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"
41 extern void *cgroups_main(void *ptr);
43 volatile sig_atomic_t netdata_exit = 0;
45 void netdata_cleanup_and_exit(int ret)
51 // let it log a few more error messages
52 error_log_limit_reset();
55 if(ftruncate(pidfd, 0) != 0)
56 error("Cannot truncate pidfile '%s'.", pidfile);
63 if(unlink(pidfile) != 0)
64 error("Cannot unlink pidfile '%s'.", pidfile);
67 info("NetData exiting. Bye bye...");
71 struct netdata_static_thread {
81 void (*init_routine) (void);
82 void *(*start_routine) (void *);
85 struct netdata_static_thread static_threads[] = {
86 {"tc", "plugins", "tc", 1, NULL, NULL, tc_main},
87 {"idlejitter", "plugins", "idlejitter", 1, NULL, NULL, cpuidlejitter_main},
88 {"proc", "plugins", "proc", 1, NULL, NULL, proc_main},
89 {"cgroups", "plugins", "cgroups", 1, NULL, NULL, cgroups_main},
91 #ifdef INTERNAL_PLUGIN_NFACCT
92 // nfacct requires root access
93 // so, we build it as an external plugin with setuid to root
94 {"nfacct", "plugins", "nfacct", 1, NULL, NULL, nfacct_main},
97 {"plugins.d", NULL, NULL, 1, NULL, NULL, pluginsd_main},
98 {"check", "plugins", "checks", 0, NULL, NULL, checks_main},
99 {"web", NULL, NULL, 1, NULL, NULL, socket_listen_main},
100 {NULL, NULL, NULL, 0, NULL, NULL, NULL}
103 int killpid(pid_t pid, int sig)
106 debug(D_EXIT, "Request to kill pid %d", pid);
109 if(kill(pid, 0) == -1) {
112 error("Request to kill pid %d, but it is not running.", pid);
116 error("Request to kill pid %d, but I do not have enough permissions.", pid);
120 error("Request to kill pid %d, but I received an error.", pid);
126 ret = kill(pid, sig);
130 error("Cannot kill pid %d, but it is not running.", pid);
134 error("Cannot kill pid %d, but I do not have enough permissions.", pid);
138 error("Cannot kill pid %d, but I received an error.", pid);
151 struct web_client *w;
152 for(w = web_clients; w ; w = w->next) {
153 debug(D_EXIT, "Stopping web client %s", w->client_ip);
154 pthread_cancel(w->thread);
155 pthread_join(w->thread, NULL);
159 for (i = 0; static_threads[i].name != NULL ; i++) {
160 if(static_threads[i].thread) {
161 debug(D_EXIT, "Stopping %s thread", static_threads[i].name);
162 pthread_cancel(*static_threads[i].thread);
163 pthread_join(*static_threads[i].thread, NULL);
164 static_threads[i].thread = NULL;
169 info("Killing tc-qos-helper procees");
170 if(killpid(tc_child_pid, SIGTERM) != -1)
171 waitid(P_PID, (id_t) tc_child_pid, &info, WEXITED);
176 for(cd = pluginsd_root ; cd ; cd = cd->next) {
177 debug(D_EXIT, "Stopping %s plugin thread", cd->id);
178 pthread_cancel(cd->thread);
179 pthread_join(cd->thread, NULL);
181 if(cd->pid && !cd->obsolete) {
182 debug(D_EXIT, "killing %s plugin process", cd->id);
183 if(killpid(cd->pid, SIGTERM) != -1)
184 waitid(P_PID, (id_t) cd->pid, &info, WEXITED);
188 // if, for any reason there is any child exited
190 waitid(P_PID, 0, &info, WEXITED|WNOHANG);
192 debug(D_EXIT, "All threads/childs stopped.");
196 int main(int argc, char **argv)
199 int config_loaded = 0;
201 size_t wanted_stacksize = 0, stacksize = 0;
204 // global initialization
207 // set the name for logging
208 program_name = "netdata";
210 // parse the arguments
211 for(i = 1; i < argc ; i++) {
212 if(strcmp(argv[i], "-c") == 0 && (i+1) < argc) {
213 if(load_config(argv[i+1], 1) != 1) {
214 error("Cannot load configuration file %s.", argv[i+1]);
218 debug(D_OPTIONS, "Configuration loaded from %s.", argv[i+1]);
