3 extern void *cgroups_main(void *ptr);
5 void netdata_cleanup_and_exit(int ret) {
8 error_log_limit_unlimited();
10 debug(D_EXIT, "Called: netdata_cleanup_and_exit()");
11 #ifdef NETDATA_INTERNAL_CHECKS
19 if(unlink(pidfile) != 0)
20 error("Cannot unlink pidfile '%s'.", pidfile);
23 info("NetData exiting. Bye bye...");
27 struct netdata_static_thread {
37 void (*init_routine) (void);
38 void *(*start_routine) (void *);
39 } static_threads[] = {
40 #ifdef INTERNAL_PLUGIN_NFACCT
41 // nfacct requires root access
42 // so, we build it as an external plugin with setuid to root
43 {"nfacct", "plugins", "nfacct", 1, NULL, NULL, nfacct_main},
46 {"tc", "plugins", "tc", 1, NULL, NULL, tc_main},
47 {"idlejitter", "plugins", "idlejitter", 1, NULL, NULL, cpuidlejitter_main},
49 {"proc", "plugins", "proc", 1, NULL, NULL, proc_main},
51 {"freebsd", "plugins", "freebsd", 1, NULL, NULL, freebsd_main},
52 #endif /* __FreeBSD__ */
53 {"cgroups", "plugins", "cgroups", 1, NULL, NULL, cgroups_main},
54 {"check", "plugins", "checks", 0, NULL, NULL, checks_main},
55 {"backends", NULL, NULL, 1, NULL, NULL, backends_main},
56 {"health", NULL, NULL, 1, NULL, NULL, health_main},
57 {"plugins.d", NULL, NULL, 1, NULL, NULL, pluginsd_main},
58 {"web", NULL, NULL, 1, NULL, NULL, socket_listen_main_multi_threaded},
59 {"web-single-threaded", NULL, NULL, 0, NULL, NULL, socket_listen_main_single_threaded},
60 {NULL, NULL, NULL, 0, NULL, NULL, NULL}
63 void web_server_threading_selection(void) {
64 int threaded = config_get_boolean("global", "multi threaded web server", 1);
67 for(i = 0; static_threads[i].name ; i++) {
68 if(static_threads[i].start_routine == socket_listen_main_multi_threaded)
69 static_threads[i].enabled = threaded?1:0;
71 if(static_threads[i].start_routine == socket_listen_main_single_threaded)
72 static_threads[i].enabled = threaded?0:1;
75 web_client_timeout = (int) config_get_number("global", "disconnect idle web clients after seconds", DEFAULT_DISCONNECT_IDLE_WEB_CLIENTS_AFTER_SECONDS);
77 web_donotrack_comply = config_get_boolean("global", "respect web browser do not track policy", web_donotrack_comply);
79 #ifdef NETDATA_WITH_ZLIB
80 web_enable_gzip = config_get_boolean("global", "enable web responses gzip compression", web_enable_gzip);
82 char *s = config_get("global", "web compression strategy", "default");
83 if(!strcmp(s, "default"))
84 web_gzip_strategy = Z_DEFAULT_STRATEGY;
85 else if(!strcmp(s, "filtered"))
86 web_gzip_strategy = Z_FILTERED;
87 else if(!strcmp(s, "huffman only"))
88 web_gzip_strategy = Z_HUFFMAN_ONLY;
89 else if(!strcmp(s, "rle"))
90 web_gzip_strategy = Z_RLE;
91 else if(!strcmp(s, "fixed"))
92 web_gzip_strategy = Z_FIXED;
94 error("Invalid compression strategy '%s'. Valid strategies are 'default', 'filtered', 'huffman only', 'rle' and 'fixed'. Proceeding with 'default'.", s);
95 web_gzip_strategy = Z_DEFAULT_STRATEGY;
98 web_gzip_level = (int)config_get_number("global", "web compression level", 3);
99 if(web_gzip_level < 1) {
100 error("Invalid compression level %d. Valid levels are 1 (fastest) to 9 (best ratio). Proceeding with level 1 (fastest compression).", web_gzip_level);
103 else if(web_gzip_level > 9) {
