcheck, int srvpos)
Gaetan Rivet707b52f2020-02-21 18:14:59 +01001483{
Christopher Faulet61cc8522020-04-20 14:54:42 +02001484 struct task *t;
Gaetan Rivet707b52f2020-02-21 18:14:59 +01001485
Willy Tarreaubeeabf52021-10-01 18:23:30 +02001486 /* task for the check. Process-based checks exclusively run on thread 1. */
Christopher Faulet61cc8522020-04-20 14:54:42 +02001487 if (check->type == PR_O2_EXT_CHK)
Willy Tarreaua89c1912021-10-20 18:43:30 +02001488 t = task_new_on(0);
Christopher Faulet177a0e62022-04-12 17:47:07 +02001489 else
Willy Tarreaubeeabf52021-10-01 18:23:30 +02001490 t = task_new_anywhere();
Gaetan Rivet707b52f2020-02-21 18:14:59 +01001491
Christopher Faulet54e85cb2022-01-06 08:46:56 +01001492 if (!t)
1493 goto fail_alloc_task;
1494
Christopher Faulet61cc8522020-04-20 14:54:42 +02001495 check->task = t;
1496 t->process = process_chk;
1497 t->context = check;
Gaetan Rivet707b52f2020-02-21 18:14:59 +01001498
Christopher Faulet61cc8522020-04-20 14:54:42 +02001499 if (mininter < srv_getinter(check))
1500 mininter = srv_getinter(check);
1501
Willy Tarreaub93758c2023-05-17 07:39:57 +02001502 if (global.spread_checks > 0) {
1503 int rnd;
1504
1505 rnd = srv_getinter(check) * global.spread_checks / 100;
1506 rnd -= (int) (2 * rnd * (ha_random32() / 4294967295.0));
1507 mininter += rnd;
1508 }
1509
Christopher Faulet61cc8522020-04-20 14:54:42 +02001510 if (global.max_spread_checks && mininter > global.max_spread_checks)
1511 mininter = global.max_spread_checks;
1512
1513 /* check this every ms */
Willy Tarreauc7b93082023-05-16 19:19:36 +02001514 t->expire = tick_add(now_ms, MS_TO_TICKS(mininter * srvpos / nbcheck));
Willy Tarreau69530f52023-04-28 09:16:15 +02001515 check->start = now_ns;
Christopher Faulet61cc8522020-04-20 14:54:42 +02001516 task_queue(t);
1517
1518 return 1;
Christopher Faulet54e85cb2022-01-06 08:46:56 +01001519
1520 fail_alloc_task:
Christopher Faulet54e85cb2022-01-06 08:46:56 +01001521 ha_alert("Starting [%s:%s] check: out of memory.\n",
1522 check->server->proxy->id, check->server->id);
1523 return 0;
Gaetan Rivet707b52f2020-02-21 18:14:59 +01001524}
1525
Christopher Faulet61cc8522020-04-20 14:54:42 +02001526/*
1527 * Start health-check.
1528 * Returns 0 if OK, ERR_FATAL on error, and prints the error in this case.
1529 */
1530static int start_checks()
1531{
1532
1533 struct proxy *px;
1534 struct server *s;
Christopher Faulet61cc8522020-04-20 14:54:42 +02001535 int nbcheck=0, mininter=0, srvpos=0;
1536
1537 /* 0- init the dummy frontend used to create all checks sessions */
1538 init_new_proxy(&checks_fe);
Christopher Faulet0f1fc232021-04-16 10:49:07 +02001539 checks_fe.id = strdup("CHECKS-FE");
Christopher Faulet61cc8522020-04-20 14:54:42 +02001540 checks_fe.cap = PR_CAP_FE | PR_CAP_BE;
1541 checks_fe.mode = PR_MODE_TCP;
1542 checks_fe.maxconn = 0;
1543 checks_fe.conn_retries = CONN_RETRIES;
1544 checks_fe.options2 |= PR_O2_INDEPSTR | PR_O2_SMARTCON | PR_O2_SMARTACC;
1545 checks_fe.timeout.client = TICK_ETERNITY;
1546
1547 /* 1- count the checkers to run simultaneously.
1548 * We also determine the minimum interval among all of those which
1549 * have an interval larger than SRV_CHK_INTER_THRES. This interval
1550 * will be used to spread their start-up date. Those which have
1551 * a shorter interval will start independently and will not dictate
1552 * too short an interval for all others.
1553 */
1554 for (px = proxies_list; px; px = px->next) {
1555 for (s = px->srv; s; s = s->next) {
Christopher Faulet61cc8522020-04-20 14:54:42 +02001556 if (s->check.state & CHK_ST_CONFIGURED) {
1557 nbcheck++;
1558 if ((srv_getinter(&s->check) >= SRV_CHK_INTER_THRES) &&
1559 (!mininter || mininter > srv_getinter(&s->check)))
1560 mininter = srv_getinter(&s->check);
Christopher Faulet5c288742020-03-31 08:15:58 +02001561 }
1562
Christopher Faulet61cc8522020-04-20 14:54:42 +02001563 if (s->agent.state & CHK_ST_CONFIGURED) {
1564 nbcheck++;
1565 if ((srv_getinter(&s->agent) >= SRV_CHK_INTER_THRES) &&
1566 (!mininter || mininter > srv_getinter(&s->agent)))
1567 mininter = srv_getinter(&s->agent);
1568 }
Christopher Faulet5c288742020-03-31 08:15:58 +02001569 }
Christopher Faulet61cc8522020-04-20 14:54:42 +02001570 }
Christopher Fauletb7d30092020-03-30 15:19:03 +02001571
Christopher Faulet61cc8522020-04-20 14:54:42 +02001572 if (!nbcheck)
Christopher Fauletfc633b62020-11-06 15:24:23 +01001573 return ERR_NONE;
Christopher Fauletb7d30092020-03-30 15:19:03 +02001574
Christopher Faulet61cc8522020-04-20 14:54:42 +02001575 srand((unsigned)time(NULL));
Christopher Fauletb7d30092020-03-30 15:19:03 +02001576
William Dauchyf4300902021-02-06 20:47:50 +01001577 /* 2- start them as far as possible from each other. For this, we will
1578 * start them after their interval is set to the min interval divided
1579 * by the number of servers, weighted by the server's position in the
1580 * list.