223 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++; }
224 else if(strcmp(argv[i], "-p") == 0 && (i+1) < argc) { config_set("global", "port", argv[i+1]); i++; }
225 else if(strcmp(argv[i], "-u") == 0 && (i+1) < argc) { config_set("global", "run as user", argv[i+1]); i++; }
226 else if(strcmp(argv[i], "-l") == 0 && (i+1) < argc) { config_set("global", "history", argv[i+1]); i++; }
227 else if(strcmp(argv[i], "-t") == 0 && (i+1) < argc) { config_set("global", "update every", argv[i+1]); i++; }
228 else if(strcmp(argv[i], "-ch") == 0 && (i+1) < argc) { config_set("global", "host access prefix", argv[i+1]); i++; }
229 else if(strcmp(argv[i], "-stacksize") == 0 && (i+1) < argc) { config_set("global", "pthread stack size", argv[i+1]); i++; }
230 else if(strcmp(argv[i], "-nodaemon") == 0 || strcmp(argv[i], "-nd") == 0) dont_fork = 1;
231 else if(strcmp(argv[i], "-pidfile") == 0 && (i+1) < argc) {
233 strncpy(pidfile, argv[i], FILENAME_MAX);
234 pidfile[FILENAME_MAX] = '\0';
236 else if(strcmp(argv[i], "--unittest") == 0) {
237 rrd_update_every = 1;
238 if(run_all_mockup_tests()) exit(1);
239 if(unit_test_storage()) exit(1);
240 fprintf(stderr, "\n\nALL TESTS PASSED\n\n");
244 fprintf(stderr, "Cannot understand option '%s'.\n", argv[i]);
245 fprintf(stderr, "\nUSAGE: %s [-d] [-l LINES_TO_SAVE] [-u UPDATE_TIMER] [-p LISTEN_PORT] [-df debug flags].\n\n", argv[0]);
246 fprintf(stderr, " -c CONFIG FILE the configuration file to load. Default: %s.\n", CONFIG_DIR "/" CONFIG_FILENAME);
247 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);
248 fprintf(stderr, " -t UPDATE_TIMER can be from 1 to %d seconds. Default: %d.\n", UPDATE_EVERY_MAX, UPDATE_EVERY);
249 fprintf(stderr, " -p LISTEN_PORT can be from 1 to %d. Default: %d.\n", 65535, LISTEN_PORT);
250 fprintf(stderr, " -u USERNAME can be any system username to run as. Default: none.\n");
251 fprintf(stderr, " -ch path to access host /proc and /sys when running in a container. Default: empty.\n");
252 fprintf(stderr, " -nd or -nodeamon to disable forking in the background. Default: unset.\n");
253 fprintf(stderr, " -df FLAGS debug options. Default: 0x%08llx.\n", debug_flags);
254 fprintf(stderr, " -stacksize BYTES to overwrite the pthread stack size.\n");
255 fprintf(stderr, " -pidfile FILENAME to save a pid while running.\n");
260 if(!config_loaded) load_config(NULL, 0);
262 // prepare configuration environment variables for the plugins
263 setenv("NETDATA_CONFIG_DIR" , config_get("global", "config directory" , CONFIG_DIR) , 1);
264 setenv("NETDATA_PLUGINS_DIR", config_get("global", "plugins directory" , PLUGINS_DIR), 1);
265 setenv("NETDATA_WEB_DIR" , config_get("global", "web files directory", WEB_DIR) , 1);
266 setenv("NETDATA_CACHE_DIR" , config_get("global", "cache directory" , CACHE_DIR) , 1);
267 setenv("NETDATA_LIB_DIR" , config_get("global", "lib directory" , VARLIB_DIR) , 1);
268 setenv("NETDATA_LOG_DIR" , config_get("global", "log directory" , LOG_DIR) , 1);
269 setenv("NETDATA_HOST_PREFIX", config_get("global", "host access prefix" , "") , 1);
270 setenv("HOME" , config_get("global", "home directory" , CACHE_DIR) , 1);
272 // avoid extended to stat(/etc/localtime)
273 // http://stackoverflow.com/questions/4554271/how-to-avoid-excessive-stat-etc-localtime-calls-in-strftime-on-linux
274 setenv("TZ", ":/etc/localtime", 0);
276 // cd to /tmp to avoid any plugins writing files at random places