104 error("Invalid compression level %d. Valid levels are 1 (fastest) to 9 (best ratio). Proceeding with level 9 (best compression).", web_gzip_level);
107 #endif /* NETDATA_WITH_ZLIB */
111 int killpid(pid_t pid, int sig)
114 debug(D_EXIT, "Request to kill pid %d", pid);
117 if(kill(pid, 0) == -1) {
120 error("Request to kill pid %d, but it is not running.", pid);
124 error("Request to kill pid %d, but I do not have enough permissions.", pid);
128 error("Request to kill pid %d, but I received an error.", pid);
134 ret = kill(pid, sig);
138 error("Cannot kill pid %d, but it is not running.", pid);
142 error("Cannot kill pid %d, but I do not have enough permissions.", pid);
146 error("Cannot kill pid %d, but I received an error.", pid);
159 struct web_client *w;
160 for(w = web_clients; w ; w = w->next) {
161 debug(D_EXIT, "Stopping web client %s", w->client_ip);
162 pthread_cancel(w->thread);
163 pthread_join(w->thread, NULL);
167 for (i = 0; static_threads[i].name != NULL ; i++) {
168 if(static_threads[i].thread) {
169 debug(D_EXIT, "Stopping %s thread", static_threads[i].name);
170 pthread_cancel(*static_threads[i].thread);
171 pthread_join(*static_threads[i].thread, NULL);
172 static_threads[i].thread = NULL;
177 debug(D_EXIT, "Killing tc-qos-helper procees");
178 if(killpid(tc_child_pid, SIGTERM) != -1)
179 waitid(P_PID, (id_t) tc_child_pid, &info, WEXITED);
184 for(cd = pluginsd_root ; cd ; cd = cd->next) {
185 debug(D_EXIT, "Stopping %s plugin thread", cd->id);
186 pthread_cancel(cd->thread);
187 pthread_join(cd->thread, NULL);
189 if(cd->pid && !cd->obsolete) {
190 debug(D_EXIT, "killing %s plugin process", cd->id);
191 if(killpid(cd->pid, SIGTERM) != -1)
192 waitid(P_PID, (id_t) cd->pid, &info, WEXITED);
196 // if, for any reason there is any child exited
198 waitid(P_PID, 0, &info, WEXITED|WNOHANG);
200 debug(D_EXIT, "All threads/childs stopped.");
203 struct option_def options[] = {
204 // opt description arg name default value
205 {'c', "Load alternate configuration file", "config_file", CONFIG_DIR "/" CONFIG_FILENAME},
206 {'D', "Disable fork into background", NULL, NULL},
207 {'h', "Display help message", NULL, NULL},
208 {'P', "File to save a pid while running", "FILE", NULL},
209 {'i', "The IP address to listen to.", "address", "All addresses"},
210 {'k', "Check daemon configuration.", NULL, NULL},
211 {'p', "Port to listen. Can be from 1 to 65535.", "port_number", "19999"},
212 {'s', "Path to access host /proc and /sys when running in a container.", "PATH", NULL},
213 {'t', "The frequency in seconds, for data collection. \
214 Same as 'update every' config file option.", "seconds", "1"},
215 {'u', "System username to run as.", "username", "netdata"},
216 {'v', "Version of the program", NULL, NULL},
217 {'W', "vendor options.", "stacksize=N|unittest|debug_flags=N", NULL},
220 void help(int exitcode) {
227 int num_opts = sizeof(options) / sizeof(struct option_def);
231 // Compute maximum argument length
232 for( i = 0; i < num_opts; i++ ) {
233 if(options[i].arg_name) {
234 int len_arg = (int)strlen(options[i].arg_name);
235 if(len_arg > max_len_arg) max_len_arg = len_arg;
239 fprintf(stream, "SYNOPSIS: netdata [options]\n");
240 fprintf(stream, "\n");
241 fprintf(stream, "Options:\n");
243 // Output options description.
