12 #include <sys/types.h>
15 #include <sys/resource.h>
20 #include "web_server.h"
22 #include "appconfig.h"
23 #include "web_client.h"
27 #include "unit_test.h"
29 #include "plugins_d.h"
30 #include "plugin_idlejitter.h"
31 #include "plugin_tc.h"
32 #include "plugin_checks.h"
33 #include "plugin_proc.h"
34 #include "plugin_nfacct.h"
40 void netdata_cleanup_and_exit(int ret)
44 //unlink("/var/run/netdata.pid");
45 info("NetData exiting. Bye bye...");
49 struct netdata_static_thread {
59 void (*init_routine) (void);
60 void *(*start_routine) (void *);
63 struct netdata_static_thread static_threads[] = {
64 {"tc", "plugins", "tc", 1, NULL, NULL, tc_main},
65 {"idlejitter", "plugins", "idlejitter", 1, NULL, NULL, cpuidlejitter_main},
66 {"proc", "plugins", "proc", 1, NULL, NULL, proc_main},
68 #ifdef INTERNAL_PLUGIN_NFACCT
69 // nfacct requires root access
70 // so, we build it as an external plugin with setuid to root
71 {"nfacct", "plugins", "nfacct", 1, NULL, NULL, nfacct_main},
74 {"plugins.d", NULL, NULL, 1, NULL, NULL, pluginsd_main},
75 {"check", "plugins", "checks", 0, NULL, NULL, checks_main},
76 {"web", NULL, NULL, 1, NULL, NULL, socket_listen_main},
77 {NULL, NULL, NULL, 0, NULL, NULL, NULL}
80 int killpid(pid_t pid, int sig)
83 debug(D_EXIT, "Request to kill pid %d", pid);
86 if(kill(pid, 0) == -1) {
89 error("Request to kill pid %d, but it is not running.", pid);
93 error("Request to kill pid %d, but I do not have enough permissions.", pid);
97 error("Request to kill pid %d, but I received an error.", pid);
105 old = signal(sig, SIG_IGN);
107 error("Cannot overwrite signal handler for signal %d", sig);
111 ret = kill(pid, sig);
113 if(signal(sig, old) == SIG_ERR)
114 error("Cannot restore signal handler for signal %d", sig);
119 error("Cannot kill pid %d, but it is not running.", pid);
123 error("Cannot kill pid %d, but I do not have enough permissions.", pid);
127 error("Cannot kill pid %d, but I received an error.", pid);
140 struct web_client *w;
141 for(w = web_clients; w ; w = w->next) {
142 debug(D_EXIT, "Stopping web client %s", w->client_ip);
143 pthread_cancel(w->thread);
144 pthread_join(w->thread, NULL);
148 for (i = 0; static_threads[i].name != NULL ; i++) {
149 if(static_threads[i].thread) {
150 debug(D_EXIT, "Stopping %s thread", static_threads[i].name);
151 pthread_cancel(*static_threads[i].thread);
152 pthread_join(*static_threads[i].thread, NULL);
153 static_threads[i].thread = NULL;
158 info("Killing tc-qos-helper procees");
159 if(killpid(tc_child_pid, SIGTERM) != -1)
160 waitid(P_PID, tc_child_pid, &info, WEXITED);
165 for(cd = pluginsd_root ; cd ; cd = cd->next) {
166 debug(D_EXIT, "Stopping %s plugin thread", cd->id);
167 pthread_cancel(cd->thread);
168 pthread_join(cd->thread, NULL);
170 if(cd->pid && !cd->obsolete) {
171 debug(D_EXIT, "killing %s plugin process", cd->id);
172 if(killpid(cd->pid, SIGTERM) != -1)
173 waitid(P_PID, cd->pid, &info, WEXITED);
177 // if, for any reason there is any child exited
179 waitid(P_PID, 0, &info, WEXITED|WNOHANG);
181 debug(D_EXIT, "All threads/childs stopped.");
185 int main(int argc, char **argv)
188 int config_loaded = 0;
191 // global initialization
194 // set the name for logging
195 program_name = "netdata";
197 // parse the arguments
198 for(i = 1; i < argc ; i++) {
199 if(strcmp(argv[i], "-c") == 0 && (i+1) < argc) {
200 if(load_config(argv[i+1], 1) != 1) {
