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},
48 {"proc", "plugins", "proc", 1, NULL, NULL, proc_main},
49 {"cgroups", "plugins", "cgroups", 1, NULL, NULL, cgroups_main},
50 {"check", "plugins", "checks", 0, NULL, NULL, checks_main},
51 {"backends", NULL, NULL, 1, NULL, NULL, backends_main},
52 {"health", NULL, NULL, 1, NULL, NULL, health_main},
53 {"plugins.d", NULL, NULL, 1, NULL, NULL, pluginsd_main},
54 {"web", NULL, NULL, 1, NULL, NULL, socket_listen_main_multi_threaded},
55 {"web-single-threaded", NULL, NULL, 0, NULL, NULL, socket_listen_main_single_threaded},
56 {NULL, NULL, NULL, 0, NULL, NULL, NULL}
59 void web_server_threading_selection(void) {
60 int threaded = config_get_boolean("global", "multi threaded web server", 1);
63 for(i = 0; static_threads[i].name ; i++) {
64 if(static_threads[i].start_routine == socket_listen_main_multi_threaded)
65 static_threads[i].enabled = threaded?1:0;
67 if(static_threads[i].start_routine == socket_listen_main_single_threaded)
68 static_threads[i].enabled = threaded?0:1;
71 web_client_timeout = (int) config_get_number("global", "disconnect idle web clients after seconds", DEFAULT_DISCONNECT_IDLE_WEB_CLIENTS_AFTER_SECONDS);
73 web_donotrack_comply = config_get_boolean("global", "respect web browser do not track policy", web_donotrack_comply);
75 #ifdef NETDATA_WITH_ZLIB
76 web_enable_gzip = config_get_boolean("global", "enable web responses gzip compression", web_enable_gzip);
78 char *s = config_get("global", "web compression strategy", "default");
79 if(!strcmp(s, "default"))
80 web_gzip_strategy = Z_DEFAULT_STRATEGY;
81 else if(!strcmp(s, "filtered"))
82 web_gzip_strategy = Z_FILTERED;
83 else if(!strcmp(s, "huffman only"))
84 web_gzip_strategy = Z_HUFFMAN_ONLY;
85 else if(!strcmp(s, "rle"))
86 web_gzip_strategy = Z_RLE;
87 else if(!strcmp(s, "fixed"))
88 web_gzip_strategy = Z_FIXED;
90 error("Invalid compression strategy '%s'. Valid strategies are 'default', 'filtered', 'huffman only', 'rle' and 'fixed'. Proceeding with 'default'.", s);
91 web_gzip_strategy = Z_DEFAULT_STRATEGY;
94 web_gzip_level = (int)config_get_number("global", "web compression level", 3);
95 if(web_gzip_level < 1) {
96 error("Invalid compression level %d. Valid levels are 1 (fastest) to 9 (best ratio). Proceeding with level 1 (fastest compression).", web_gzip_level);
99 else if(web_gzip_level > 9) {
100 error("Invalid compression level %d. Valid levels are 1 (fastest) to 9 (best ratio). Proceeding with level 9 (best compression).", web_gzip_level);
103 #endif /* NETDATA_WITH_ZLIB */
107 int killpid(pid_t pid, int sig)
110 debug(D_EXIT, "Request to kill pid %d", pid);
113 if(kill(pid, 0) == -1) {
116 error("Request to kill pid %d, but it is not running.", pid);
120 error("Request to kill pid %d, but I do not have enough permissions.", pid);
124 error("Request to kill pid %d, but I received an error.", pid);
130 ret = kill(pid, sig);
134 error("Cannot kill pid %d, but it is not running.", pid);
138 error("Cannot kill pid %d, but I do not have enough permissions.", pid);
142 error("Cannot kill pid %d, but I received an error.", pid);
155 struct web_client *w;
156 for(w = web_clients; w ; w = w->next) {
157 debug(D_EXIT, "Stopping web client %s", w->client_ip);
158 pthread_cancel(w->thread);
159 pthread_join(w->thread, NULL);
163 for (i = 0; static_threads[i].name != NULL ; i++) {
164 if(static_threads[i].thread) {
165 debug(D_EXIT, "Stopping %s thread", static_threads[i].name);
166 pthread_cancel(*static_threads[i].thread);
167 pthread_join(*static_threads[i].thread, NULL);
168 static_threads[i].thread = NULL;
173 debug(D_EXIT, "Killing tc-qos-helper procees");
174 if(killpid(tc_child_pid, SIGTERM) != -1)
175 waitid(P_PID, (id_t) tc_child_pid, &info, WEXITED);
180 for(cd = pluginsd_root ; cd ; cd = cd->next) {
181 debug(D_EXIT, "Stopping %s plugin thread", cd->id);
182 pthread_cancel(cd->thread);
183 pthread_join(cd->thread, NULL);
185 if(cd->pid && !cd->obsolete) {
186 debug(D_EXIT, "killing %s plugin process", cd->id);
187 if(killpid(cd->pid, SIGTERM) != -1)
188 waitid(P_PID, (id_t) cd->pid, &info, WEXITED);
192 // if, for any reason there is any child exited
194 waitid(P_PID, 0, &info, WEXITED|WNOHANG);
196 debug(D_EXIT, "All threads/childs stopped.");
199 struct option_def options[] = {