244 for( i = 0; i < num_opts; i++ ) {
245 fprintf(stream, " -%c %-*s %s", options[i].val, max_len_arg, options[i].arg_name ? options[i].arg_name : "", options[i].description);
246 if(options[i].default_value) {
247 fprintf(stream, " Default: %s\n", options[i].default_value);
249 fprintf(stream, "\n");
257 // TODO: Remove this function with the nix major release.
258 void remove_option(int opt_index, int *argc, char **argv) {
260 // remove the options.
263 for(i = opt_index; i < *argc; i++) {
267 } while(argv[i][0] != '-' && opt_index >= *argc);
270 static const char *verify_required_directory(const char *dir) {
272 fatal("Cannot cd to directory '%s'", dir);
274 DIR *d = opendir(dir);
276 fatal("Cannot examine the contents of directory '%s'", dir);
282 int main(int argc, char **argv)
284 char *hostname = "localhost";
285 int i, check_config = 0;
286 int config_loaded = 0;
288 size_t wanted_stacksize = 0, stacksize = 0;
291 // set the name for logging
292 program_name = "netdata";
294 // parse depercated options
295 // TODO: Remove this block with the next major release.
299 if(strcmp(argv[i], "-pidfile") == 0 && (i+1) < argc) {
300 strncpyz(pidfile, argv[i+1], FILENAME_MAX);
301 fprintf(stderr, "%s: deprecated option -- %s -- please use -P instead.\n", argv[0], argv[i]);
302 remove_option(i, &argc, argv);
304 else if(strcmp(argv[i], "-nodaemon") == 0 || strcmp(argv[i], "-nd") == 0) {
306 fprintf(stderr, "%s: deprecated option -- %s -- please use -D instead.\n ", argv[0], argv[i]);
307 remove_option(i, &argc, argv);
309 else if(strcmp(argv[i], "-ch") == 0 && (i+1) < argc) {
310 config_set("global", "host access prefix", argv[i+1]);
311 fprintf(stderr, "%s: deprecated option -- %s -- please use -s instead.\n", argv[0], argv[i]);
312 remove_option(i, &argc, argv);
314 else if(strcmp(argv[i], "-l") == 0 && (i+1) < argc) {
315 config_set("global", "history", argv[i+1]);
316 fprintf(stderr, "%s: deprecated option -- %s -- This option will be removed with V2.*.\n", argv[0], argv[i]);
317 remove_option(i, &argc, argv);
325 int num_opts = sizeof(options) / sizeof(struct option_def);
326 char optstring[(num_opts * 2) + 1];
329 for( i = 0; i < num_opts; i++ ) {
330 optstring[string_i] = options[i].val;
332 if(options[i].arg_name) {
333 optstring[string_i] = ':';
339 while( (opt = getopt(argc, argv, optstring)) != -1 ) {
342 if(load_config(optarg, 1) != 1) {
343 error("Cannot load configuration file %s.", optarg);
347 debug(D_OPTIONS, "Configuration loaded from %s.", optarg);
358 config_set("global", "bind to", optarg);
365 strncpy(pidfile, optarg, FILENAME_MAX);
366 pidfile[FILENAME_MAX] = '\0';
369 config_set("global", "default port", optarg);
372 config_set("global", "host access prefix", optarg);
375 config_set("global", "update every", optarg);
378 config_set("global", "run as user", optarg);
381 // TODO: Outsource version to makefile which can compute version from git.