Christopher Faulet61cc8522020-04-20 14:54:42 +02001581 */
1582 for (px = proxies_list; px; px = px->next) {
1583 if ((px->options2 & PR_O2_CHK_ANY) == PR_O2_EXT_CHK) {
1584 if (init_pid_list()) {
1585 ha_alert("Starting [%s] check: out of memory.\n", px->id);
1586 return ERR_ALERT | ERR_FATAL;
1587 }
1588 }
Christopher Fauletb7d30092020-03-30 15:19:03 +02001589
Christopher Faulet61cc8522020-04-20 14:54:42 +02001590 for (s = px->srv; s; s = s->next) {
1591 /* A task for the main check */
1592 if (s->check.state & CHK_ST_CONFIGURED) {
1593 if (s->check.type == PR_O2_EXT_CHK) {
1594 if (!prepare_external_check(&s->check))
1595 return ERR_ALERT | ERR_FATAL;
Christopher Fauletb7d30092020-03-30 15:19:03 +02001596 }
Christopher Faulet61cc8522020-04-20 14:54:42 +02001597 if (!start_check_task(&s->check, mininter, nbcheck, srvpos))
1598 return ERR_ALERT | ERR_FATAL;
1599 srvpos++;
Christopher Faulet98572322020-03-30 13:16:44 +02001600 }
Christopher Fauletfd6c2292020-03-25 18:20:15 +01001601
Christopher Faulet61cc8522020-04-20 14:54:42 +02001602 /* A task for a auxiliary agent check */
1603 if (s->agent.state & CHK_ST_CONFIGURED) {
1604 if (!start_check_task(&s->agent, mininter, nbcheck, srvpos)) {
1605 return ERR_ALERT | ERR_FATAL;
1606 }
1607 srvpos++;
1608 }
Christopher Fauletfd6c2292020-03-25 18:20:15 +01001609 }
Christopher Fauletfd6c2292020-03-25 18:20:15 +01001610 }
Christopher Fauletfc633b62020-11-06 15:24:23 +01001611 return ERR_NONE;
Christopher Faulet61cc8522020-04-20 14:54:42 +02001612}
Christopher Fauletfd6c2292020-03-25 18:20:15 +01001613
Christopher Fauletfd6c2292020-03-25 18:20:15 +01001614
Christopher Faulet61cc8522020-04-20 14:54:42 +02001615/*
1616 * Return value:
1617 * the port to be used for the health check
1618 * 0 in case no port could be found for the check
1619 */
1620static int srv_check_healthcheck_port(struct check *chk)
1621{
1622 int i = 0;
1623 struct server *srv = NULL;
1624
1625 srv = chk->server;
1626
William Dauchyf4300902021-02-06 20:47:50 +01001627 /* by default, we use the health check port configured */
Christopher Faulet61cc8522020-04-20 14:54:42 +02001628 if (chk->port > 0)
1629 return chk->port;
1630
1631 /* try to get the port from check_core.addr if check.port not set */
1632 i = get_host_port(&chk->addr);
1633 if (i > 0)
1634 return i;
1635
1636 /* try to get the port from server address */
1637 /* prevent MAPPORTS from working at this point, since checks could
1638 * not be performed in such case (MAPPORTS impose a relative ports
1639 * based on live traffic)
1640 */
1641 if (srv->flags & SRV_F_MAPPORTS)
1642 return 0;
1643
1644 i = srv->svc_port; /* by default */
1645 if (i > 0)
1646 return i;
1647
1648 return 0;
Christopher Fauletfd6c2292020-03-25 18:20:15 +01001649}
1650
Christopher Faulet61cc8522020-04-20 14:54:42 +02001651/* Initializes an health-check attached to the server <srv>. Non-zero is returned
1652 * if an error occurred.
1653 */
Amaury Denoyelle3c2ab1a2021-07-22 16:04:40 +02001654int init_srv_check(struct server *srv)
Christopher Fauletfd6c2292020-03-25 18:20:15 +01001655{
Christopher Faulet61cc8522020-04-20 14:54:42 +02001656 const char *err;
1657 struct tcpcheck_rule *r;
Christopher Fauletfc633b62020-11-06 15:24:23 +01001658 int ret = ERR_NONE;
Amaury Denoyelle0519bd42020-11-13 12:34:56 +01001659 int check_type;
Christopher Fauletfd6c2292020-03-25 18:20:15 +01001660
Christopher Faulet6ecd5932021-01-12 17:29:45 +01001661 if (!srv->do_check || !(srv->proxy->cap & PR_CAP_BE))
Christopher Faulet61cc8522020-04-20 14:54:42 +02001662 goto out;
Christopher Fauletfd6c2292020-03-25 18:20:15 +01001663
Amaury Denoyelle0519bd42020-11-13 12:34:56 +01001664 check_type = srv->check.tcpcheck_rules->flags & TCPCHK_RULES_PROTO_CHK;
Christopher Fauletf50f4e92020-03-30 19:52:29 +02001665
Amaury Denoyelle7d098be2022-03-09 14:20:10 +01001666 if (!(srv->flags & SRV_F_DYNAMIC)) {
1667 /* If neither a port nor an addr was specified and no check
1668 * transport layer is forced, then the transport layer used by
1669 * the checks is the same as for the production traffic.
1670 * Otherwise we use raw_sock by default, unless one is
1671 * specified.
1672 */
1673 if (!srv->check.port && !is_addr(&srv->check.addr)) {
1674 if (!srv->check.use_ssl && srv->use_ssl != -1) {
1675 srv->check.use_ssl = srv->use_ssl;
1676 srv->check.xprt = srv->xprt;
1677 }
1678 else if (srv->check.use_ssl == 1)
1679 srv->check.xprt = xprt_get(XPRT_SSL);
1680 srv->check.send_proxy |= (srv->pp_opts);
Christopher Fauletfd6c2292020-03-25 18:20:15 +01001681 }
Christopher Faulet61cc8522020-04-20 14:54:42 +02001682 else if (srv->check.use_ssl == 1)
1683 srv->check.xprt = xprt_get(XPRT_SSL);
Amaury Denoyelle7d098be2022-03-09 14:20:10 +01001684 }
1685 else {
1686 /* For dynamic servers, check-ssl and check-send-proxy must be
Ilya Shipitsin3b64a282022-07-29 22:26:53 +05001687 * explicitly defined even if the check port was not
Amaury Denoyelle7d098be2022-03-09 14:20:10 +01001688 * overridden.