200 // opt description arg name default value
201 {'c', "Load alternate configuration file", "config_file", CONFIG_DIR "/" CONFIG_FILENAME},
202 {'D', "Disable fork into background", NULL, NULL},
203 {'h', "Display help message", NULL, NULL},
204 {'P', "File to save a pid while running", "FILE", NULL},
205 {'i', "The IP address to listen to.", "address", "All addresses"},
206 {'k', "Check daemon configuration.", NULL, NULL},
207 {'p', "Port to listen. Can be from 1 to 65535.", "port_number", "19999"},
208 {'s', "Path to access host /proc and /sys when running in a container.", "PATH", NULL},
209 {'t', "The frequency in seconds, for data collection. \
210 Same as 'update every' config file option.", "seconds", "1"},
211 {'u', "System username to run as.", "username", "netdata"},
212 {'v', "Version of the program", NULL, NULL},
213 {'W', "vendor options.", "stacksize=N|unittest|debug_flags=N", NULL},
216 void help(int exitcode) {
223 int num_opts = sizeof(options) / sizeof(struct option_def);
227 // Compute maximum argument length
228 for( i = 0; i < num_opts; i++ ) {
229 if(options[i].arg_name) {
230 int len_arg = (int)strlen(options[i].arg_name);
231 if(len_arg > max_len_arg) max_len_arg = len_arg;
235 fprintf(stream, "SYNOPSIS: netdata [options]\n");
236 fprintf(stream, "\n");
237 fprintf(stream, "Options:\n");
239 // Output options description.
240 for( i = 0; i < num_opts; i++ ) {
241 fprintf(stream, " -%c %-*s %s", options[i].val, max_len_arg, options[i].arg_name ? options[i].arg_name : "", options[i].description);
242 if(options[i].default_value) {
243 fprintf(stream, " Default: %s\n", options[i].default_value);
245 fprintf(stream, "\n");
253 // TODO: Remove this function with the nix major release.
254 void remove_option(int opt_index, int *argc, char **argv) {
256 // remove the options.
259 for(i = opt_index; i < *argc; i++) {
263 } while(argv[i][0] != '-' && opt_index >= *argc);
266 static const char *verify_required_directory(const char *dir) {
268 fatal("Cannot cd to directory '%s'", dir);
270 DIR *d = opendir(dir);
272 fatal("Cannot examine the contents of directory '%s'", dir);
278 int main(int argc, char **argv)
280 char *hostname = "localhost";
281 int i, check_config = 0;
282 int config_loaded = 0;
284 size_t wanted_stacksize = 0, stacksize = 0;
287 // set the name for logging
288 program_name = "netdata";
290 // parse depercated options
291 // TODO: Remove this block with the next major release.
295 if(strcmp(argv[i], "-pidfile") == 0 && (i+1) < argc) {
296 strncpyz(pidfile, argv[i+1], FILENAME_MAX);
297 fprintf(stderr, "%s: deprecated option -- %s -- please use -P instead.\n", argv[0], argv[i]);
298 remove_option(i, &argc, argv);
300 else if(strcmp(argv[i], "-nodaemon") == 0 || strcmp(argv[i], "-nd") == 0) {
302 fprintf(stderr, "%s: deprecated option -- %s -- please use -D instead.\n ", argv[0], argv[i]);
303 remove_option(i, &argc, argv);
305 else if(strcmp(argv[i], "-ch") == 0 && (i+1) < argc) {
306 config_set("global", "host access prefix", argv[i+1]);
307 fprintf(stderr, "%s: deprecated option -- %s -- please use -s instead.\n", argv[0], argv[i]);
308 remove_option(i, &argc, argv);
310 else if(strcmp(argv[i], "-l") == 0 && (i+1) < argc) {
311 config_set("global", "history", argv[i+1]);
312 fprintf(stderr, "%s: deprecated option -- %s -- This option will be removed with V2.*.\n", argv[0], argv[i]);
313 remove_option(i, &argc, argv);
321 int num_opts = sizeof(options) / sizeof(struct option_def);
322 char optstring[(num_opts * 2) + 1];
325 for( i = 0; i < num_opts; i++ ) {
326 optstring[string_i] = options[i].val;
328 if(options[i].arg_name) {
329 optstring[string_i] = ':';
335 while( (opt = getopt(argc, argv, optstring)) != -1 ) {
338 if(load_config(optarg, 1) != 1) {
339 error("Cannot load configuration file %s.", optarg);
343 debug(D_OPTIONS, "Configuration loaded from %s.", optarg);
354 config_set("global", "bind to", optarg);
361 strncpy(pidfile, optarg, FILENAME_MAX);
362 pidfile[FILENAME_MAX] = '\0';
365 config_set("global", "default port", optarg);
368 config_set("global", "host access prefix", optarg);
371 config_set("global", "update every", optarg);
374 config_set("global", "run as user", optarg);
377 // TODO: Outsource version to makefile which can compute version from git.