382 printf("netdata %s\n", VERSION);
386 char* stacksize_string = "stacksize=";
387 char* debug_flags_string = "debug_flags=";
388 if(strcmp(optarg, "unittest") == 0) {
389 rrd_update_every = 1;
390 if(run_all_mockup_tests()) exit(1);
391 if(unit_test_storage()) exit(1);
392 fprintf(stderr, "\n\nALL TESTS PASSED\n\n");
394 } else if(strncmp(optarg, stacksize_string, strlen(stacksize_string)) == 0) {
395 optarg += strlen(stacksize_string);
396 config_set("global", "pthread stack size", optarg);
397 } else if(strncmp(optarg, debug_flags_string, strlen(debug_flags_string)) == 0) {
398 optarg += strlen(debug_flags_string);
399 config_set("global", "debug flags", optarg);
400 debug_flags = strtoull(optarg, NULL, 0);
412 load_config(NULL, 0);
415 char *pmax = config_get("global", "glibc malloc arena max for plugins", "1");
417 setenv("MALLOC_ARENA_MAX", pmax, 1);
419 #if defined(HAVE_C_MALLOPT)
420 int i = config_get_number("global", "glibc malloc arena max for netdata", 1);
422 mallopt(M_ARENA_MAX, 1);
425 char *config_dir = config_get("global", "config directory", CONFIG_DIR);
427 // prepare configuration environment variables for the plugins
428 setenv("NETDATA_CONFIG_DIR" , verify_required_directory(config_dir) , 1);
429 setenv("NETDATA_PLUGINS_DIR", verify_required_directory(config_get("global", "plugins directory" , PLUGINS_DIR)), 1);
430 setenv("NETDATA_WEB_DIR" , verify_required_directory(config_get("global", "web files directory", WEB_DIR)) , 1);
431 setenv("NETDATA_CACHE_DIR" , verify_required_directory(config_get("global", "cache directory" , CACHE_DIR)) , 1);
432 setenv("NETDATA_LIB_DIR" , verify_required_directory(config_get("global", "lib directory" , VARLIB_DIR)) , 1);
433 setenv("NETDATA_LOG_DIR" , verify_required_directory(config_get("global", "log directory" , LOG_DIR)) , 1);
435 setenv("NETDATA_HOST_PREFIX", config_get("global", "host access prefix" , "") , 1);
436 setenv("HOME" , config_get("global", "home directory" , CACHE_DIR) , 1);
438 // disable buffering for python plugins
439 setenv("PYTHONUNBUFFERED", "1", 1);
441 // avoid flood calls to stat(/etc/localtime)
442 // http://stackoverflow.com/questions/4554271/how-to-avoid-excessive-stat-etc-localtime-calls-in-strftime-on-linux
443 setenv("TZ", ":/etc/localtime", 0);
445 // work while we are cd into config_dir
446 // to allow the plugins refer to their config
447 // files using relative filenames
448 if(chdir(config_dir) == -1)
449 fatal("Cannot cd to '%s'", config_dir);
451 char path[1024 + 1], *p = getenv("PATH");
452 if(!p) p = "/bin:/usr/bin";
453 snprintfz(path, 1024, "%s:%s", p, "/sbin:/usr/sbin:/usr/local/bin:/usr/local/sbin");
454 setenv("PATH", config_get("plugins", "PATH environment variable", path), 1);
459 char *flags = config_get("global", "debug flags", "0x00000000");
460 setenv("NETDATA_DEBUG_FLAGS", flags, 1);
462 debug_flags = strtoull(flags, NULL, 0);
463 debug(D_OPTIONS, "Debug flags set to '0x%8llx'.", debug_flags);
465 if(debug_flags != 0) {
466 struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
467 if(setrlimit(RLIMIT_CORE, &rl) != 0)
468 error("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
471 prctl(PR_SET_DUMPABLE, 1, 0, 0, 0);
472 #endif /* __FreeBSD__ */
475 // --------------------------------------------------------------------
477 #ifdef MADV_MERGEABLE
478 enable_ksm = config_get_boolean("global", "memory deduplication (ksm)", enable_ksm);
480 #warning "Kernel memory deduplication (KSM) is not available"
