12 #include <sys/types.h>
15 #include <sys/resource.h>
17 #include <sys/prctl.h>
22 #include "web_server.h"
24 #include "appconfig.h"
25 #include "web_client.h"
29 #include "unit_test.h"
31 #include "plugins_d.h"
32 #include "plugin_idlejitter.h"
33 #include "plugin_tc.h"
34 #include "plugin_checks.h"
35 #include "plugin_proc.h"
36 #include "plugin_nfacct.h"
40 extern void *cgroups_main(void *ptr);
44 void netdata_cleanup_and_exit(int ret)
50 // let it log a few more error messages
51 error_log_limit_reset();
54 if(ftruncate(pidfd, 0) != 0)
55 error("Cannot truncate pidfile '%s'.", pidfile);
62 if(unlink(pidfile) != 0)
63 error("Cannot unlink pidfile '%s'.", pidfile);
66 info("NetData exiting. Bye bye...");
70 struct netdata_static_thread {
80 void (*init_routine) (void);
81 void *(*start_routine) (void *);
84 struct netdata_static_thread static_threads[] = {
85 {"tc", "plugins", "tc", 1, NULL, NULL, tc_main},
86 {"idlejitter", "plugins", "idlejitter", 1, NULL, NULL, cpuidlejitter_main},
87 {"proc", "plugins", "proc", 1, NULL, NULL, proc_main},
88 {"cgroups", "plugins", "cgroups", 1, NULL, NULL, cgroups_main},
90 #ifdef INTERNAL_PLUGIN_NFACCT
91 // nfacct requires root access
92 // so, we build it as an external plugin with setuid to root
93 {"nfacct", "plugins", "nfacct", 1, NULL, NULL, nfacct_main},
96 {"plugins.d", NULL, NULL, 1, NULL, NULL, pluginsd_main},
97 {"check", "plugins", "checks", 0, NULL, NULL, checks_main},
98 {"web", NULL, NULL, 1, NULL, NULL, socket_listen_main},
99 {NULL, NULL, NULL, 0, NULL, NULL, NULL}
102 int killpid(pid_t pid, int sig)
105 debug(D_EXIT, "Request to kill pid %d", pid);
108 if(kill(pid, 0) == -1) {
111 error("Request to kill pid %d, but it is not running.", pid);
115 error("Request to kill pid %d, but I do not have enough permissions.", pid);
119 error("Request to kill pid %d, but I received an error.", pid);
127 old = signal(sig, SIG_IGN);
129 error("Cannot overwrite signal handler for signal %d", sig);
133 ret = kill(pid, sig);
135 if(signal(sig, old) == SIG_ERR)
136 error("Cannot restore signal handler for signal %d", sig);
141 error("Cannot kill pid %d, but it is not running.", pid);
145 error("Cannot kill pid %d, but I do not have enough permissions.", pid);
149 error("Cannot kill pid %d, but I received an error.", pid);
162 struct web_client *w;
163 for(w = web_clients; w ; w = w->next) {
164 debug(D_EXIT, "Stopping web client %s", w->client_ip);
165 pthread_cancel(w->thread);
166 pthread_join(w->thread, NULL);
170 for (i = 0; static_threads[i].name != NULL ; i++) {
171 if(static_threads[i].thread) {
172 debug(D_EXIT, "Stopping %s thread", static_threads[i].name);
173 pthread_cancel(*static_threads[i].thread);
174 pthread_join(*static_threads[i].thread, NULL);
175 static_threads[i].thread = NULL;
180 info("Killing tc-qos-helper procees");
181 if(killpid(tc_child_pid, SIGTERM) != -1)
182 waitid(P_PID, (id_t) tc_child_pid, &info, WEXITED);
187 for(cd = pluginsd_root ; cd ; cd = cd->next) {
188 debug(D_EXIT, "Stopping %s plugin thread", cd->id);
189 pthread_cancel(cd->thread);
190 pthread_join(cd->thread, NULL);
192 if(cd->pid && !cd->obsolete) {
193 debug(D_EXIT, "killing %s plugin process", cd->id);
194 if(killpid(cd->pid, SIGTERM) != -1)
195 waitid(P_PID, (id_t) cd->pid, &info, WEXITED);