378 printf("netdata %s\n", VERSION);
382 char* stacksize_string = "stacksize=";
383 char* debug_flags_string = "debug_flags=";
384 if(strcmp(optarg, "unittest") == 0) {
385 rrd_update_every = 1;
386 if(run_all_mockup_tests()) exit(1);
387 if(unit_test_storage()) exit(1);
388 fprintf(stderr, "\n\nALL TESTS PASSED\n\n");
390 } else if(strncmp(optarg, stacksize_string, strlen(stacksize_string)) == 0) {
391 optarg += strlen(stacksize_string);
392 config_set("global", "pthread stack size", optarg);
393 } else if(strncmp(optarg, debug_flags_string, strlen(debug_flags_string)) == 0) {
394 optarg += strlen(debug_flags_string);
395 config_set("global", "debug flags", optarg);
396 debug_flags = strtoull(optarg, NULL, 0);
408 load_config(NULL, 0);
411 char *pmax = config_get("global", "glibc malloc arena max for plugins", "1");
413 setenv("MALLOC_ARENA_MAX", pmax, 1);
415 #if defined(HAVE_C_MALLOPT)
416 int i = config_get_number("global", "glibc malloc arena max for netdata", 1);
418 mallopt(M_ARENA_MAX, 1);
421 char *config_dir = config_get("global", "config directory", CONFIG_DIR);
423 // prepare configuration environment variables for the plugins
424 setenv("NETDATA_CONFIG_DIR" , verify_required_directory(config_dir) , 1);
425 setenv("NETDATA_PLUGINS_DIR", verify_required_directory(config_get("global", "plugins directory" , PLUGINS_DIR)), 1);
426 setenv("NETDATA_WEB_DIR" , verify_required_directory(config_get("global", "web files directory", WEB_DIR)) , 1);
427 setenv("NETDATA_CACHE_DIR" , verify_required_directory(config_get("global", "cache directory" , CACHE_DIR)) , 1);
428 setenv("NETDATA_LIB_DIR" , verify_required_directory(config_get("global", "lib directory" , VARLIB_DIR)) , 1);
429 setenv("NETDATA_LOG_DIR" , verify_required_directory(config_get("global", "log directory" , LOG_DIR)) , 1);
431 setenv("NETDATA_HOST_PREFIX", config_get("global", "host access prefix" , "") , 1);
432 setenv("HOME" , config_get("global", "home directory" , CACHE_DIR) , 1);
434 // disable buffering for python plugins
435 setenv("PYTHONUNBUFFERED", "1", 1);
437 // avoid flood calls to stat(/etc/localtime)
438 // http://stackoverflow.com/questions/4554271/how-to-avoid-excessive-stat-etc-localtime-calls-in-strftime-on-linux
439 setenv("TZ", ":/etc/localtime", 0);
441 // work while we are cd into config_dir
442 // to allow the plugins refer to their config
443 // files using relative filenames
444 if(chdir(config_dir) == -1)
445 fatal("Cannot cd to '%s'", config_dir);
447 char path[1024 + 1], *p = getenv("PATH");
448 if(!p) p = "/bin:/usr/bin";
449 snprintfz(path, 1024, "%s:%s", p, "/sbin:/usr/sbin:/usr/local/bin:/usr/local/sbin");
450 setenv("PATH", config_get("plugins", "PATH environment variable", path), 1);
455 char *flags = config_get("global", "debug flags", "0x00000000");
456 setenv("NETDATA_DEBUG_FLAGS", flags, 1);
458 debug_flags = strtoull(flags, NULL, 0);
459 debug(D_OPTIONS, "Debug flags set to '0x%8llx'.", debug_flags);
462 if(debug_flags != 0) {
463 struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
464 if(setrlimit(RLIMIT_CORE, &rl) != 0)
465 error("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
467 prctl(PR_SET_DUMPABLE, 1, 0, 0, 0);
469 #endif /* __FreeBSD__ */
471 // --------------------------------------------------------------------
473 #ifdef MADV_MERGEABLE
474 enable_ksm = config_get_boolean("global", "memory deduplication (ksm)", enable_ksm);
476 #warning "Kernel memory deduplication (KSM) is not available"