483 // --------------------------------------------------------------------
485 global_host_prefix = config_get("global", "host access prefix", "");
486 setenv("NETDATA_HOST_PREFIX", global_host_prefix, 1);
490 get_system_pid_max();
492 // --------------------------------------------------------------------
494 stdout_filename = config_get("global", "debug log", LOG_DIR "/debug.log");
495 stderr_filename = config_get("global", "error log", LOG_DIR "/error.log");
496 stdaccess_filename = config_get("global", "access log", LOG_DIR "/access.log");
498 error_log_throttle_period_backup =
499 error_log_throttle_period = config_get_number("global", "errors flood protection period", error_log_throttle_period);
500 setenv("NETDATA_ERRORS_THROTTLE_PERIOD", config_get("global", "errors flood protection period" , ""), 1);
502 error_log_errors_per_period = (unsigned long)config_get_number("global", "errors to trigger flood protection", error_log_errors_per_period);
503 setenv("NETDATA_ERRORS_PER_PERIOD" , config_get("global", "errors to trigger flood protection", ""), 1);
506 stdout_filename = stderr_filename = stdaccess_filename = "system";
507 error_log_throttle_period = 0;
508 error_log_errors_per_period = 0;
510 error_log_limit_unlimited();
512 // --------------------------------------------------------------------
514 rrd_memory_mode = rrd_memory_mode_id(config_get("global", "memory mode", rrd_memory_mode_name(rrd_memory_mode)));
516 // --------------------------------------------------------------------
519 char hostnamebuf[HOSTNAME_MAX + 1];
520 if(gethostname(hostnamebuf, HOSTNAME_MAX) == -1)
521 error("WARNING: Cannot get machine hostname.");
522 hostname = config_get("global", "hostname", hostnamebuf);
523 debug(D_OPTIONS, "hostname set to '%s'", hostname);
524 setenv("NETDATA_HOSTNAME", hostname, 1);
527 // --------------------------------------------------------------------
529 rrd_default_history_entries = (int) config_get_number("global", "history", RRD_DEFAULT_HISTORY_ENTRIES);
530 if(rrd_default_history_entries < 5 || rrd_default_history_entries > RRD_HISTORY_ENTRIES_MAX) {
531 error("Invalid history entries %d given. Defaulting to %d.", rrd_default_history_entries, RRD_DEFAULT_HISTORY_ENTRIES);
532 rrd_default_history_entries = RRD_DEFAULT_HISTORY_ENTRIES;
535 debug(D_OPTIONS, "save lines set to %d.", rrd_default_history_entries);
538 // --------------------------------------------------------------------
540 rrd_update_every = (int) config_get_number("global", "update every", UPDATE_EVERY);
541 if(rrd_update_every < 1 || rrd_update_every > 600) {
542 error("Invalid data collection frequency (update every) %d given. Defaulting to %d.", rrd_update_every, UPDATE_EVERY_MAX);
543 rrd_update_every = UPDATE_EVERY;
545 else debug(D_OPTIONS, "update timer set to %d.", rrd_update_every);
547 // let the plugins know the min update_every
550 snprintfz(buf, 15, "%d", rrd_update_every);
551 setenv("NETDATA_UPDATE_EVERY", buf, 1);
554 // --------------------------------------------------------------------
556 // block signals while initializing threads.
557 // this causes the threads to block signals.
560 if(pthread_sigmask(SIG_BLOCK, &sigset, NULL) == -1)
561 error("Could not block signals for threads");
563 // Catch signals which we want to use
567 // ingore all signals while we run in a signal handler
568 sigfillset(&sa.sa_mask);
570 // INFO: If we add signals here we have to unblock them
571 // at popen.c when running a external plugin.
573 // Ignore SIGPIPE completely.