277 if(chdir("/tmp")) error("netdata: ERROR: Cannot cd to /tmp");
279 char *input_log_file = NULL;
280 char *output_log_file = NULL;
281 char *error_log_file = NULL;
282 char *access_log_file = NULL;
287 // --------------------------------------------------------------------
289 sprintf(buffer, "0x%08llx", 0ULL);
290 char *flags = config_get("global", "debug flags", buffer);
291 setenv("NETDATA_DEBUG_FLAGS", flags, 1);
293 debug_flags = strtoull(flags, NULL, 0);
294 debug(D_OPTIONS, "Debug flags set to '0x%8llx'.", debug_flags);
296 if(debug_flags != 0) {
297 struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
298 if(setrlimit(RLIMIT_CORE, &rl) != 0)
299 info("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
300 prctl(PR_SET_DUMPABLE, 1, 0, 0, 0);
303 // --------------------------------------------------------------------
305 #ifdef MADV_MERGEABLE
306 enable_ksm = config_get_boolean("global", "memory deduplication (ksm)", enable_ksm);
308 #warning "Kernel memory deduplication (KSM) is not available"
311 // --------------------------------------------------------------------
314 global_host_prefix = config_get("global", "host access prefix", "");
315 setenv("NETDATA_HOST_PREFIX", global_host_prefix, 1);
317 // --------------------------------------------------------------------
319 output_log_file = config_get("global", "debug log", LOG_DIR "/debug.log");
320 if(strcmp(output_log_file, "syslog") == 0) {
321 output_log_syslog = 1;
322 output_log_file = NULL;
324 else if(strcmp(output_log_file, "none") == 0) {
325 output_log_syslog = 0;
326 output_log_file = NULL;
328 else output_log_syslog = 0;
330 // --------------------------------------------------------------------
332 error_log_file = config_get("global", "error log", LOG_DIR "/error.log");
333 if(strcmp(error_log_file, "syslog") == 0) {
334 error_log_syslog = 1;
335 error_log_file = NULL;
337 else if(strcmp(error_log_file, "none") == 0) {
338 error_log_syslog = 0;
339 error_log_file = NULL;
340 // optimization - do not even generate debug log entries
342 else error_log_syslog = 0;
344 error_log_throttle_period = config_get_number("global", "errors flood protection period", error_log_throttle_period);
345 setenv("NETDATA_ERRORS_THROTTLE_PERIOD", config_get("global", "errors flood protection period" , ""), 1);
347 error_log_errors_per_period = config_get_number("global", "errors to trigger flood protection", error_log_errors_per_period);
348 setenv("NETDATA_ERRORS_PER_PERIOD" , config_get("global", "errors to trigger flood protection", ""), 1);
350 // --------------------------------------------------------------------
352 access_log_file = config_get("global", "access log", LOG_DIR "/access.log");
353 if(strcmp(access_log_file, "syslog") == 0) {
354 access_log_syslog = 1;
355 access_log_file = NULL;
357 else if(strcmp(access_log_file, "none") == 0) {
358 access_log_syslog = 0;
359 access_log_file = NULL;
361 else access_log_syslog = 0;
363 // --------------------------------------------------------------------
365 rrd_memory_mode = rrd_memory_mode_id(config_get("global", "memory mode", rrd_memory_mode_name(rrd_memory_mode)));
367 // --------------------------------------------------------------------
369 if(gethostname(buffer, HOSTNAME_MAX) == -1)
370 error("WARNING: Cannot get machine hostname.");
371 hostname = config_get("global", "hostname", buffer);
372 debug(D_OPTIONS, "hostname set to '%s'", hostname);