479 // --------------------------------------------------------------------
481 global_host_prefix = config_get("global", "host access prefix", "");
482 setenv("NETDATA_HOST_PREFIX", global_host_prefix, 1);
486 get_system_pid_max();
488 // --------------------------------------------------------------------
490 stdout_filename = config_get("global", "debug log", LOG_DIR "/debug.log");
491 stderr_filename = config_get("global", "error log", LOG_DIR "/error.log");
492 stdaccess_filename = config_get("global", "access log", LOG_DIR "/access.log");
494 error_log_throttle_period_backup =
495 error_log_throttle_period = config_get_number("global", "errors flood protection period", error_log_throttle_period);
496 setenv("NETDATA_ERRORS_THROTTLE_PERIOD", config_get("global", "errors flood protection period" , ""), 1);
498 error_log_errors_per_period = (unsigned long)config_get_number("global", "errors to trigger flood protection", error_log_errors_per_period);
499 setenv("NETDATA_ERRORS_PER_PERIOD" , config_get("global", "errors to trigger flood protection", ""), 1);
502 stdout_filename = stderr_filename = stdaccess_filename = "system";
503 error_log_throttle_period = 0;
504 error_log_errors_per_period = 0;
506 error_log_limit_unlimited();
508 // --------------------------------------------------------------------
510 rrd_memory_mode = rrd_memory_mode_id(config_get("global", "memory mode", rrd_memory_mode_name(rrd_memory_mode)));
512 // --------------------------------------------------------------------
515 char hostnamebuf[HOSTNAME_MAX + 1];
516 if(gethostname(hostnamebuf, HOSTNAME_MAX) == -1)
517 error("WARNING: Cannot get machine hostname.");
518 hostname = config_get("global", "hostname", hostnamebuf);
519 debug(D_OPTIONS, "hostname set to '%s'", hostname);
520 setenv("NETDATA_HOSTNAME", hostname, 1);
523 // --------------------------------------------------------------------
525 rrd_default_history_entries = (int) config_get_number("global", "history", RRD_DEFAULT_HISTORY_ENTRIES);
526 if(rrd_default_history_entries < 5 || rrd_default_history_entries > RRD_HISTORY_ENTRIES_MAX) {
527 error("Invalid history entries %d given. Defaulting to %d.", rrd_default_history_entries, RRD_DEFAULT_HISTORY_ENTRIES);
528 rrd_default_history_entries = RRD_DEFAULT_HISTORY_ENTRIES;
531 debug(D_OPTIONS, "save lines set to %d.", rrd_default_history_entries);
534 // --------------------------------------------------------------------
536 rrd_update_every = (int) config_get_number("global", "update every", UPDATE_EVERY);
537 if(rrd_update_every < 1 || rrd_update_every > 600) {
538 error("Invalid data collection frequency (update every) %d given. Defaulting to %d.", rrd_update_every, UPDATE_EVERY_MAX);
539 rrd_update_every = UPDATE_EVERY;
541 else debug(D_OPTIONS, "update timer set to %d.", rrd_update_every);
543 // let the plugins know the min update_every
546 snprintfz(buf, 15, "%d", rrd_update_every);
547 setenv("NETDATA_UPDATE_EVERY", buf, 1);
550 // --------------------------------------------------------------------
552 // block signals while initializing threads.
553 // this causes the threads to block signals.
556 if(pthread_sigmask(SIG_BLOCK, &sigset, NULL) == -1)
557 error("Could not block signals for threads");
559 // Catch signals which we want to use
563 // ingore all signals while we run in a signal handler
564 sigfillset(&sa.sa_mask);
566 // INFO: If we add signals here we have to unblock them
567 // at popen.c when running a external plugin.
569 // Ignore SIGPIPE completely.