574 sa.sa_handler = SIG_IGN;
575 if(sigaction(SIGPIPE, &sa, NULL) == -1)
576 error("Failed to change signal handler for SIGPIPE");
578 sa.sa_handler = sig_handler_exit;
579 if(sigaction(SIGINT, &sa, NULL) == -1)
580 error("Failed to change signal handler for SIGINT");
582 sa.sa_handler = sig_handler_exit;
583 if(sigaction(SIGTERM, &sa, NULL) == -1)
584 error("Failed to change signal handler for SIGTERM");
586 sa.sa_handler = sig_handler_logrotate;
587 if(sigaction(SIGHUP, &sa, NULL) == -1)
588 error("Failed to change signal handler for SIGHUP");
590 // save database on SIGUSR1
591 sa.sa_handler = sig_handler_save;
592 if(sigaction(SIGUSR1, &sa, NULL) == -1)
593 error("Failed to change signal handler for SIGUSR1");
595 // reload health configuration on SIGUSR2
596 sa.sa_handler = sig_handler_reload_health;
597 if(sigaction(SIGUSR2, &sa, NULL) == -1)
598 error("Failed to change signal handler for SIGUSR2");
600 // --------------------------------------------------------------------
602 i = pthread_attr_init(&attr);
604 fatal("pthread_attr_init() failed with code %d.", i);
606 i = pthread_attr_getstacksize(&attr, &stacksize);
608 fatal("pthread_attr_getstacksize() failed with code %d.", i);
610 debug(D_OPTIONS, "initial pthread stack size is %zu bytes", stacksize);
612 wanted_stacksize = (size_t)config_get_number("global", "pthread stack size", (long)stacksize);
614 // --------------------------------------------------------------------
616 for (i = 0; static_threads[i].name != NULL ; i++) {
617 struct netdata_static_thread *st = &static_threads[i];
619 if(st->config_name) st->enabled = config_get_boolean(st->config_section, st->config_name, st->enabled);
620 if(st->enabled && st->init_routine) st->init_routine();
623 // --------------------------------------------------------------------
625 // get the user we should run
626 // IMPORTANT: this is required before web_files_uid()
627 user = config_get("global", "run as user" , (getuid() == 0)?NETDATA_USER:"");
629 // IMPORTANT: these have to run once, while single threaded
630 web_files_uid(); // IMPORTANT: web_files_uid() before web_files_gid()
633 // --------------------------------------------------------------------
636 create_listen_sockets();
639 // initialize the log files
640 open_all_log_files();
642 #ifdef NETDATA_INTERNAL_CHECKS
643 if(debug_flags != 0) {
644 struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
645 if(setrlimit(RLIMIT_CORE, &rl) != 0)
646 error("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
648 prctl(PR_SET_DUMPABLE, 1, 0, 0, 0);
649 #endif /* __FreeBSD__ */
651 #endif /* NETDATA_INTERNAL_CHECKS */
653 // fork, switch user, create pid file, set process priority
654 if(become_daemon(dont_fork, user) == -1)
655 fatal("Cannot daemonize myself.");
657 info("NetData started on pid %d", getpid());
660 // ------------------------------------------------------------------------
661 // get default pthread stack size
663 if(stacksize < wanted_stacksize) {
664 i = pthread_attr_setstacksize(&attr, wanted_stacksize);
666 fatal("pthread_attr_setstacksize() to %zu bytes, failed with code %d.", wanted_stacksize, i);
668 debug(D_SYSTEM, "Successfully set pthread stacksize to %zu bytes", wanted_stacksize);
671 // ------------------------------------------------------------------------
672 // initialize rrd host
674 rrdhost_init(hostname);
676 // ------------------------------------------------------------------------
677 // initialize the registry
681 // ------------------------------------------------------------------------
682 // initialize health monitoring
689 // ------------------------------------------------------------------------
690 // enable log flood protection
692 error_log_limit_reset();
694 // ------------------------------------------------------------------------
697 web_server_threading_selection();
699 for (i = 0; static_threads[i].name != NULL ; i++) {
700 struct netdata_static_thread *st = &static_threads[i];
703 st->thread = mallocz(sizeof(pthread_t));
705 debug(D_SYSTEM, "Starting thread %s.", st->name);
707 if(pthread_create(st->thread, &attr, st->start_routine, NULL))
708 error("failed to create new thread for %s.", st->name);
710 else if(pthread_detach(*st->thread))
711 error("Cannot request detach of newly created %s thread.", st->name);
713 else debug(D_SYSTEM, "Not starting thread %s.", st->name);
716 // ------------------------------------------------------------------------
717 // block signals while initializing threads.
721 if(pthread_sigmask(SIG_UNBLOCK, &sigset, NULL) == -1) {
722 error("Could not unblock signals for threads");
725 // Handle flags set in the signal handler.
729 debug(D_EXIT, "Exit main loop of netdata.");
730 netdata_cleanup_and_exit(0);