374 // --------------------------------------------------------------------
376 rrd_default_history_entries = (int) config_get_number("global", "history", RRD_DEFAULT_HISTORY_ENTRIES);
377 if(rrd_default_history_entries < 5 || rrd_default_history_entries > RRD_HISTORY_ENTRIES_MAX) {
378 info("Invalid save lines %d given. Defaulting to %d.", rrd_default_history_entries, RRD_DEFAULT_HISTORY_ENTRIES);
379 rrd_default_history_entries = RRD_DEFAULT_HISTORY_ENTRIES;
382 debug(D_OPTIONS, "save lines set to %d.", rrd_default_history_entries);
385 // --------------------------------------------------------------------
387 rrd_update_every = (int) config_get_number("global", "update every", UPDATE_EVERY);
388 if(rrd_update_every < 1 || rrd_update_every > 600) {
389 info("Invalid update timer %d given. Defaulting to %d.", rrd_update_every, UPDATE_EVERY_MAX);
390 rrd_update_every = UPDATE_EVERY;
392 else debug(D_OPTIONS, "update timer set to %d.", rrd_update_every);
394 // let the plugins know the min update_every
397 snprintf(buf, 50, "%d", rrd_update_every);
398 setenv("NETDATA_UPDATE_EVERY", buf, 1);
401 // --------------------------------------------------------------------
403 // block signals while initializing threads.
404 // this causes the threads to block signals.
408 if(pthread_sigmask(SIG_BLOCK, &sigset, NULL) == -1) {
409 error("Could not block signals for threads");
412 // Catch signals which we want to use to quit savely
414 sigemptyset(&sa.sa_mask);
415 sigaddset(&sa.sa_mask, SIGHUP);
416 sigaddset(&sa.sa_mask, SIGINT);
417 sigaddset(&sa.sa_mask, SIGTERM);
418 sa.sa_handler = sig_handler;
419 if(sigaction(SIGHUP, &sa, NULL) == -1) {
420 error("Failed to change signal handler for SIGHUP");
422 if(sigaction(SIGINT, &sa, NULL) == -1) {
423 error("Failed to change signal handler for SIGINT");
425 if(sigaction(SIGTERM, &sa, NULL) == -1) {
426 error("Failed to change signal handler for SIGTERM");
428 // Ignore SIGPIPE completely.
429 // INFO: If we add signals here we have to unblock them
430 // at popen.c when running a external plugin.
431 sa.sa_handler = SIG_IGN;
432 if(sigaction(SIGPIPE, &sa, NULL) == -1) {
433 error("Failed to change signal handler for SIGTERM");
436 // --------------------------------------------------------------------
438 i = pthread_attr_init(&attr);
440 fatal("pthread_attr_init() failed with code %d.", i);
442 i = pthread_attr_getstacksize(&attr, &stacksize);
444 fatal("pthread_attr_getstacksize() failed with code %d.", i);
446 debug(D_OPTIONS, "initial pthread stack size is %zu bytes", stacksize);
448 wanted_stacksize = config_get_number("global", "pthread stack size", stacksize);
450 // --------------------------------------------------------------------
452 for (i = 0; static_threads[i].name != NULL ; i++) {
453 struct netdata_static_thread *st = &static_threads[i];
455 if(st->config_name) st->enabled = config_get_boolean(st->config_section, st->config_name, st->enabled);
456 if(st->enabled && st->init_routine) st->init_routine();
459 // --------------------------------------------------------------------
461 // get the user we should run
462 // IMPORTANT: this is required before web_files_uid()
463 user = config_get("global", "run as user" , (getuid() == 0)?NETDATA_USER:"");
465 // IMPORTANT: these have to run once, while single threaded
466 web_files_uid(); // IMPORTANT: web_files_uid() before web_files_gid()
469 // --------------------------------------------------------------------