199 // if, for any reason there is any child exited
201 waitid(P_PID, 0, &info, WEXITED|WNOHANG);
203 debug(D_EXIT, "All threads/childs stopped.");
207 int main(int argc, char **argv)
210 int config_loaded = 0;
212 size_t wanted_stacksize = 0, stacksize = 0;
215 // global initialization
218 // set the name for logging
219 program_name = "netdata";
221 // parse the arguments
222 for(i = 1; i < argc ; i++) {
223 if(strcmp(argv[i], "-c") == 0 && (i+1) < argc) {
224 if(load_config(argv[i+1], 1) != 1) {
225 error("Cannot load configuration file %s.", argv[i+1]);
229 debug(D_OPTIONS, "Configuration loaded from %s.", argv[i+1]);
234 else if(strcmp(argv[i], "-df") == 0 && (i+1) < argc) { config_set("global", "debug flags", argv[i+1]); debug_flags = strtoull(argv[i+1], NULL, 0); i++; }
235 else if(strcmp(argv[i], "-p") == 0 && (i+1) < argc) { config_set("global", "port", argv[i+1]); i++; }
236 else if(strcmp(argv[i], "-u") == 0 && (i+1) < argc) { config_set("global", "run as user", argv[i+1]); i++; }
237 else if(strcmp(argv[i], "-l") == 0 && (i+1) < argc) { config_set("global", "history", argv[i+1]); i++; }
238 else if(strcmp(argv[i], "-t") == 0 && (i+1) < argc) { config_set("global", "update every", argv[i+1]); i++; }
239 else if(strcmp(argv[i], "-ch") == 0 && (i+1) < argc) { config_set("global", "host access prefix", argv[i+1]); i++; }
240 else if(strcmp(argv[i], "-stacksize") == 0 && (i+1) < argc) { config_set("global", "pthread stack size", argv[i+1]); i++; }
241 else if(strcmp(argv[i], "-nodaemon") == 0 || strcmp(argv[i], "-nd") == 0) dont_fork = 1;
242 else if(strcmp(argv[i], "-pidfile") == 0 && (i+1) < argc) {
244 strncpy(pidfile, argv[i], FILENAME_MAX);
245 pidfile[FILENAME_MAX] = '\0';
247 else if(strcmp(argv[i], "--unittest") == 0) {
248 rrd_update_every = 1;
249 if(run_all_mockup_tests()) exit(1);
250 if(unit_test_storage()) exit(1);
251 fprintf(stderr, "\n\nALL TESTS PASSED\n\n");
255 fprintf(stderr, "Cannot understand option '%s'.\n", argv[i]);
256 fprintf(stderr, "\nUSAGE: %s [-d] [-l LINES_TO_SAVE] [-u UPDATE_TIMER] [-p LISTEN_PORT] [-df debug flags].\n\n", argv[0]);
257 fprintf(stderr, " -c CONFIG FILE the configuration file to load. Default: %s.\n", CONFIG_DIR "/" CONFIG_FILENAME);
258 fprintf(stderr, " -l LINES_TO_SAVE can be from 5 to %d lines in JSON data. Default: %d.\n", RRD_HISTORY_ENTRIES_MAX, RRD_DEFAULT_HISTORY_ENTRIES);
259 fprintf(stderr, " -t UPDATE_TIMER can be from 1 to %d seconds. Default: %d.\n", UPDATE_EVERY_MAX, UPDATE_EVERY);
260 fprintf(stderr, " -p LISTEN_PORT can be from 1 to %d. Default: %d.\n", 65535, LISTEN_PORT);
261 fprintf(stderr, " -u USERNAME can be any system username to run as. Default: none.\n");
262 fprintf(stderr, " -ch path to access host /proc and /sys when running in a container. Default: empty.\n");
263 fprintf(stderr, " -nd or -nodeamon to disable forking in the background. Default: unset.\n");
264 fprintf(stderr, " -df FLAGS debug options. Default: 0x%08llx.\n", debug_flags);
265 fprintf(stderr, " -stacksize BYTES to overwrite the pthread stack size.\n");
266 fprintf(stderr, " -pidfile FILENAME to save a pid while running.\n");
271 if(!config_loaded) load_config(NULL, 0);
273 // prepare configuration environment variables for the plugins
274 setenv("NETDATA_CONFIG_DIR" , config_get("global", "config directory" , CONFIG_DIR) , 1);
