]> arthur.barton.de Git - netdata.git/blobdiff - src/main.c
fixed all crashes; empty responses are now formatted properly; fixed exit procedure...
[netdata.git] / src / main.c
index 662d4ace6c3a0560cbf96449b905a8e17b68148a..d73dd5901d85a071185efae4869df6c2c2da3d3d 100755 (executable)
@@ -1,3 +1,6 @@
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
 #include <stdio.h>
 #include <stdlib.h>
 #include <unistd.h>
 #include <errno.h>
 #include <sys/types.h>
 #include <sys/wait.h>
+#include <sys/time.h>
+#include <sys/resource.h>
 
+#include "common.h"
 #include "log.h"
 #include "daemon.h"
 #include "web_server.h"
 #include "popen.h"
-#include "config.h"
+#include "appconfig.h"
 #include "web_client.h"
 #include "rrd.h"
 #include "rrd2json.h"
 #include "plugin_tc.h"
 #include "plugin_checks.h"
 #include "plugin_proc.h"
+#include "plugin_nfacct.h"
 
 #include "main.h"
 
+int netdata_exit = 0;
+
+void netdata_cleanup_and_exit(int ret)
+{
+       kill_childs();
+       rrdset_free_all();
+       //unlink("/var/run/netdata.pid");
+       info("NetData exiting. Bye bye...");
+       exit(ret);
+}
+
 struct netdata_static_thread {
        char *name;
 
        char *config_section;
        char *config_name;
 
+       int enabled;
+
        pthread_t *thread;
+
+       void (*init_routine) (void);
        void *(*start_routine) (void *);
 };
 
 struct netdata_static_thread static_threads[] = {
-       {"tc",                  "plugins",      "tc",                   NULL, tc_main},
-       {"idlejitter",  "plugins",      "idlejitter",   NULL, cpuidlejitter_main},
-       {"proc",                "plugins",      "proc",                 NULL, proc_main},
-       {"plugins.d",   NULL,           NULL,                   NULL, pluginsd_main},
-       {"check",               "plugins",      "checks",               NULL, checks_main},
-       {"web",                 NULL,           NULL,                   NULL, socket_listen_main},
-       {NULL,                  NULL,           NULL,                   NULL, NULL}
+       {"tc",                  "plugins",      "tc",                   1, NULL, NULL,  tc_main},
+       {"idlejitter",  "plugins",      "idlejitter",   1, NULL, NULL,  cpuidlejitter_main},
+       {"proc",                "plugins",      "proc",                 1, NULL, NULL,  proc_main},
+
+#ifdef INTERNAL_PLUGIN_NFACCT
+       // nfacct requires root access
+       // so, we build it as an external plugin with setuid to root
+       {"nfacct",              "plugins",      "nfacct",               1, NULL, NULL,  nfacct_main},
+#endif
+
+       {"plugins.d",   NULL,           NULL,                   1, NULL, NULL,  pluginsd_main},
+       {"check",               "plugins",      "checks",               0, NULL, NULL,  checks_main},
+       {"web",                 NULL,           NULL,                   1, NULL, NULL,  socket_listen_main},
+       {NULL,                  NULL,           NULL,                   0, NULL, NULL,  NULL}
 };
 