1689 */
1690 if (srv->check.use_ssl == 1)
1691 srv->check.xprt = xprt_get(XPRT_SSL);
Christopher Fauletfd6c2292020-03-25 18:20:15 +01001692 }
1693
Christopher Faulet12882cf2020-04-23 15:50:18 +02001694 /* Inherit the mux protocol from the server if not already defined for
1695 * the check
1696 */
Amaury Denoyelle0519bd42020-11-13 12:34:56 +01001697 if (srv->mux_proto && !srv->check.mux_proto &&
1698 ((srv->mux_proto->mode == PROTO_MODE_HTTP && check_type == TCPCHK_RULES_HTTP_CHK) ||
1699 (srv->mux_proto->mode == PROTO_MODE_TCP && check_type != TCPCHK_RULES_HTTP_CHK))) {
Christopher Faulet12882cf2020-04-23 15:50:18 +02001700 srv->check.mux_proto = srv->mux_proto;
Amaury Denoyelle0519bd42020-11-13 12:34:56 +01001701 }
Amaury Denoyelle7c148902020-11-13 12:34:57 +01001702 /* test that check proto is valid if explicitly defined */
1703 else if (srv->check.mux_proto &&
1704 ((srv->check.mux_proto->mode == PROTO_MODE_HTTP && check_type != TCPCHK_RULES_HTTP_CHK) ||
1705 (srv->check.mux_proto->mode == PROTO_MODE_TCP && check_type == TCPCHK_RULES_HTTP_CHK))) {
1706 ha_alert("config: %s '%s': server '%s' uses an incompatible MUX protocol for the selected check type\n",
1707 proxy_type_str(srv->proxy), srv->proxy->id, srv->id);
1708 ret |= ERR_ALERT | ERR_FATAL;
1709 goto out;
1710 }
Christopher Faulet12882cf2020-04-23 15:50:18 +02001711
Christopher Faulet61cc8522020-04-20 14:54:42 +02001712 /* validate <srv> server health-check settings */
Christopher Fauletf50f4e92020-03-30 19:52:29 +02001713
Christopher Faulet61cc8522020-04-20 14:54:42 +02001714 /* We need at least a service port, a check port or the first tcp-check
1715 * rule must be a 'connect' one when checking an IPv4/IPv6 server.
1716 */
1717 if ((srv_check_healthcheck_port(&srv->check) != 0) ||
1718 (!is_inet_addr(&srv->check.addr) && (is_addr(&srv->check.addr) || !is_inet_addr(&srv->addr))))
1719 goto init;
Christopher Fauletf50f4e92020-03-30 19:52:29 +02001720
Christopher Faulet61cc8522020-04-20 14:54:42 +02001721 if (!srv->proxy->tcpcheck_rules.list || LIST_ISEMPTY(srv->proxy->tcpcheck_rules.list)) {
1722 ha_alert("config: %s '%s': server '%s' has neither service port nor check port.\n",
1723 proxy_type_str(srv->proxy), srv->proxy->id, srv->id);
1724 ret |= ERR_ALERT | ERR_ABORT;
1725 goto out;
1726 }
Christopher Fauletfd6c2292020-03-25 18:20:15 +01001727
Christopher Faulet61cc8522020-04-20 14:54:42 +02001728 /* search the first action (connect / send / expect) in the list */
1729 r = get_first_tcpcheck_rule(&srv->proxy->tcpcheck_rules);
1730 if (!r || (r->action != TCPCHK_ACT_CONNECT) || (!r->connect.port && !get_host_port(&r->connect.addr))) {
1731 ha_alert("config: %s '%s': server '%s' has neither service port nor check port "
1732 "nor tcp_check rule 'connect' with port information.\n",
1733 proxy_type_str(srv->proxy), srv->proxy->id, srv->id);
1734 ret |= ERR_ALERT | ERR_ABORT;
1735 goto out;
1736 }
Christopher Fauletfd6c2292020-03-25 18:20:15 +01001737
Christopher Faulet61cc8522020-04-20 14:54:42 +02001738 /* scan the tcp-check ruleset to ensure a port has been configured */
1739 list_for_each_entry(r, srv->proxy->tcpcheck_rules.list, list) {
Willy Tarreauacff3092021-07-22 11:06:41 +02001740 if ((r->action == TCPCHK_ACT_CONNECT) && (!r->connect.port && !get_host_port(&r->connect.addr))) {
Christopher Faulet61cc8522020-04-20 14:54:42 +02001741 ha_alert("config: %s '%s': server '%s' has neither service port nor check port, "
1742 "and a tcp_check rule 'connect' with no port information.\n",
1743 proxy_type_str(srv->proxy), srv->proxy->id, srv->id);
1744 ret |= ERR_ALERT | ERR_ABORT;
1745 goto out;
Christopher Faulete5870d82020-04-15 11:32:03 +02001746 }
Christopher Faulete5870d82020-04-15 11:32:03 +02001747 }
1748
Christopher Faulet61cc8522020-04-20 14:54:42 +02001749 init:
Christopher Faulet61cc8522020-04-20 14:54:42 +02001750 err = init_check(&srv->check, srv->proxy->options2 & PR_O2_CHK_ANY);
1751 if (err) {
1752 ha_alert("config: %s '%s': unable to init check for server '%s' (%s).\n",
1753 proxy_type_str(srv->proxy), srv->proxy->id, srv->id, err);
1754 ret |= ERR_ALERT | ERR_ABORT;
1755 goto out;
Christopher Faulete5870d82020-04-15 11:32:03 +02001756 }
Christopher Faulet61cc8522020-04-20 14:54:42 +02001757 srv->check.state |= CHK_ST_CONFIGURED | CHK_ST_ENABLED;
Amaury Denoyellebc2ebfa2021-08-25 15:34:53 +02001758 srv_take(srv);
Amaury Denoyelle403dce82021-07-29 15:39:43 +02001759
1760 /* Only increment maxsock for servers from the configuration. Dynamic
1761 * servers at the moment are not taken into account for the estimation
1762 * of the resources limits.
1763 */
1764 if (global.mode & MODE_STARTING)
1765 global.maxsock++;
Christopher Faulete5870d82020-04-15 11:32:03 +02001766
Christopher Faulet61cc8522020-04-20 14:54:42 +02001767 out:
1768 return ret;
Christopher Faulete5870d82020-04-15 11:32:03 +02001769}
1770
Christopher Faulet61cc8522020-04-20 14:54:42 +02001771/* Initializes an agent-check attached to the server <srv>. Non-zero is returned
1772 * if an error occurred.
1773 */
Amaury Denoyelle3c2ab1a2021-07-22 16:04:40 +02001774int init_srv_agent_check(struct server *srv)
Christopher Faulete5870d82020-04-15 11:32:03 +02001775{
Christopher Faulet61cc8522020-04-20 14:54:42 +02001776 struct tcpcheck_rule *chk;
1777 const char *err;
Christopher Fauletfc633b62020-11-06 15:24:23 +01001778 int ret = ERR_NONE;
Christopher Faulete5870d82020-04-15 11:32:03 +02001779
Christopher Faulet6ecd5932021-01-12 17:29:45 +01001780 if (!srv->do_agent || !(srv->proxy->cap & PR_CAP_BE))
Christopher Faulet61cc8522020-04-20 14:54:42 +02001781 goto out;
Christopher Faulete5870d82020-04-15 11:32:03 +02001782
Ilya Shipitsinc02a23f2020-05-06 00:53:22 +05001783 /* If there is no connect rule preceding all send / expect rules, an
Christopher Faulet61cc8522020-04-20 14:54:42 +02001784 * implicit one is inserted before all others.