275 setenv("NETDATA_PLUGINS_DIR", config_get("global", "plugins directory" , PLUGINS_DIR), 1);
276 setenv("NETDATA_WEB_DIR" , config_get("global", "web files directory", WEB_DIR) , 1);
277 setenv("NETDATA_CACHE_DIR" , config_get("global", "cache directory" , CACHE_DIR) , 1);
278 setenv("NETDATA_LIB_DIR" , config_get("global", "lib directory" , VARLIB_DIR) , 1);
279 setenv("NETDATA_LOG_DIR" , config_get("global", "log directory" , LOG_DIR) , 1);
280 setenv("NETDATA_HOST_PREFIX", config_get("global", "host access prefix" , "") , 1);
281 setenv("HOME" , config_get("global", "home directory" , CACHE_DIR) , 1);
283 // avoid extended to stat(/etc/localtime)
284 // http://stackoverflow.com/questions/4554271/how-to-avoid-excessive-stat-etc-localtime-calls-in-strftime-on-linux
285 setenv("TZ", ":/etc/localtime", 0);
287 // cd to /tmp to avoid any plugins writing files at random places
288 if(chdir("/tmp")) error("netdata: ERROR: Cannot cd to /tmp");
290 char *input_log_file = NULL;
291 char *output_log_file = NULL;
292 char *error_log_file = NULL;
293 char *access_log_file = NULL;
298 // --------------------------------------------------------------------
300 sprintf(buffer, "0x%08llx", 0ULL);
301 char *flags = config_get("global", "debug flags", buffer);
302 setenv("NETDATA_DEBUG_FLAGS", flags, 1);
304 debug_flags = strtoull(flags, NULL, 0);
305 debug(D_OPTIONS, "Debug flags set to '0x%8llx'.", debug_flags);
307 if(debug_flags != 0) {
308 struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
309 if(setrlimit(RLIMIT_CORE, &rl) != 0)
310 info("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
311 prctl(PR_SET_DUMPABLE, 1, 0, 0, 0);
314 // --------------------------------------------------------------------
316 #ifdef MADV_MERGEABLE
317 enable_ksm = config_get_boolean("global", "memory deduplication (ksm)", enable_ksm);
319 #warning "Kernel memory deduplication (KSM) is not available"
322 // --------------------------------------------------------------------
325 global_host_prefix = config_get("global", "host access prefix", "");
326 setenv("NETDATA_HOST_PREFIX", global_host_prefix, 1);
328 // --------------------------------------------------------------------
330 output_log_file = config_get("global", "debug log", LOG_DIR "/debug.log");
331 if(strcmp(output_log_file, "syslog") == 0) {
332 output_log_syslog = 1;
333 output_log_file = NULL;
335 else if(strcmp(output_log_file, "none") == 0) {
336 output_log_syslog = 0;
337 output_log_file = NULL;
339 else output_log_syslog = 0;
341 // --------------------------------------------------------------------
343 error_log_file = config_get("global", "error log", LOG_DIR "/error.log");
344 if(strcmp(error_log_file, "syslog") == 0) {
345 error_log_syslog = 1;
346 error_log_file = NULL;
348 else if(strcmp(error_log_file, "none") == 0) {
349 error_log_syslog = 0;
350 error_log_file = NULL;
351 // optimization - do not even generate debug log entries
353 else error_log_syslog = 0;
355 error_log_throttle_period = config_get_number("global", "errors flood protection period", error_log_throttle_period);
356 setenv("NETDATA_ERRORS_THROTTLE_PERIOD", config_get("global", "errors flood protection period" , ""), 1);
358 error_log_errors_per_period = config_get_number("global", "errors to trigger flood protection", error_log_errors_per_period);