+int killpid(pid_t pid, int sig)
+{
+       int ret = -1;
+       debug(D_EXIT, "Request to kill pid %d", pid);
+
+       errno = 0;
+       if(kill(pid, 0) == -1) {
+               switch(errno) {
+                       case ESRCH:
+                               error("Request to kill pid %d, but it is not running.", pid);
+                               break;
+
+                       case EPERM:
+                               error("Request to kill pid %d, but I do not have enough permissions.", pid);
+                               break;
+
+                       default:
+                               error("Request to kill pid %d, but I received an error.", pid);
+                               break;
+               }
+       }
+       else {
+               errno = 0;
+
+               void (*old)(int);
+               old = signal(sig, SIG_IGN);
+               if(old == SIG_ERR) {
+                       error("Cannot overwrite signal handler for signal %d", sig);
+                       old = sig_handler;
+               }
+
+               ret = kill(pid, sig);
+
+               if(signal(sig, old) == SIG_ERR)
+                       error("Cannot restore signal handler for signal %d", sig);
+
+               if(ret == -1) {
+                       switch(errno) {
+                               case ESRCH:
+                                       error("Cannot kill pid %d, but it is not running.", pid);
+                                       break;
+
+                               case EPERM:
+                                       error("Cannot kill pid %d, but I do not have enough permissions.", pid);
+                                       break;
+
+                               default:
+                                       error("Cannot kill pid %d, but I received an error.", pid);
+                                       break;
+                       }
+               }
+       }
+
+       return ret;
+}
+
 void kill_childs()
 {
        siginfo_t info;
@@ -59,7 +144,7 @@ void kill_childs()
                pthread_join(w->thread, NULL);
        }
 
-       int i;  
+       int i;
        for (i = 0; static_threads[i].name != NULL ; i++) {
                if(static_threads[i].thread) {
                        debug(D_EXIT, "Stopping %s thread", static_threads[i].name);
@@ -70,9 +155,9 @@ void kill_childs()
        }
 
        if(tc_child_pid) {
-               debug(D_EXIT, "Killing tc-qos-helper procees");
-               kill(tc_child_pid, SIGTERM);
-               waitid(tc_child_pid, 0, &info, WEXITED);
+               info("Killing tc-qos-helper procees");
+               if(killpid(tc_child_pid, SIGTERM) != -1)
+                       waitid(P_PID, tc_child_pid, &info, WEXITED);
        }
        tc_child_pid = 0;
 
@@ -84,11 +169,15 @@ void kill_childs()
 
                if(cd->pid && !cd->obsolete) {
                        debug(D_EXIT, "killing %s plugin process", cd->id);
-                       kill(cd->pid, SIGTERM);
-                       waitid(cd->pid, 0, &info, WEXITED);
+                       if(killpid(cd->pid, SIGTERM) != -1)
+                               waitid(P_PID, cd->pid, &info, WEXITED);
                }
        }
 
+       // if, for any reason there is any child exited
+       // catch it here
+       waitid(P_PID, 0, &info, WEXITED|WNOHANG);
+
        debug(D_EXIT, "All threads/childs stopped.");
 }
 
@@ -97,6 +186,13 @@ int main(int argc, char **argv)
 {
        int i;
        int config_loaded = 0;
+       int dont_fork = 0;
+
+       // global initialization
+       get_HZ();
+
+       // set the name for logging
+       program_name = "netdata";
 
        // parse  the arguments
        for(i = 1; i < argc ; i++) {
@@ -116,10 +212,12 @@ int main(int argc, char **argv)
                else if(strcmp(argv[i], "-u")  == 0 && (i+1) < argc) { config_set("global", "run as user",  argv[i+1]); i++; }
                else if(strcmp(argv[i], "-l")  == 0 && (i+1) < argc) { config_set("global", "history",      argv[i+1]); i++; }
                else if(strcmp(argv[i], "-t")  == 0 && (i+1) < argc) { config_set("global", "update every", argv[i+1]); i++; }
+               else if(strcmp(argv[i], "-ch") == 0 && (i+1) < argc) { config_set("global", "host access prefix", argv[i+1]); i++; }
+               else if(strcmp(argv[i], "-nodeamon") == 0 || strcmp(argv[i], "-nd") == 0) dont_fork = 1;
                else if(strcmp(argv[i], "--unittest")  == 0) {
                        if(unit_test_storage()) exit(1);
                        exit(0);
-                       update_every = 1;
+                       rrd_update_every = 1;
                        if(unit_test(1000000, 0)) exit(1);
                        if(unit_test(1000000, 500000)) exit(1);
                        if(unit_test(1000000, 100000)) exit(1);
@@ -139,21 +237,38 @@ int main(int argc, char **argv)
                        fprintf(stderr, "Cannot understand option '%s'.\n", argv[i]);
                        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]);
                        fprintf(stderr, "  -c CONFIG FILE the configuration file to load. Default: %s.\n", CONFIG_DIR "/" CONFIG_FILENAME);
-                       fprintf(stderr, "  -l LINES_TO_SAVE can be from 5 to %d lines in JSON data. Default: %d.\n", HISTORY_MAX, HISTORY);
+                       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);
                        fprintf(stderr, "  -t UPDATE_TIMER can be from 1 to %d seconds. Default: %d.\n", UPDATE_EVERY_MAX, UPDATE_EVERY);
                        fprintf(stderr, "  -p LISTEN_PORT can be from 1 to %d. Default: %d.\n", 65535, LISTEN_PORT);
                        fprintf(stderr, "  -u USERNAME can be any system username to run as. Default: none.\n");
-                       fprintf(stderr, "  -df FLAGS debug options. Default: 0x%8llx.\n", debug_flags);
+                       fprintf(stderr, "  -ch path to access host /proc and /sys when running in a container. Default: empty.\n");
+                       fprintf(stderr, "  -nd or -nodeamon to disable forking in the background. Default: unset.\n");
+                       fprintf(stderr, "  -df FLAGS debug options. Default: 0x%08llx.\n", debug_flags);
                        exit(1);
                }
        }
 