1785 */
1786 chk = get_first_tcpcheck_rule(srv->agent.tcpcheck_rules);
1787 if (!chk || chk->action != TCPCHK_ACT_CONNECT) {
1788 chk = calloc(1, sizeof(*chk));
1789 if (!chk) {
Amaury Denoyelle11124302021-06-04 18:22:08 +02001790 ha_alert("%s '%s': unable to add implicit tcp-check connect rule"
Christopher Faulet61cc8522020-04-20 14:54:42 +02001791 " to agent-check for server '%s' (out of memory).\n",
1792 proxy_type_str(srv->proxy), srv->proxy->id, srv->id);
1793 ret |= ERR_ALERT | ERR_FATAL;
1794 goto out;
Christopher Faulete5870d82020-04-15 11:32:03 +02001795 }
Christopher Faulet61cc8522020-04-20 14:54:42 +02001796 chk->action = TCPCHK_ACT_CONNECT;
1797 chk->connect.options = (TCPCHK_OPT_DEFAULT_CONNECT|TCPCHK_OPT_IMPLICIT);
Willy Tarreau2b718102021-04-21 07:32:39 +02001798 LIST_INSERT(srv->agent.tcpcheck_rules->list, &chk->list);
Christopher Faulete5870d82020-04-15 11:32:03 +02001799 }
1800
Christopher Faulet871dd822022-08-24 11:38:03 +02001801 /* <chk> is always defined here and it is a CONNECT action. If there is
1802 * a preset variable, it means there is an agent string defined and data
1803 * will be sent after the connect.
1804 */
1805 if (!LIST_ISEMPTY(&srv->agent.tcpcheck_rules->preset_vars))
1806 chk->connect.options |= TCPCHK_OPT_HAS_DATA;
1807
Christopher Faulete5870d82020-04-15 11:32:03 +02001808
Christopher Faulet61cc8522020-04-20 14:54:42 +02001809 err = init_check(&srv->agent, PR_O2_TCPCHK_CHK);
1810 if (err) {
1811 ha_alert("config: %s '%s': unable to init agent-check for server '%s' (%s).\n",
1812 proxy_type_str(srv->proxy), srv->proxy->id, srv->id, err);
1813 ret |= ERR_ALERT | ERR_ABORT;
1814 goto out;
Christopher Faulete5870d82020-04-15 11:32:03 +02001815 }
1816
Christopher Faulet61cc8522020-04-20 14:54:42 +02001817 if (!srv->agent.inter)
1818 srv->agent.inter = srv->check.inter;
1819
1820 srv->agent.state |= CHK_ST_CONFIGURED | CHK_ST_ENABLED | CHK_ST_AGENT;
Amaury Denoyellebc2ebfa2021-08-25 15:34:53 +02001821 srv_take(srv);
Amaury Denoyelle403dce82021-07-29 15:39:43 +02001822
1823 /* Only increment maxsock for servers from the configuration. Dynamic
1824 * servers at the moment are not taken into account for the estimation
1825 * of the resources limits.
1826 */
1827 if (global.mode & MODE_STARTING)
1828 global.maxsock++;
Christopher Faulet61cc8522020-04-20 14:54:42 +02001829
1830 out:
1831 return ret;
Christopher Faulete5870d82020-04-15 11:32:03 +02001832}
1833
Christopher Faulet61cc8522020-04-20 14:54:42 +02001834static void deinit_srv_check(struct server *srv)
1835{
Aurelien DARRAGON32483ec2023-03-09 11:56:14 +01001836 if (srv->check.state & CHK_ST_CONFIGURED) {
Christopher Faulet61cc8522020-04-20 14:54:42 +02001837 free_check(&srv->check);
Aurelien DARRAGON32483ec2023-03-09 11:56:14 +01001838 /* it is safe to drop now since the main server reference is still held by the proxy */
1839 srv_drop(srv);
1840 }
Christopher Faulet61cc8522020-04-20 14:54:42 +02001841 srv->check.state &= ~CHK_ST_CONFIGURED & ~CHK_ST_ENABLED;
1842 srv->do_check = 0;
1843}
Christopher Faulete5870d82020-04-15 11:32:03 +02001844
Christopher Faulet61cc8522020-04-20 14:54:42 +02001845
1846static void deinit_srv_agent_check(struct server *srv)
1847{
Aurelien DARRAGON32483ec2023-03-09 11:56:14 +01001848 if (srv->agent.state & CHK_ST_CONFIGURED) {
Christopher Faulet61cc8522020-04-20 14:54:42 +02001849 free_check(&srv->agent);
Aurelien DARRAGON32483ec2023-03-09 11:56:14 +01001850 /* it is safe to drop now since the main server reference is still held by the proxy */
1851 srv_drop(srv);
1852 }
Christopher Faulet61cc8522020-04-20 14:54:42 +02001853
1854 srv->agent.state &= ~CHK_ST_CONFIGURED & ~CHK_ST_ENABLED & ~CHK_ST_AGENT;
1855 srv->do_agent = 0;
Christopher Faulete5870d82020-04-15 11:32:03 +02001856}
1857
Willy Tarreaucee013e2020-06-05 11:40:38 +02001858REGISTER_POST_SERVER_CHECK(init_srv_check);
1859REGISTER_POST_SERVER_CHECK(init_srv_agent_check);
Willy Tarreaucee013e2020-06-05 11:40:38 +02001860REGISTER_POST_CHECK(start_checks);
Christopher Faulet61cc8522020-04-20 14:54:42 +02001861
Willy Tarreaucee013e2020-06-05 11:40:38 +02001862REGISTER_SERVER_DEINIT(deinit_srv_check);
1863REGISTER_SERVER_DEINIT(deinit_srv_agent_check);
Christopher Faulet61cc8522020-04-20 14:54:42 +02001864
Christopher Faulet61cc8522020-04-20 14:54:42 +02001865
1866/**************************************************************************/
1867/************************** Check sample fetches **************************/
1868/**************************************************************************/
Christopher Fauletfd6c2292020-03-25 18:20:15 +01001869
Christopher Faulet61cc8522020-04-20 14:54:42 +02001870static struct sample_fetch_kw_list smp_kws = {ILH, {
Christopher Faulet61cc8522020-04-20 14:54:42 +02001871 { /* END */ },
1872}};
1873
1874INITCALL1(STG_REGISTER, sample_register_fetches, &smp_kws);
1875
1876
1877/**************************************************************************/
1878/************************ Check's parsing functions ***********************/
1879/**************************************************************************/
Christopher Fauletce8111e2020-04-06 15:04:11 +02001880/* Parse the "addr" server keyword */
1881static int srv_parse_addr(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
1882 char **errmsg)
1883{
1884 struct sockaddr_storage *sk;
Christopher Fauletce8111e2020-04-06 15:04:11 +02001885 int port1, port2, err_code = 0;
1886
1887
1888 if (!*args[*cur_arg+1]) {