359 setenv("NETDATA_ERRORS_PER_PERIOD" , config_get("global", "errors to trigger flood protection", ""), 1);
361 // --------------------------------------------------------------------
363 access_log_file = config_get("global", "access log", LOG_DIR "/access.log");
364 if(strcmp(access_log_file, "syslog") == 0) {
365 access_log_syslog = 1;
366 access_log_file = NULL;
368 else if(strcmp(access_log_file, "none") == 0) {
369 access_log_syslog = 0;
370 access_log_file = NULL;
372 else access_log_syslog = 0;
374 // --------------------------------------------------------------------
376 rrd_memory_mode = rrd_memory_mode_id(config_get("global", "memory mode", rrd_memory_mode_name(rrd_memory_mode)));
378 // --------------------------------------------------------------------
380 if(gethostname(buffer, HOSTNAME_MAX) == -1)
381 error("WARNING: Cannot get machine hostname.");
382 hostname = config_get("global", "hostname", buffer);
383 debug(D_OPTIONS, "hostname set to '%s'", hostname);
385 // --------------------------------------------------------------------
387 rrd_default_history_entries = (int) config_get_number("global", "history", RRD_DEFAULT_HISTORY_ENTRIES);
388 if(rrd_default_history_entries < 5 || rrd_default_history_entries > RRD_HISTORY_ENTRIES_MAX) {
389 info("Invalid save lines %d given. Defaulting to %d.", rrd_default_history_entries, RRD_DEFAULT_HISTORY_ENTRIES);
390 rrd_default_history_entries = RRD_DEFAULT_HISTORY_ENTRIES;
393 debug(D_OPTIONS, "save lines set to %d.", rrd_default_history_entries);
396 // --------------------------------------------------------------------
398 rrd_update_every = (int) config_get_number("global", "update every", UPDATE_EVERY);
399 if(rrd_update_every < 1 || rrd_update_every > 600) {
400 info("Invalid update timer %d given. Defaulting to %d.", rrd_update_every, UPDATE_EVERY_MAX);
401 rrd_update_every = UPDATE_EVERY;
403 else debug(D_OPTIONS, "update timer set to %d.", rrd_update_every);
405 // let the plugins know the min update_every
408 snprintf(buf, 50, "%d", rrd_update_every);
409 setenv("NETDATA_UPDATE_EVERY", buf, 1);
412 // --------------------------------------------------------------------
414 i = pthread_attr_init(&attr);
416 fatal("pthread_attr_init() failed with code %d.", i);
418 i = pthread_attr_getstacksize(&attr, &stacksize);
420 fatal("pthread_attr_getstacksize() failed with code %d.", i);
422 debug(D_OPTIONS, "initial pthread stack size is %zu bytes", stacksize);
424 wanted_stacksize = config_get_number("global", "pthread stack size", stacksize);
426 // --------------------------------------------------------------------
428 for (i = 0; static_threads[i].name != NULL ; i++) {
429 struct netdata_static_thread *st = &static_threads[i];
431 if(st->config_name) st->enabled = config_get_boolean(st->config_section, st->config_name, st->enabled);
432 if(st->enabled && st->init_routine) st->init_routine();
435 // --------------------------------------------------------------------
437 // get the user we should run
438 // IMPORTANT: this is required before web_files_uid()
439 user = config_get("global", "run as user" , (getuid() == 0)?NETDATA_USER:"");
441 // IMPORTANT: these have to run once, while single threaded
442 web_files_uid(); // IMPORTANT: web_files_uid() before web_files_gid()