        if(!config_loaded) load_config(NULL, 0);
 
+       // prepare configuration environment variables for the plugins
+       setenv("NETDATA_CONFIG_DIR" , config_get("global", "config directory"   , CONFIG_DIR) , 1);
+       setenv("NETDATA_PLUGINS_DIR", config_get("global", "plugins directory"  , PLUGINS_DIR), 1);
+       setenv("NETDATA_WEB_DIR"    , config_get("global", "web files directory", WEB_DIR)    , 1);
+       setenv("NETDATA_CACHE_DIR"  , config_get("global", "cache directory"    , CACHE_DIR)  , 1);
+       setenv("NETDATA_LOG_DIR"    , config_get("global", "log directory"      , LOG_DIR)    , 1);
+
+       // avoid extended to stat(/etc/localtime)
+       // http://stackoverflow.com/questions/4554271/how-to-avoid-excessive-stat-etc-localtime-calls-in-strftime-on-linux
+       setenv("TZ", ":/etc/localtime", 0);
+
+       // cd to /tmp to avoid any plugins writing files at random places
+       if(chdir("/tmp")) fprintf(stderr, "netdata: ERROR: Cannot cd to /tmp: %s", strerror(errno));
+
        char *input_log_file = NULL;
        char *output_log_file = NULL;
        char *error_log_file = NULL;
        char *access_log_file = NULL;
+       char *user = NULL;
        {
                char buffer[1024];
 
@@ -161,9 +276,22 @@ int main(int argc, char **argv)
 
                sprintf(buffer, "0x%08llx", 0ULL);
                char *flags = config_get("global", "debug flags", buffer);
+               setenv("NETDATA_DEBUG_FLAGS", flags, 1);
+
                debug_flags = strtoull(flags, NULL, 0);
                debug(D_OPTIONS, "Debug flags set to '0x%8llx'.", debug_flags);
 
+               if(debug_flags != 0) {
+                       struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
+                       if(setrlimit(RLIMIT_CORE, &rl) != 0)
+                               info("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
+               }
+
+               // --------------------------------------------------------------------
+
+               global_host_prefix = config_get("global", "host access prefix", "");
+               setenv("NETDATA_HOST_PREFIX", global_host_prefix, 1);
+
                // --------------------------------------------------------------------
 
                output_log_file = config_get("global", "debug log", LOG_DIR "/debug.log");
@@ -207,7 +335,7 @@ int main(int argc, char **argv)
 
                // --------------------------------------------------------------------
 
-               memory_mode = memory_mode_id(config_get("global", "memory mode", memory_mode_name(memory_mode)));
+               rrd_memory_mode = rrd_memory_mode_id(config_get("global", "memory mode", rrd_memory_mode_name(rrd_memory_mode)));
 