201 error("Cannot load configuration file %s.", argv[i+1]);
205 debug(D_OPTIONS, "Configuration loaded from %s.", argv[i+1]);
210 else if(strcmp(argv[i], "-df") == 0 && (i+1) < argc) { config_set("global", "debug flags", argv[i+1]); i++; }
211 else if(strcmp(argv[i], "-p") == 0 && (i+1) < argc) { config_set("global", "port", argv[i+1]); i++; }
212 else if(strcmp(argv[i], "-u") == 0 && (i+1) < argc) { config_set("global", "run as user", argv[i+1]); i++; }
213 else if(strcmp(argv[i], "-l") == 0 && (i+1) < argc) { config_set("global", "history", argv[i+1]); i++; }
214 else if(strcmp(argv[i], "-t") == 0 && (i+1) < argc) { config_set("global", "update every", argv[i+1]); i++; }
215 else if(strcmp(argv[i], "-ch") == 0 && (i+1) < argc) { config_set("global", "host access prefix", argv[i+1]); i++; }
216 else if(strcmp(argv[i], "-nodeamon") == 0 || strcmp(argv[i], "-nd") == 0) dont_fork = 1;
217 else if(strcmp(argv[i], "--unittest") == 0) {
218 if(unit_test_storage()) exit(1);
220 rrd_update_every = 1;
221 if(unit_test(1000000, 0)) exit(1);
222 if(unit_test(1000000, 500000)) exit(1);
223 if(unit_test(1000000, 100000)) exit(1);
224 if(unit_test(1000000, 900000)) exit(1);
225 if(unit_test(2000000, 0)) exit(1);
226 if(unit_test(2000000, 500000)) exit(1);
227 if(unit_test(2000000, 100000)) exit(1);
228 if(unit_test(2000000, 900000)) exit(1);
229 if(unit_test(500000, 500000)) exit(1);
230 //if(unit_test(500000, 100000)) exit(1); // FIXME: the expected numbers are wrong
231 //if(unit_test(500000, 900000)) exit(1); // FIXME: the expected numbers are wrong
232 if(unit_test(500000, 0)) exit(1);
233 fprintf(stderr, "\n\nALL TESTS PASSED\n\n");
237 fprintf(stderr, "Cannot understand option '%s'.\n", argv[i]);
238 fprintf(stderr, "\nUSAGE: %s [-d] [-l LINES_TO_SAVE] [-u UPDATE_TIMER] [-p LISTEN_PORT] [-dl debug log file] [-df debug flags].\n\n", argv[0]);
239 fprintf(stderr, " -c CONFIG FILE the configuration file to load. Default: %s.\n", CONFIG_DIR "/" CONFIG_FILENAME);
240 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);
241 fprintf(stderr, " -t UPDATE_TIMER can be from 1 to %d seconds. Default: %d.\n", UPDATE_EVERY_MAX, UPDATE_EVERY);
242 fprintf(stderr, " -p LISTEN_PORT can be from 1 to %d. Default: %d.\n", 65535, LISTEN_PORT);
243 fprintf(stderr, " -u USERNAME can be any system username to run as. Default: none.\n");
244 fprintf(stderr, " -ch path to access host /proc and /sys when running in a container. Default: empty.\n");
245 fprintf(stderr, " -nd or -nodeamon to disable forking in the background. Default: unset.\n");
246 fprintf(stderr, " -df FLAGS debug options. Default: 0x%08llx.\n", debug_flags);
251 if(!config_loaded) load_config(NULL, 0);
253 // prepare configuration environment variables for the plugins
254 setenv("NETDATA_CONFIG_DIR" , config_get("global", "config directory" , CONFIG_DIR) , 1);
255 setenv("NETDATA_PLUGINS_DIR", config_get("global", "plugins directory" , PLUGINS_DIR), 1);
256 setenv("NETDATA_WEB_DIR" , config_get("global", "web files directory", WEB_DIR) , 1);
257 setenv("NETDATA_CACHE_DIR" , config_get("global", "cache directory" , CACHE_DIR) , 1);
258 setenv("NETDATA_LOG_DIR" , config_get("global", "log directory" , LOG_DIR) , 1);
260 // avoid extended to stat(/etc/localtime)
261 // http://stackoverflow.com/questions/4554271/how-to-avoid-excessive-stat-etc-localtime-calls-in-strftime-on-linux
262 setenv("TZ", ":/etc/localtime", 0);