445 // --------------------------------------------------------------------
447 listen_backlog = (int) config_get_number("global", "http port listen backlog", LISTEN_BACKLOG);
449 listen_port = (int) config_get_number("global", "port", LISTEN_PORT);
450 if(listen_port < 1 || listen_port > 65535) {
451 info("Invalid listen port %d given. Defaulting to %d.", listen_port, LISTEN_PORT);
452 listen_port = LISTEN_PORT;
454 else debug(D_OPTIONS, "Listen port set to %d.", listen_port);
457 char *ipv = config_get("global", "ip version", "any");
458 if(!strcmp(ipv, "any") || !strcmp(ipv, "both") || !strcmp(ipv, "all")) ip = 0;
459 else if(!strcmp(ipv, "ipv4") || !strcmp(ipv, "IPV4") || !strcmp(ipv, "IPv4") || !strcmp(ipv, "4")) ip = 4;
460 else if(!strcmp(ipv, "ipv6") || !strcmp(ipv, "IPV6") || !strcmp(ipv, "IPv6") || !strcmp(ipv, "6")) ip = 6;
461 else info("Cannot understand ip version '%s'. Assuming 'any'.", ipv);
463 if(ip == 0 || ip == 6) listen_fd = create_listen_socket6(config_get("global", "bind socket to IP", "*"), listen_port, listen_backlog);
465 listen_fd = create_listen_socket4(config_get("global", "bind socket to IP", "*"), listen_port, listen_backlog);
466 if(listen_fd >= 0 && ip != 4) info("Managed to open an IPv4 socket on port %d.", listen_port);
469 if(listen_fd < 0) fatal("Cannot listen socket.");
472 // never become a problem
473 if(nice(20) == -1) error("Cannot lower my CPU priority.");
475 if(become_daemon(dont_fork, 0, user, input_log_file, output_log_file, error_log_file, access_log_file, &access_fd, &stdaccess) == -1) {
476 fatal("Cannot demonize myself.");
480 if(debug_flags != 0) {
481 struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
482 if(setrlimit(RLIMIT_CORE, &rl) != 0)
483 info("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
485 prctl(PR_SET_DUMPABLE, 1, 0, 0, 0);
488 if(output_log_syslog || error_log_syslog || access_log_syslog)
489 openlog("netdata", LOG_PID, LOG_DAEMON);
491 info("NetData started on pid %d", getpid());
495 for (i = 1 ; i < 65 ;i++) {
497 case SIGKILL: // not catchable
498 case SIGSTOP: // not catchable
508 signal(i, sig_handler);
513 // ------------------------------------------------------------------------
514 // get default pthread stack size
516 if(stacksize < wanted_stacksize) {
517 i = pthread_attr_setstacksize(&attr, wanted_stacksize);
519 fatal("pthread_attr_setstacksize() to %zu bytes, failed with code %d.", wanted_stacksize, i);
521 info("Successfully set pthread stacksize to %zu bytes", wanted_stacksize);
524 // ------------------------------------------------------------------------
527 for (i = 0; static_threads[i].name != NULL ; i++) {
528 struct netdata_static_thread *st = &static_threads[i];
531 st->thread = malloc(sizeof(pthread_t));
533 fatal("Cannot allocate pthread_t memory");
535 info("Starting thread %s.", st->name);
537 if(pthread_create(st->thread, &attr, st->start_routine, NULL))
538 error("failed to create new thread for %s.", st->name);
540 else if(pthread_detach(*st->thread))
541 error("Cannot request detach of newly created %s thread.", st->name);
543 else info("Not starting thread %s.", st->name);
546 // for future use - the main thread
548 if(netdata_exit != 0) {
551 if(netdata_exit > 5) {
552 netdata_cleanup_and_exit(0);