                // --------------------------------------------------------------------
 
@@ -218,59 +346,81 @@ int main(int argc, char **argv)
 
                // --------------------------------------------------------------------
 
-               save_history = config_get_number("global", "history", HISTORY);
-               if(save_history < 5 || save_history > HISTORY_MAX) {
-                       fprintf(stderr, "Invalid save lines %d given. Defaulting to %d.\n", save_history, HISTORY);
-                       save_history = HISTORY;
+               rrd_default_history_entries = config_get_number("global", "history", RRD_DEFAULT_HISTORY_ENTRIES);
+               if(rrd_default_history_entries < 5 || rrd_default_history_entries > RRD_HISTORY_ENTRIES_MAX) {
+                       fprintf(stderr, "Invalid save lines %d given. Defaulting to %d.\n", rrd_default_history_entries, RRD_DEFAULT_HISTORY_ENTRIES);
+                       rrd_default_history_entries = RRD_DEFAULT_HISTORY_ENTRIES;
                }
                else {
-                       debug(D_OPTIONS, "save lines set to %d.", save_history);
+                       debug(D_OPTIONS, "save lines set to %d.", rrd_default_history_entries);
                }
 
                // --------------------------------------------------------------------
 
-               prepare_rundir();
-               char *user = config_get("global", "run as user", (getuid() == 0)?"nobody":"");
-               if(*user) {
-                       if(become_user(user) != 0) {
-                               fprintf(stderr, "Cannot become user %s.\n", user);
-                               exit(1);
-                       }
-                       else debug(D_OPTIONS, "Successfully became user %s.", user);
+               rrd_update_every = config_get_number("global", "update every", UPDATE_EVERY);
+               if(rrd_update_every < 1 || rrd_update_every > 600) {
+                       fprintf(stderr, "Invalid update timer %d given. Defaulting to %d.\n", rrd_update_every, UPDATE_EVERY_MAX);
+                       rrd_update_every = UPDATE_EVERY;
+               }
+               else debug(D_OPTIONS, "update timer set to %d.", rrd_update_every);
+
+               // let the plugins know the min update_every
+               {
+                       char buffer[50];
+                       snprintf(buffer, 50, "%d", rrd_update_every);
+                       setenv("NETDATA_UPDATE_EVERY", buffer, 1);
                }
 
                // --------------------------------------------------------------------
 
-               update_every = config_get_number("global", "update every", UPDATE_EVERY);
-               if(update_every < 1 || update_every > 600) {
-                       fprintf(stderr, "Invalid update timer %d given. Defaulting to %d.\n", update_every, UPDATE_EVERY_MAX);
-                       update_every = UPDATE_EVERY;
+               for (i = 0; static_threads[i].name != NULL ; i++) {
+                       struct netdata_static_thread *st = &static_threads[i];
+
+                       if(st->config_name) st->enabled = config_get_boolean(st->config_section, st->config_name, st->enabled);
+                       if(st->enabled && st->init_routine) st->init_routine();
                }
-               else debug(D_OPTIONS, "update timer set to %d.", update_every);
 