1889 memprintf(errmsg, "'%s' expects <ipv4|ipv6> as argument.", args[*cur_arg]);
1890 goto error;
1891 }
1892
Willy Tarreau65ec4e32020-09-16 19:17:08 +02001893 sk = str2sa_range(args[*cur_arg+1], NULL, &port1, &port2, NULL, NULL, errmsg, NULL, NULL,
1894 PA_O_RESOLVE | PA_O_PORT_OK | PA_O_STREAM | PA_O_CONNECT);
Christopher Fauletce8111e2020-04-06 15:04:11 +02001895 if (!sk) {
1896 memprintf(errmsg, "'%s' : %s", args[*cur_arg], *errmsg);
1897 goto error;
1898 }
1899
William Dauchy1c921cd2021-02-03 22:30:08 +01001900 srv->check.addr = *sk;
1901 /* if agentaddr was never set, we can use addr */
1902 if (!(srv->flags & SRV_F_AGENTADDR))
1903 srv->agent.addr = *sk;
Christopher Fauletce8111e2020-04-06 15:04:11 +02001904
1905 out:
1906 return err_code;
1907
1908 error:
1909 err_code |= ERR_ALERT | ERR_FATAL;
1910 goto out;
1911}
1912
Christopher Fauletcbba66c2020-04-06 14:26:30 +02001913/* Parse the "agent-addr" server keyword */
1914static int srv_parse_agent_addr(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
1915 char **errmsg)
1916{
William Dauchy1c921cd2021-02-03 22:30:08 +01001917 struct sockaddr_storage sk;
Christopher Fauletcbba66c2020-04-06 14:26:30 +02001918 int err_code = 0;
1919
1920 if (!*(args[*cur_arg+1])) {
1921 memprintf(errmsg, "'%s' expects an address as argument.", args[*cur_arg]);
1922 goto error;
1923 }
William Dauchy1c921cd2021-02-03 22:30:08 +01001924 memset(&sk, 0, sizeof(sk));
1925 if (str2ip(args[*cur_arg + 1], &sk) == NULL) {
Christopher Fauletcbba66c2020-04-06 14:26:30 +02001926 memprintf(errmsg, "parsing agent-addr failed. Check if '%s' is correct address.", args[*cur_arg+1]);
1927 goto error;
1928 }
William Dauchy1c921cd2021-02-03 22:30:08 +01001929 set_srv_agent_addr(srv, &sk);
Christopher Fauletcbba66c2020-04-06 14:26:30 +02001930
1931 out:
1932 return err_code;
1933
1934 error:
1935 err_code |= ERR_ALERT | ERR_FATAL;
1936 goto out;
1937}
1938
1939/* Parse the "agent-check" server keyword */
1940static int srv_parse_agent_check(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
1941 char **errmsg)
1942{
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001943 struct tcpcheck_ruleset *rs = NULL;
1944 struct tcpcheck_rules *rules = srv->agent.tcpcheck_rules;
1945 struct tcpcheck_rule *chk;
1946 int err_code = 0;
1947
1948 if (srv->do_agent)
1949 goto out;
1950
Christopher Faulet6ecd5932021-01-12 17:29:45 +01001951 if (!(curpx->cap & PR_CAP_BE)) {
1952 memprintf(errmsg, "'%s' ignored because %s '%s' has no backend capability",
1953 args[*cur_arg], proxy_type_str(curpx), curpx->id);
1954 return ERR_WARN;
1955 }
1956
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001957 if (!rules) {
1958 rules = calloc(1, sizeof(*rules));
1959 if (!rules) {
1960 memprintf(errmsg, "out of memory.");
1961 goto error;
1962 }
1963 LIST_INIT(&rules->preset_vars);
1964 srv->agent.tcpcheck_rules = rules;
1965 }
1966 rules->list = NULL;
1967 rules->flags = 0;
1968
Christopher Faulet61cc8522020-04-20 14:54:42 +02001969 rs = find_tcpcheck_ruleset("*agent-check");
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001970 if (rs)
1971 goto ruleset_found;
1972
Christopher Faulet61cc8522020-04-20 14:54:42 +02001973 rs = create_tcpcheck_ruleset("*agent-check");
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001974 if (rs == NULL) {
1975 memprintf(errmsg, "out of memory.");
1976 goto error;
1977 }
1978
Christopher Fauletb50b3e62020-05-05 18:43:43 +02001979 chk = parse_tcpcheck_send((char *[]){"tcp-check", "send-lf", "%[var(check.agent_string)]", ""},
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001980 1, curpx, &rs->rules, srv->conf.file, srv->conf.line, errmsg);
1981 if (!chk) {
1982 memprintf(errmsg, "'%s': %s", args[*cur_arg], *errmsg);
1983 goto error;
1984 }
1985 chk->index = 0;
Willy Tarreau2b718102021-04-21 07:32:39 +02001986 LIST_APPEND(&rs->rules, &chk->list);
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001987
1988 chk = parse_tcpcheck_expect((char *[]){"tcp-check", "expect", "custom", ""},
Christopher Faulete5870d82020-04-15 11:32:03 +02001989 1, curpx, &rs->rules, TCPCHK_RULES_AGENT_CHK,
1990 srv->conf.file, srv->conf.line, errmsg);
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001991 if (!chk) {
1992 memprintf(errmsg, "'%s': %s", args[*cur_arg], *errmsg);
1993 goto error;
1994 }
1995 chk->expect.custom = tcpcheck_agent_expect_reply;
1996 chk->index = 1;
Willy Tarreau2b718102021-04-21 07:32:39 +02001997 LIST_APPEND(&rs->rules, &chk->list);
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001998
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02001999 ruleset_found:
2000 rules->list = &rs->rules;
Christopher Faulet1faf18a2020-11-25 16:43:12 +01002001 rules->flags &= ~(TCPCHK_RULES_PROTO_CHK|TCPCHK_RULES_UNUSED_RS);
Christopher Faulet404f9192020-04-09 23:13:54 +02002002 rules->flags |= TCPCHK_RULES_AGENT_CHK;
Christopher Fauletcbba66c2020-04-06 14:26:30 +02002003 srv->do_agent = 1;
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02002004
2005 out:
Dirkjan Bussinkdfee2172021-06-18 19:57:49 +00002006 return err_code;
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02002007
2008 error:
2009 deinit_srv_agent_check(srv);
Christopher Faulet61cc8522020-04-20 14:54:42 +02002010 free_tcpcheck_ruleset(rs);
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02002011 err_code |= ERR_ALERT | ERR_FATAL;
2012 goto out;
Christopher Fauletcbba66c2020-04-06 14:26:30 +02002013}
2014
2015/* Parse the "agent-inter" server keyword */
2016static int srv_parse_agent_inter(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