264 // cd to /tmp to avoid any plugins writing files at random places
265 if(chdir("/tmp")) error("netdata: ERROR: Cannot cd to /tmp");
267 char *input_log_file = NULL;
268 char *output_log_file = NULL;
269 char *error_log_file = NULL;
270 char *access_log_file = NULL;
275 // --------------------------------------------------------------------
277 sprintf(buffer, "0x%08llx", 0ULL);
278 char *flags = config_get("global", "debug flags", buffer);
279 setenv("NETDATA_DEBUG_FLAGS", flags, 1);
281 debug_flags = strtoull(flags, NULL, 0);
282 debug(D_OPTIONS, "Debug flags set to '0x%8llx'.", debug_flags);
284 if(debug_flags != 0) {
285 struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
286 if(setrlimit(RLIMIT_CORE, &rl) != 0)
287 info("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
290 // --------------------------------------------------------------------
292 global_host_prefix = config_get("global", "host access prefix", "");
293 setenv("NETDATA_HOST_PREFIX", global_host_prefix, 1);
295 // --------------------------------------------------------------------
297 output_log_file = config_get("global", "debug log", LOG_DIR "/debug.log");
298 if(strcmp(output_log_file, "syslog") == 0) {
299 output_log_syslog = 1;
300 output_log_file = NULL;
302 else if(strcmp(output_log_file, "none") == 0) {
303 output_log_syslog = 0;
304 output_log_file = NULL;
306 else output_log_syslog = 0;
308 // --------------------------------------------------------------------
311 error_log_file = config_get("global", "error log", LOG_DIR "/error.log");
312 if(strcmp(error_log_file, "syslog") == 0) {
313 error_log_syslog = 1;
314 error_log_file = NULL;
316 else if(strcmp(error_log_file, "none") == 0) {
317 error_log_syslog = 0;
318 error_log_file = NULL;
319 silent = 1; // optimization - do not even generate debug log entries
321 else error_log_syslog = 0;
323 // --------------------------------------------------------------------
325 access_log_file = config_get("global", "access log", LOG_DIR "/access.log");
326 if(strcmp(access_log_file, "syslog") == 0) {
327 access_log_syslog = 1;
328 access_log_file = NULL;
330 else if(strcmp(access_log_file, "none") == 0) {
331 access_log_syslog = 0;
332 access_log_file = NULL;
334 else access_log_syslog = 0;
336 // --------------------------------------------------------------------
338 rrd_memory_mode = rrd_memory_mode_id(config_get("global", "memory mode", rrd_memory_mode_name(rrd_memory_mode)));
340 // --------------------------------------------------------------------
342 if(gethostname(buffer, HOSTNAME_MAX) == -1)
343 error("WARNING: Cannot get machine hostname.");
344 hostname = config_get("global", "hostname", buffer);
345 debug(D_OPTIONS, "hostname set to '%s'", hostname);
347 // --------------------------------------------------------------------
349 rrd_default_history_entries = config_get_number("global", "history", RRD_DEFAULT_HISTORY_ENTRIES);
350 if(rrd_default_history_entries < 5 || rrd_default_history_entries > RRD_HISTORY_ENTRIES_MAX) {
351 fprintf(stderr, "Invalid save lines %d given. Defaulting to %d.\n", rrd_default_history_entries, RRD_DEFAULT_HISTORY_ENTRIES);
352 rrd_default_history_entries = RRD_DEFAULT_HISTORY_ENTRIES;
355 debug(D_OPTIONS, "save lines set to %d.", rrd_default_history_entries);
358 // --------------------------------------------------------------------
360 rrd_update_every = config_get_number("global", "update every", UPDATE_EVERY);