471 listen_backlog = (int) config_get_number("global", "http port listen backlog", LISTEN_BACKLOG);
473 listen_port = (int) config_get_number("global", "port", LISTEN_PORT);
474 if(listen_port < 1 || listen_port > 65535) {
475 info("Invalid listen port %d given. Defaulting to %d.", listen_port, LISTEN_PORT);
476 listen_port = LISTEN_PORT;
478 else debug(D_OPTIONS, "Listen port set to %d.", listen_port);
481 char *ipv = config_get("global", "ip version", "any");
482 if(!strcmp(ipv, "any") || !strcmp(ipv, "both") || !strcmp(ipv, "all")) ip = 0;
483 else if(!strcmp(ipv, "ipv4") || !strcmp(ipv, "IPV4") || !strcmp(ipv, "IPv4") || !strcmp(ipv, "4")) ip = 4;
484 else if(!strcmp(ipv, "ipv6") || !strcmp(ipv, "IPV6") || !strcmp(ipv, "IPv6") || !strcmp(ipv, "6")) ip = 6;
485 else info("Cannot understand ip version '%s'. Assuming 'any'.", ipv);
487 if(ip == 0 || ip == 6) listen_fd = create_listen_socket6(config_get("global", "bind socket to IP", "*"), listen_port, listen_backlog);
489 listen_fd = create_listen_socket4(config_get("global", "bind socket to IP", "*"), listen_port, listen_backlog);
490 if(listen_fd >= 0 && ip != 4) info("Managed to open an IPv4 socket on port %d.", listen_port);
493 if(listen_fd < 0) fatal("Cannot listen socket.");
496 // never become a problem
497 if(nice(20) == -1) error("Cannot lower my CPU priority.");
499 if(become_daemon(dont_fork, 0, user, input_log_file, output_log_file, error_log_file, access_log_file, &access_fd, &stdaccess) == -1) {
500 fatal("Cannot demonize myself.");
504 if(debug_flags != 0) {
505 struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
506 if(setrlimit(RLIMIT_CORE, &rl) != 0)
507 info("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
509 prctl(PR_SET_DUMPABLE, 1, 0, 0, 0);
512 if(output_log_syslog || error_log_syslog || access_log_syslog)
513 openlog("netdata", LOG_PID, LOG_DAEMON);
515 info("NetData started on pid %d", getpid());
519 // ------------------------------------------------------------------------
520 // get default pthread stack size
522 if(stacksize < wanted_stacksize) {
523 i = pthread_attr_setstacksize(&attr, wanted_stacksize);
525 fatal("pthread_attr_setstacksize() to %zu bytes, failed with code %d.", wanted_stacksize, i);
527 info("Successfully set pthread stacksize to %zu bytes", wanted_stacksize);
530 // --------------------------------------------------------------------
531 // initialize the registry
536 // ------------------------------------------------------------------------
539 for (i = 0; static_threads[i].name != NULL ; i++) {
540 struct netdata_static_thread *st = &static_threads[i];
543 st->thread = malloc(sizeof(pthread_t));
545 fatal("Cannot allocate pthread_t memory");
547 info("Starting thread %s.", st->name);
549 if(pthread_create(st->thread, &attr, st->start_routine, NULL))
550 error("failed to create new thread for %s.", st->name);
552 else if(pthread_detach(*st->thread))
553 error("Cannot request detach of newly created %s thread.", st->name);
555 else info("Not starting thread %s.", st->name);
558 // ------------------------------------------------------------------------
559 // block signals while initializing threads.
563 if(pthread_sigmask(SIG_UNBLOCK, &sigset, NULL) == -1) {
564 error("Could not unblock signals for threads");
567 // Handle flags set in the signal handler.
571 info("Exit main loop of netdata.");
572 netdata_cleanup_and_exit(0);