2017 char **errmsg)
2018{
2019 const char *err = NULL;
2020 unsigned int delay;
2021 int err_code = 0;
2022
2023 if (!*(args[*cur_arg+1])) {
2024 memprintf(errmsg, "'%s' expects a delay as argument.", args[*cur_arg]);
2025 goto error;
2026 }
2027
2028 err = parse_time_err(args[*cur_arg+1], &delay, TIME_UNIT_MS);
2029 if (err == PARSE_TIME_OVER) {
2030 memprintf(errmsg, "timer overflow in argument <%s> to <%s> of server %s, maximum value is 2147483647 ms (~24.8 days).",
2031 args[*cur_arg+1], args[*cur_arg], srv->id);
2032 goto error;
2033 }
2034 else if (err == PARSE_TIME_UNDER) {
2035 memprintf(errmsg, "timer underflow in argument <%s> to <%s> of server %s, minimum non-null value is 1 ms.",
2036 args[*cur_arg+1], args[*cur_arg], srv->id);
2037 goto error;
2038 }
2039 else if (err) {
2040 memprintf(errmsg, "unexpected character '%c' in 'agent-inter' argument of server %s.",
2041 *err, srv->id);
2042 goto error;
2043 }
2044 if (delay <= 0) {
2045 memprintf(errmsg, "invalid value %d for argument '%s' of server %s.",
2046 delay, args[*cur_arg], srv->id);
2047 goto error;
2048 }
2049 srv->agent.inter = delay;
2050
2051 out:
2052 return err_code;
2053
2054 error:
2055 err_code |= ERR_ALERT | ERR_FATAL;
2056 goto out;
2057}
2058
2059/* Parse the "agent-port" server keyword */
2060static int srv_parse_agent_port(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
2061 char **errmsg)
2062{
2063 int err_code = 0;
2064
2065 if (!*(args[*cur_arg+1])) {
2066 memprintf(errmsg, "'%s' expects a port number as argument.", args[*cur_arg]);
2067 goto error;
2068 }
2069
Amaury Denoyelle403dce82021-07-29 15:39:43 +02002070 /* Only increment maxsock for servers from the configuration. Dynamic
2071 * servers at the moment are not taken into account for the estimation
2072 * of the resources limits.
2073 */
2074 if (global.mode & MODE_STARTING)
2075 global.maxsock++;
2076
William Dauchy4858fb22021-02-03 22:30:09 +01002077 set_srv_agent_port(srv, atol(args[*cur_arg + 1]));
Christopher Fauletcbba66c2020-04-06 14:26:30 +02002078
2079 out:
2080 return err_code;
2081
2082 error:
2083 err_code |= ERR_ALERT | ERR_FATAL;
2084 goto out;
2085}
2086
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02002087int set_srv_agent_send(struct server *srv, const char *send)
2088{
2089 struct tcpcheck_rules *rules = srv->agent.tcpcheck_rules;
2090 struct tcpcheck_var *var = NULL;
2091 char *str;
2092
2093 str = strdup(send);
Christopher Fauletb61caf42020-04-21 10:57:42 +02002094 var = create_tcpcheck_var(ist("check.agent_string"));
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02002095 if (str == NULL || var == NULL)
2096 goto error;
2097
2098 free_tcpcheck_vars(&rules->preset_vars);
2099
2100 var->data.type = SMP_T_STR;
2101 var->data.u.str.area = str;
2102 var->data.u.str.data = strlen(str);
2103 LIST_INIT(&var->list);
Willy Tarreau2b718102021-04-21 07:32:39 +02002104 LIST_APPEND(&rules->preset_vars, &var->list);
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02002105
2106 return 1;
2107
2108 error:
2109 free(str);
2110 free(var);
2111 return 0;
2112}
Christopher Fauletcbba66c2020-04-06 14:26:30 +02002113
2114/* Parse the "agent-send" server keyword */
2115static int srv_parse_agent_send(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
2116 char **errmsg)
2117{
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02002118 struct tcpcheck_rules *rules = srv->agent.tcpcheck_rules;
Christopher Fauletcbba66c2020-04-06 14:26:30 +02002119 int err_code = 0;
2120
2121 if (!*(args[*cur_arg+1])) {
2122 memprintf(errmsg, "'%s' expects a string as argument.", args[*cur_arg]);
2123 goto error;
2124 }
2125
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02002126 if (!rules) {
2127 rules = calloc(1, sizeof(*rules));
2128 if (!rules) {
2129 memprintf(errmsg, "out of memory.");
2130 goto error;
2131 }
2132 LIST_INIT(&rules->preset_vars);
2133 srv->agent.tcpcheck_rules = rules;
2134 }
2135
2136 if (!set_srv_agent_send(srv, args[*cur_arg+1])) {
Christopher Fauletcbba66c2020-04-06 14:26:30 +02002137 memprintf(errmsg, "out of memory.");
2138 goto error;
2139 }
2140
2141 out:
2142 return err_code;
2143
2144 error:
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02002145 deinit_srv_agent_check(srv);
Christopher Fauletcbba66c2020-04-06 14:26:30 +02002146 err_code |= ERR_ALERT | ERR_FATAL;
2147 goto out;
2148}
2149
2150/* Parse the "no-agent-send" server keyword */
2151static int srv_parse_no_agent_check(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
2152 char **errmsg)
2153{
Christopher Faulet0ae3d1d2020-04-06 17:54:24 +02002154 deinit_srv_agent_check(srv);
Christopher Fauletcbba66c2020-04-06 14:26:30 +02002155 return 0;
2156}
2157
Christopher Fauletce8111e2020-04-06 15:04:11 +02002158/* Parse the "check" server keyword */
2159static int srv_parse_check(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
2160 char **errmsg)
2161{
Christopher Faulet6ecd5932021-01-12 17:29:45 +01002162 if (!(curpx->cap & PR_CAP_BE)) {
2163 memprintf(errmsg, "'%s' ignored because %s '%s' has no backend capability",
2164 args[*cur_arg], proxy_type_str(curpx), curpx->id);
2165 return ERR_WARN;
2166 }
2167
Christopher Fauletce8111e2020-04-06 15:04:11 +02002168 srv->do_check = 1;
2169 return 0;
2170}
2171
2172/* Parse the "check-send-proxy" server keyword */
2173static int srv_parse_check_send_proxy(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
2174 char **errmsg)
2175{
2176 srv->check.send_proxy = 1;
2177 return 0;