                // --------------------------------------------------------------------
 
+               prepare_rundir();
+               user = config_get("global", "run as user"    , (getuid() == 0)?NETDATA_USER:"");
+               web_files_uid();
+
+               // --------------------------------------------------------------------
+
+               listen_backlog = config_get_number("global", "http port listen backlog", LISTEN_BACKLOG);
+
                listen_port = config_get_number("global", "port", LISTEN_PORT);
                if(listen_port < 1 || listen_port > 65535) {
                        fprintf(stderr, "Invalid listen port %d given. Defaulting to %d.\n", listen_port, LISTEN_PORT);
                        listen_port = LISTEN_PORT;
                }
-               else debug(D_OPTIONS, "listen port set to %d.", listen_port);
+               else debug(D_OPTIONS, "Listen port set to %d.", listen_port);
+
+               int ip = 0;
+               char *ipv = config_get("global", "ip version", "any");
+               if(!strcmp(ipv, "any") || !strcmp(ipv, "both") || !strcmp(ipv, "all")) ip = 0;
+               else if(!strcmp(ipv, "ipv4") || !strcmp(ipv, "IPV4") || !strcmp(ipv, "IPv4") || !strcmp(ipv, "4")) ip = 4;
+               else if(!strcmp(ipv, "ipv6") || !strcmp(ipv, "IPV6") || !strcmp(ipv, "IPv6") || !strcmp(ipv, "6")) ip = 6;
+               else fprintf(stderr, "Cannot understand ip version '%s'. Assumming 'any'.", ipv);
+
+               if(ip == 0 || ip == 6) listen_fd = create_listen_socket6(listen_port, listen_backlog);
+               if(listen_fd < 0) {
+                       listen_fd = create_listen_socket4(listen_port, listen_backlog);
+                       if(listen_fd >= 0 && ip != 4) fprintf(stderr, "Managed to open an IPv4 socket on port %d.", listen_port);
+               }
 
-               listen_fd = create_listen_socket(listen_port);
+               if(listen_fd < 0) fatal("Cannot listen socket.");
        }
 
        // never become a problem
-       if(nice(20) == -1) {
-               fprintf(stderr, "Cannot lower my CPU priority. Error: %s.\n", strerror(errno));
-       }
+       if(nice(20) == -1) fprintf(stderr, "Cannot lower my CPU priority. Error: %s.\n", strerror(errno));
 
-       if(become_daemon(0, input_log_file, output_log_file, error_log_file, access_log_file, &access_fd, &stdaccess) == -1) {
-               fprintf(stderr, "Cannot demonize myself (%s).", strerror(errno));
+       if(become_daemon(dont_fork, 0, user, input_log_file, output_log_file, error_log_file, access_log_file, &access_fd, &stdaccess) == -1) {
+               fatal("Cannot demonize myself (%s).", strerror(errno));
                exit(1);
        }
 
-
        if(output_log_syslog || error_log_syslog || access_log_syslog)
                openlog("netdata", LOG_PID, LOG_DAEMON);
 
@@ -278,29 +428,49 @@ int main(int argc, char **argv)
 
 
        // catch all signals
-       for (i = 1 ; i < 65 ;i++) if(i != SIGSEGV && i != SIGFPE) signal(i,  sig_handler);
-       
+       for (i = 1 ; i < 65 ;i++) {
+               switch(i) {
+                       case SIGSEGV:
+                       case SIGFPE:
+                       case SIGCHLD:
+                               signal(i, SIG_DFL);
+                               break;
+
+                       default:
+                               signal(i,  sig_handler);
+                               break;
+               }
+       }
+
        for (i = 0; static_threads[i].name != NULL ; i++) {
                struct netdata_static_thread *st = &static_threads[i];
-               int doit = 1;
-
-               if(st->config_name) doit = config_get_boolean(st->config_section, st->config_name, doit);
 
-               if(doit) {
+               if(st->enabled) {
                        st->thread = malloc(sizeof(pthread_t));
 
+                       info("Starting thread %s.", st->name);
+
                        if(pthread_create(st->thread, NULL, st->start_routine, NULL))
                                error("failed to create new thread for %s.", st->name);
 
                        else if(pthread_detach(*st->thread))
                                error("Cannot request detach of newly created %s thread.", st->name);
                }
+               else info("Not starting thread %s.", st->name);
        }
 
-       // the main process - the web server listener
-       // this never ends
-       // socket_listen_main(NULL);
-       while(1) process_childs(1);
+       // for future use - the main thread
+       while(1) {
+               if(netdata_exit != 0) {
+                       netdata_exit++;
+
+                       if(netdata_exit > 5) {
+                               netdata_cleanup_and_exit(0);
+                               exit(0);
+                       }
+               }
+               sleep(2);
+       }
 
        exit(0);
 }