570 sa.sa_handler = SIG_IGN;
571 if(sigaction(SIGPIPE, &sa, NULL) == -1)
572 error("Failed to change signal handler for SIGPIPE");
574 sa.sa_handler = sig_handler_exit;
575 if(sigaction(SIGINT, &sa, NULL) == -1)
576 error("Failed to change signal handler for SIGINT");
578 sa.sa_handler = sig_handler_exit;
579 if(sigaction(SIGTERM, &sa, NULL) == -1)
580 error("Failed to change signal handler for SIGTERM");
582 sa.sa_handler = sig_handler_logrotate;
583 if(sigaction(SIGHUP, &sa, NULL) == -1)
584 error("Failed to change signal handler for SIGHUP");
586 // save database on SIGUSR1
587 sa.sa_handler = sig_handler_save;
588 if(sigaction(SIGUSR1, &sa, NULL) == -1)
589 error("Failed to change signal handler for SIGUSR1");
591 // reload health configuration on SIGUSR2
592 sa.sa_handler = sig_handler_reload_health;
593 if(sigaction(SIGUSR2, &sa, NULL) == -1)
594 error("Failed to change signal handler for SIGUSR2");
596 // --------------------------------------------------------------------
598 i = pthread_attr_init(&attr);
600 fatal("pthread_attr_init() failed with code %d.", i);
602 i = pthread_attr_getstacksize(&attr, &stacksize);
604 fatal("pthread_attr_getstacksize() failed with code %d.", i);
606 debug(D_OPTIONS, "initial pthread stack size is %zu bytes", stacksize);
608 wanted_stacksize = (size_t)config_get_number("global", "pthread stack size", (long)stacksize);
610 // --------------------------------------------------------------------
612 for (i = 0; static_threads[i].name != NULL ; i++) {
613 struct netdata_static_thread *st = &static_threads[i];
615 if(st->config_name) st->enabled = config_get_boolean(st->config_section, st->config_name, st->enabled);
616 if(st->enabled && st->init_routine) st->init_routine();
619 // --------------------------------------------------------------------
621 // get the user we should run
622 // IMPORTANT: this is required before web_files_uid()
623 user = config_get("global", "run as user" , (getuid() == 0)?NETDATA_USER:"");
625 // IMPORTANT: these have to run once, while single threaded
626 web_files_uid(); // IMPORTANT: web_files_uid() before web_files_gid()
629 // --------------------------------------------------------------------
632 create_listen_sockets();
635 // initialize the log files
636 open_all_log_files();
639 #ifdef NETDATA_INTERNAL_CHECKS
640 if(debug_flags != 0) {
641 struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
642 if(setrlimit(RLIMIT_CORE, &rl) != 0)
643 error("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
644 prctl(PR_SET_DUMPABLE, 1, 0, 0, 0);
646 #endif /* NETDATA_INTERNAL_CHECKS */
647 #endif /* __FreeBSD__ */
649 // fork, switch user, create pid file, set process priority
650 if(become_daemon(dont_fork, user) == -1)
651 fatal("Cannot daemonize myself.");
653 info("NetData started on pid %d", getpid());
656 // ------------------------------------------------------------------------
657 // get default pthread stack size
659 if(stacksize < wanted_stacksize) {
660 i = pthread_attr_setstacksize(&attr, wanted_stacksize);
662 fatal("pthread_attr_setstacksize() to %zu bytes, failed with code %d.", wanted_stacksize, i);
664 debug(D_SYSTEM, "Successfully set pthread stacksize to %zu bytes", wanted_stacksize);
667 // ------------------------------------------------------------------------
668 // initialize rrd host
670 rrdhost_init(hostname);
672 // ------------------------------------------------------------------------
673 // initialize the registry
677 // ------------------------------------------------------------------------
678 // initialize health monitoring
685 // ------------------------------------------------------------------------
686 // enable log flood protection
688 error_log_limit_reset();
690 // ------------------------------------------------------------------------
693 web_server_threading_selection();
695 for (i = 0; static_threads[i].name != NULL ; i++) {
696 struct netdata_static_thread *st = &static_threads[i];
699 st->thread = mallocz(sizeof(pthread_t));
701 debug(D_SYSTEM, "Starting thread %s.", st->name);
703 if(pthread_create(st->thread, &attr, st->start_routine, NULL))
704 error("failed to create new thread for %s.", st->name);
706 else if(pthread_detach(*st->thread))
707 error("Cannot request detach of newly created %s thread.", st->name);
709 else debug(D_SYSTEM, "Not starting thread %s.", st->name);
712 // ------------------------------------------------------------------------
713 // block signals while initializing threads.
717 if(pthread_sigmask(SIG_UNBLOCK, &sigset, NULL) == -1) {
718 error("Could not unblock signals for threads");
721 // Handle flags set in the signal handler.
725 debug(D_EXIT, "Exit main loop of netdata.");
726 netdata_cleanup_and_exit(0);