2178}
2179
2180/* Parse the "check-via-socks4" server keyword */
2181static int srv_parse_check_via_socks4(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
2182 char **errmsg)
2183{
2184 srv->check.via_socks4 = 1;
2185 return 0;
2186}
2187
2188/* Parse the "no-check" server keyword */
2189static int srv_parse_no_check(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
2190 char **errmsg)
2191{
2192 deinit_srv_check(srv);
2193 return 0;
2194}
2195
2196/* Parse the "no-check-send-proxy" server keyword */
2197static int srv_parse_no_check_send_proxy(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
2198 char **errmsg)
2199{
2200 srv->check.send_proxy = 0;
2201 return 0;
2202}
2203
Christopher Fauletedc6ed92020-04-23 16:27:59 +02002204/* parse the "check-proto" server keyword */
2205static int srv_parse_check_proto(char **args, int *cur_arg,
2206 struct proxy *px, struct server *newsrv, char **err)
2207{
2208 int err_code = 0;
2209
2210 if (!*args[*cur_arg + 1]) {
2211 memprintf(err, "'%s' : missing value", args[*cur_arg]);
2212 goto error;
2213 }
Tim Duesterhusdcf753a2021-03-04 17:31:47 +01002214 newsrv->check.mux_proto = get_mux_proto(ist(args[*cur_arg + 1]));
Christopher Fauletedc6ed92020-04-23 16:27:59 +02002215 if (!newsrv->check.mux_proto) {
2216 memprintf(err, "'%s' : unknown MUX protocol '%s'", args[*cur_arg], args[*cur_arg+1]);
2217 goto error;
2218 }
2219
2220 out:
2221 return err_code;
2222
2223 error:
2224 err_code |= ERR_ALERT | ERR_FATAL;
2225 goto out;
2226}
2227
2228
Christopher Fauletce8111e2020-04-06 15:04:11 +02002229/* Parse the "rise" server keyword */
2230static int srv_parse_check_rise(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
2231 char **errmsg)
2232{
2233 int err_code = 0;
2234
2235 if (!*args[*cur_arg + 1]) {
2236 memprintf(errmsg, "'%s' expects an integer argument.", args[*cur_arg]);
2237 goto error;
2238 }
2239
2240 srv->check.rise = atol(args[*cur_arg+1]);
2241 if (srv->check.rise <= 0) {
2242 memprintf(errmsg, "'%s' has to be > 0.", args[*cur_arg]);
2243 goto error;
2244 }
2245
2246 if (srv->check.health)
2247 srv->check.health = srv->check.rise;
2248
2249 out:
2250 return err_code;
2251
2252 error:
2253 deinit_srv_agent_check(srv);
2254 err_code |= ERR_ALERT | ERR_FATAL;
2255 goto out;
Christopher Fauletce8111e2020-04-06 15:04:11 +02002256}
2257
2258/* Parse the "fall" server keyword */
2259static int srv_parse_check_fall(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
2260 char **errmsg)
2261{
2262 int err_code = 0;
2263
2264 if (!*args[*cur_arg + 1]) {
2265 memprintf(errmsg, "'%s' expects an integer argument.", args[*cur_arg]);
2266 goto error;
2267 }
2268
2269 srv->check.fall = atol(args[*cur_arg+1]);
2270 if (srv->check.fall <= 0) {
2271 memprintf(errmsg, "'%s' has to be > 0.", args[*cur_arg]);
2272 goto error;
2273 }
2274
2275 out:
2276 return err_code;
2277
2278 error:
2279 deinit_srv_agent_check(srv);
2280 err_code |= ERR_ALERT | ERR_FATAL;
2281 goto out;
Christopher Fauletce8111e2020-04-06 15:04:11 +02002282}
2283
2284/* Parse the "inter" server keyword */
2285static int srv_parse_check_inter(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
2286 char **errmsg)
2287{
2288 const char *err = NULL;
2289 unsigned int delay;
2290 int err_code = 0;
2291
2292 if (!*(args[*cur_arg+1])) {
2293 memprintf(errmsg, "'%s' expects a delay as argument.", args[*cur_arg]);
2294 goto error;
2295 }
2296
2297 err = parse_time_err(args[*cur_arg+1], &delay, TIME_UNIT_MS);
2298 if (err == PARSE_TIME_OVER) {
2299 memprintf(errmsg, "timer overflow in argument <%s> to <%s> of server %s, maximum value is 2147483647 ms (~24.8 days).",
2300 args[*cur_arg+1], args[*cur_arg], srv->id);
2301 goto error;
2302 }
2303 else if (err == PARSE_TIME_UNDER) {
2304 memprintf(errmsg, "timer underflow in argument <%s> to <%s> of server %s, minimum non-null value is 1 ms.",
2305 args[*cur_arg+1], args[*cur_arg], srv->id);
2306 goto error;
2307 }
2308 else if (err) {
2309 memprintf(errmsg, "unexpected character '%c' in 'agent-inter' argument of server %s.",
2310 *err, srv->id);
2311 goto error;
2312 }
2313 if (delay <= 0) {
2314 memprintf(errmsg, "invalid value %d for argument '%s' of server %s.",
2315 delay, args[*cur_arg], srv->id);
2316 goto error;
2317 }
2318 srv->check.inter = delay;
2319
2320 out:
2321 return err_code;
2322
2323 error:
2324 err_code |= ERR_ALERT | ERR_FATAL;
2325 goto out;
2326}
2327
2328
2329/* Parse the "fastinter" server keyword */
2330static int srv_parse_check_fastinter(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
2331 char **errmsg)
2332{
2333 const char *err = NULL;
2334 unsigned int delay;
2335 int err_code = 0;
2336
2337 if (!*(args[*cur_arg+1])) {
2338 memprintf(errmsg, "'%s' expects a delay as argument.", args[*cur_arg]);
2339 goto error;
2340 }
2341
2342 err = parse_time_err(args[*cur_arg+1], &delay, TIME_UNIT_MS);
2343 if (err == PARSE_TIME_OVER) {
2344 memprintf(errmsg, "timer overflow in argument <%s> to <%s> of server %s, maximum value is 2147483647 ms (~24.8 days).",
2345 args[*cur_arg+1], args[*cur_arg], srv->id);
2346 goto error;
2347 }
2348 else if (err == PARSE_TIME_UNDER) {
2349 memprintf(errmsg, "timer underflow in argument <%s> to <%s> of server %s, minimum non-null value is 1 ms.",
2350 args[*cur_arg+1], args[*cur_arg], srv->id);
2351 goto error;
2352 }
2353 else if (err) {
2354 memprintf(errmsg, "unexpected character '%c' in 'agent-inter' argument of server %s.",
2355 *err, srv->id);
2356 goto error;
2357 }
2358 if (delay <= 0) {