361 if(rrd_update_every < 1 || rrd_update_every > 600) {
362 fprintf(stderr, "Invalid update timer %d given. Defaulting to %d.\n", rrd_update_every, UPDATE_EVERY_MAX);
363 rrd_update_every = UPDATE_EVERY;
365 else debug(D_OPTIONS, "update timer set to %d.", rrd_update_every);
367 // let the plugins know the min update_every
370 snprintf(buffer, 50, "%d", rrd_update_every);
371 setenv("NETDATA_UPDATE_EVERY", buffer, 1);
374 // --------------------------------------------------------------------
376 for (i = 0; static_threads[i].name != NULL ; i++) {
377 struct netdata_static_thread *st = &static_threads[i];
379 if(st->config_name) st->enabled = config_get_boolean(st->config_section, st->config_name, st->enabled);
380 if(st->enabled && st->init_routine) st->init_routine();
383 // --------------------------------------------------------------------
386 user = config_get("global", "run as user" , (getuid() == 0)?NETDATA_USER:"");
389 // --------------------------------------------------------------------
391 listen_backlog = config_get_number("global", "http port listen backlog", LISTEN_BACKLOG);
393 listen_port = config_get_number("global", "port", LISTEN_PORT);
394 if(listen_port < 1 || listen_port > 65535) {
395 fprintf(stderr, "Invalid listen port %d given. Defaulting to %d.\n", listen_port, LISTEN_PORT);
396 listen_port = LISTEN_PORT;
398 else debug(D_OPTIONS, "Listen port set to %d.", listen_port);
401 char *ipv = config_get("global", "ip version", "any");
402 if(!strcmp(ipv, "any") || !strcmp(ipv, "both") || !strcmp(ipv, "all")) ip = 0;
403 else if(!strcmp(ipv, "ipv4") || !strcmp(ipv, "IPV4") || !strcmp(ipv, "IPv4") || !strcmp(ipv, "4")) ip = 4;
404 else if(!strcmp(ipv, "ipv6") || !strcmp(ipv, "IPV6") || !strcmp(ipv, "IPv6") || !strcmp(ipv, "6")) ip = 6;
405 else fprintf(stderr, "Cannot understand ip version '%s'. Assumming 'any'.", ipv);
407 if(ip == 0 || ip == 6) listen_fd = create_listen_socket6(listen_port, listen_backlog);
409 listen_fd = create_listen_socket4(listen_port, listen_backlog);
410 if(listen_fd >= 0 && ip != 4) fprintf(stderr, "Managed to open an IPv4 socket on port %d.", listen_port);
413 if(listen_fd < 0) fatal("Cannot listen socket.");
416 // never become a problem
417 if(nice(20) == -1) error("Cannot lower my CPU priority.");
419 if(become_daemon(dont_fork, 0, user, input_log_file, output_log_file, error_log_file, access_log_file, &access_fd, &stdaccess) == -1) {
420 fatal("Cannot demonize myself.");
424 if(output_log_syslog || error_log_syslog || access_log_syslog)
425 openlog("netdata", LOG_PID, LOG_DAEMON);
427 info("NetData started on pid %d", getpid());
431 for (i = 1 ; i < 65 ;i++) {
433 case SIGKILL: // not catchable
434 case SIGSTOP: // not catchable
444 signal(i, sig_handler);
449 for (i = 0; static_threads[i].name != NULL ; i++) {
450 struct netdata_static_thread *st = &static_threads[i];
453 st->thread = malloc(sizeof(pthread_t));
455 info("Starting thread %s.", st->name);
457 if(pthread_create(st->thread, NULL, st->start_routine, NULL))
458 error("failed to create new thread for %s.", st->name);
460 else if(pthread_detach(*st->thread))
461 error("Cannot request detach of newly created %s thread.", st->name);
463 else info("Not starting thread %s.", st->name);
466 // for future use - the main thread
468 if(netdata_exit != 0) {
471 if(netdata_exit > 5) {
472 netdata_cleanup_and_exit(0);