do not call cleanup and exit on fatal conditions during startup (netd…

…ata#19715)

src/daemon/daemon-shutdown.h

@@ -10,4 +10,10 @@ void cancel_main_threads(void);
 void abort_on_fatal_disable(void);
 void abort_on_fatal_enable(void);
 
+#ifdef OS_WINDOWS
+void netdata_cleanup_and_exit(EXIT_REASON reason, const char *action, const char *action_result, const char *action_data);
+#else
+void netdata_cleanup_and_exit(EXIT_REASON reason, const char *action, const char *action_result, const char *action_data) NORETURN;
+#endif
+
 #endif //NETDATA_DAEMON_SHUTDOWN_H

src/daemon/daemon-systemd-watcher.c

@@ -3,6 +3,7 @@
 #include "libnetdata/libnetdata.h"
 #include "daemon-systemd-watcher.h"
 #include "daemon-service.h"
+#include "daemon-shutdown.h"
 
 #ifdef ENABLE_SYSTEMD_DBUS
 

src/daemon/main.c

@@ -229,6 +229,16 @@ int unittest_prepare_rrd(const char **user) {
     return 0;
 }
 
+static void fatal_cleanup_and_exit_cb(void) {
+    netdata_cleanup_and_exit(EXIT_REASON_FATAL, "fatal error", "exiting", NULL);
+    exit(1);
+}
+
+static void fatal_status_file_save(void) {
+    daemon_status_file_update_status(DAEMON_STATUS_NONE);
+    exit(1);
+}
+
 int netdata_main(int argc, char **argv) {
     libjudy_malloc_init();
     string_init();
@@ -753,6 +763,7 @@ int netdata_main(int argc, char **argv) {
     // initialize the log files
     nd_log_initialize();
     nd_log_register_event_cb(daemon_status_file_register_fatal);
+    nd_log_register_fatal_cb(fatal_status_file_save);
 
     netdata_conf_section_global(); // get hostname, host prefix, profile, etc
     registry_init(); // for machine_guid, must be after netdata_conf_section_global()
@@ -1057,6 +1068,7 @@ int netdata_main(int argc, char **argv) {
 
     webrtc_initialize();
 
+    nd_log_register_fatal_cb(fatal_cleanup_and_exit_cb);
     daemon_status_file_startup_step(NULL);
     daemon_status_file_update_status(DAEMON_STATUS_RUNNING);
     return 10;

src/libnetdata/libnetdata.h

@@ -56,12 +56,6 @@ char *find_and_replace(const char *src, const char *find, const char *replace, c
 bool run_command_and_copy_output_to_stdout(const char *command, int max_line_length);
 struct web_buffer *run_command_and_get_output_to_buffer(const char *command, int max_line_length);
 
-#ifdef OS_WINDOWS
-void netdata_cleanup_and_exit(EXIT_REASON reason, const char *action, const char *action_result, const char *action_data);
-#else
-void netdata_cleanup_and_exit(EXIT_REASON reason, const char *action, const char *action_result, const char *action_data) NORETURN;
-#endif
-
 extern const char *netdata_configured_host_prefix;
 
 // safe includes before O/S specific functions

src/libnetdata/log/nd_log-internals.h

@@ -126,6 +126,7 @@ struct nd_log {
 
     ND_LOG_SOURCES overwrite_process_source;
     log_event_t log_event_cb;
+    fatal_event_t fatal_event_cb;
 
     struct nd_log_source sources[_NDLS_MAX];
 

src/libnetdata/log/nd_log.c

@@ -135,6 +135,12 @@ void nd_log_register_event_cb(log_event_t cb) {
     nd_log.log_event_cb = cb;
 }
 
+// --------------------------------------------------------------------------------------------------------------------
+
+void nd_log_register_fatal_cb(fatal_event_t cb) {
+    nd_log.fatal_event_cb = cb;
+}
+
 // --------------------------------------------------------------------------------------------------------------------
 // high level logger
 
@@ -525,8 +531,11 @@ void netdata_logger_fatal(const char *file, const char *function, const unsigned
 #endif
 
 #ifdef NETDATA_INTERNAL_CHECKS
-    abort();
+    // abort();
 #endif
 
-    netdata_cleanup_and_exit(EXIT_REASON_FATAL, "FATAL", action_result, action_data);
+    if(nd_log.fatal_event_cb)
+        nd_log.fatal_event_cb();
+
+    exit(1);
 }

src/libnetdata/log/nd_log.h

@@ -36,6 +36,9 @@ ND_UUID nd_log_get_invocation_id(void);
 typedef void (*log_event_t)(const char *filename, const char *function, const char *message, const char *errno_str, const char *stack_trace, long line);
 void nd_log_register_event_cb(log_event_t cb);
 
+typedef void (*fatal_event_t)(void);
+void nd_log_register_fatal_cb(fatal_event_t cb);
+
 int nd_log_health_fd(void);
 int nd_log_collectors_fd(void);
 typedef bool (*log_formatter_callback_t)(BUFFER *wb, void *data);

src/libnetdata/required_dummies.h

@@ -3,16 +3,6 @@
 #ifndef NETDATA_LIB_DUMMIES_H
 #define NETDATA_LIB_DUMMIES_H 1
 
-// callback required by fatal()
-void netdata_cleanup_and_exit(EXIT_REASON reason, const char *action, const char *action_result, const char *action_data)
-{
-    (void)action;
-    (void)action_result;
-    (void)action_data;
-
-    exit(reason == EXIT_REASON_FATAL ? 1 : 0);
-}
-
 void rrdset_thread_rda_free(void){}
 void sender_thread_buffer_free(void){}
 void query_target_free(void){}