2359 memprintf(errmsg, "invalid value %d for argument '%s' of server %s.",
2360 delay, args[*cur_arg], srv->id);
2361 goto error;
2362 }
2363 srv->check.fastinter = delay;
2364
2365 out:
2366 return err_code;
2367
2368 error:
2369 err_code |= ERR_ALERT | ERR_FATAL;
2370 goto out;
2371}
2372
2373
2374/* Parse the "downinter" server keyword */
2375static int srv_parse_check_downinter(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
2376 char **errmsg)
2377{
2378 const char *err = NULL;
2379 unsigned int delay;
2380 int err_code = 0;
2381
2382 if (!*(args[*cur_arg+1])) {
2383 memprintf(errmsg, "'%s' expects a delay as argument.", args[*cur_arg]);
2384 goto error;
2385 }
2386
2387 err = parse_time_err(args[*cur_arg+1], &delay, TIME_UNIT_MS);
2388 if (err == PARSE_TIME_OVER) {
2389 memprintf(errmsg, "timer overflow in argument <%s> to <%s> of server %s, maximum value is 2147483647 ms (~24.8 days).",
2390 args[*cur_arg+1], args[*cur_arg], srv->id);
2391 goto error;
2392 }
2393 else if (err == PARSE_TIME_UNDER) {
2394 memprintf(errmsg, "timer underflow in argument <%s> to <%s> of server %s, minimum non-null value is 1 ms.",
2395 args[*cur_arg+1], args[*cur_arg], srv->id);
2396 goto error;
2397 }
2398 else if (err) {
2399 memprintf(errmsg, "unexpected character '%c' in 'agent-inter' argument of server %s.",
2400 *err, srv->id);
2401 goto error;
2402 }
2403 if (delay <= 0) {
2404 memprintf(errmsg, "invalid value %d for argument '%s' of server %s.",
2405 delay, args[*cur_arg], srv->id);
2406 goto error;
2407 }
2408 srv->check.downinter = delay;
2409
2410 out:
2411 return err_code;
2412
2413 error:
2414 err_code |= ERR_ALERT | ERR_FATAL;
2415 goto out;
2416}
2417
2418/* Parse the "port" server keyword */
2419static int srv_parse_check_port(char **args, int *cur_arg, struct proxy *curpx, struct server *srv,
2420 char **errmsg)
2421{
2422 int err_code = 0;
2423
2424 if (!*(args[*cur_arg+1])) {
2425 memprintf(errmsg, "'%s' expects a port number as argument.", args[*cur_arg]);
2426 goto error;
2427 }
2428
Amaury Denoyelle403dce82021-07-29 15:39:43 +02002429 /* Only increment maxsock for servers from the configuration. Dynamic
2430 * servers at the moment are not taken into account for the estimation
2431 * of the resources limits.
2432 */
2433 if (global.mode & MODE_STARTING)
2434 global.maxsock++;
2435
Christopher Fauletce8111e2020-04-06 15:04:11 +02002436 srv->check.port = atol(args[*cur_arg+1]);
William Dauchy4858fb22021-02-03 22:30:09 +01002437 /* if agentport was never set, we can use port */
2438 if (!(srv->flags & SRV_F_AGENTPORT))
2439 srv->agent.port = srv->check.port;
Christopher Fauletce8111e2020-04-06 15:04:11 +02002440
2441 out:
2442 return err_code;
2443
2444 error:
2445 err_code |= ERR_ALERT | ERR_FATAL;
2446 goto out;
2447}
2448
Christopher Fauletcbba66c2020-04-06 14:26:30 +02002449static struct srv_kw_list srv_kws = { "CHK", { }, {
Amaury Denoyelle9ecee0f2021-07-23 16:34:58 +02002450 { "addr", srv_parse_addr, 1, 1, 1 }, /* IP address to send health to or to probe from agent-check */
2451 { "agent-addr", srv_parse_agent_addr, 1, 1, 1 }, /* Enable an auxiliary agent check */
Amaury Denoyelleb65f4ca2021-08-04 11:33:14 +02002452 { "agent-check", srv_parse_agent_check, 0, 1, 1 }, /* Enable agent checks */
Amaury Denoyelle9ecee0f2021-07-23 16:34:58 +02002453 { "agent-inter", srv_parse_agent_inter, 1, 1, 1 }, /* Set the interval between two agent checks */
2454 { "agent-port", srv_parse_agent_port, 1, 1, 1 }, /* Set the TCP port used for agent checks. */
2455 { "agent-send", srv_parse_agent_send, 1, 1, 1 }, /* Set string to send to agent. */
Amaury Denoyelle2fc4d392021-07-22 16:04:59 +02002456 { "check", srv_parse_check, 0, 1, 1 }, /* Enable health checks */
Amaury Denoyelle9ecee0f2021-07-23 16:34:58 +02002457 { "check-proto", srv_parse_check_proto, 1, 1, 1 }, /* Set the mux protocol for health checks */
2458 { "check-send-proxy", srv_parse_check_send_proxy, 0, 1, 1 }, /* Enable PROXY protocol for health checks */
2459 { "check-via-socks4", srv_parse_check_via_socks4, 0, 1, 1 }, /* Enable socks4 proxy for health checks */
Amaury Denoyelle76e10e72021-03-08 17:08:01 +01002460 { "no-agent-check", srv_parse_no_agent_check, 0, 1, 0 }, /* Do not enable any auxiliary agent check */
2461 { "no-check", srv_parse_no_check, 0, 1, 0 }, /* Disable health checks */
2462 { "no-check-send-proxy", srv_parse_no_check_send_proxy, 0, 1, 0 }, /* Disable PROXY protocol for health checks */
Amaury Denoyelle9ecee0f2021-07-23 16:34:58 +02002463 { "rise", srv_parse_check_rise, 1, 1, 1 }, /* Set rise value for health checks */
2464 { "fall", srv_parse_check_fall, 1, 1, 1 }, /* Set fall value for health checks */
2465 { "inter", srv_parse_check_inter, 1, 1, 1 }, /* Set inter value for health checks */
2466 { "fastinter", srv_parse_check_fastinter, 1, 1, 1 }, /* Set fastinter value for health checks */
2467 { "downinter", srv_parse_check_downinter, 1, 1, 1 }, /* Set downinter value for health checks */
2468 { "port", srv_parse_check_port, 1, 1, 1 }, /* Set the TCP port used for health checks. */
Christopher Fauletcbba66c2020-04-06 14:26:30 +02002469 { NULL, NULL, 0 },
2470}};
2471
Christopher Fauletcbba66c2020-04-06 14:26:30 +02002472INITCALL1(STG_REGISTER, srv_register_keywords, &srv_kws);
Christopher Fauletfd6c2292020-03-25 18:20:15 +01002473
Willy Tarreaubd741542010-03-16 18:46:54 +01002474/*
Willy Tarreaubaaee002006-06-26 02:48:02 +02002475 * Local variables:
2476 * c-indent-level: 8
2477 * c-basic-offset: 8
2478 * End:
2479 */