Add systemd watchdog and adjust reload behaviour

CMakeLists.txt

@@ -38,6 +38,11 @@ option(ICINGA2_WITH_PERFDATA "Build the perfdata module" ON)
 option(ICINGA2_WITH_STUDIO "Build the Icinga Studio application" OFF)
 option(ICINGA2_WITH_TESTS "Run unit tests" ON)
 
+option (USE_SYSTEMD
+ "Configure icinga as native systemd service instead of a SysV initscript" OFF)
+
+set(HAVE_SYSTEMD ${USE_SYSTEMD})
+
 file(STRINGS icinga2.spec VERSION_LINE REGEX "^Version: ")
 string(REPLACE "Version: " "" ICINGA2_VERSION ${VERSION_LINE})
 
@@ -155,6 +160,11 @@ if(UNIX OR CYGWIN)
   list(APPEND base_OBJS $<TARGET_OBJECTS:execvpe>)
 endif()
 
+if(HAVE_SYSTEMD)
+  list(APPEND base_DEPS systemd)
+endif()
+
+
 if(EDITLINE_FOUND)
   list(APPEND base_DEPS ${EDITLINE_LIBRARIES})
   include_directories(${EDITLINE_INCLUDE_DIR})

config.h.cmake

@@ -9,6 +9,7 @@
 #cmakedefine HAVE_CXXABI_H
 #cmakedefine HAVE_NICE
 #cmakedefine HAVE_EDITLINE
+#cmakedefine HAVE_SYSTEMD
 
 #cmakedefine ICINGA2_UNITY_BUILD
 

doc/02-getting-started.md

@@ -360,6 +360,14 @@ content:
     StartLimitInterval=10
     StartLimitBurst=3
 
+Using the watchdog can also help with monitoring Icinga 2, to activate and use it add the following to the override:
+
+    WatchdogSec=30s
+
+This way Systemd will kill Icinga 2 if does not notify for over 30 seconds, a timout of less than 10 seconds is not
+recommended. When the watchdog is activated, `Restart=` can be set to `watchdog` to restart Icinga 2 in the case of a
+watchdog timeout.
+
 Run `systemctl daemon-reload && systemctl restart icinga2` to apply the changes.
 Now Systemd will always try to restart Icinga 2 (except if you run
 `systemctl stop icinga2`). After three failures in ten seconds it will stop

etc/initsystem/CMakeLists.txt

@@ -34,9 +34,6 @@ if(NOT WIN32)
     PERMISSIONS OWNER_READ OWNER_WRITE OWNER_EXECUTE GROUP_READ GROUP_EXECUTE WORLD_READ WORLD_EXECUTE
   )
 
-  option (USE_SYSTEMD
-    "Configure icinga as native systemd service instead of a SysV initscript" OFF)
-
   # required for packaging on Gentoo, see Bug #6498
   option (INSTALL_SYSTEMD_SERVICE_AND_INITSCRIPT
     "Force install both the systemd service definition file and the SysV initscript in parallel, regardless of how USE_SYSTEMD is set. Only use this for special packaging purposes and if you know what you are doing" OFF)

etc/initsystem/icinga2.service.cmake

@@ -3,10 +3,10 @@ Description=Icinga host/service/network monitoring system
 After=syslog.target network-online.target postgresql.service mariadb.service carbon-cache.service carbon-relay.service
 
 [Service]
-Type=forking
+Type=notify
 EnvironmentFile=@ICINGA2_SYSCONFIGFILE@
 ExecStartPre=@CMAKE_INSTALL_PREFIX@/lib/icinga2/prepare-dirs @ICINGA2_SYSCONFIGFILE@
-ExecStart=@CMAKE_INSTALL_FULL_SBINDIR@/icinga2 daemon -d -e ${ICINGA2_ERROR_LOG}
+ExecStart=@CMAKE_INSTALL_FULL_SBINDIR@/icinga2 daemon -e ${ICINGA2_ERROR_LOG}
 PIDFile=@ICINGA2_RUNDIR@/icinga2/icinga2.pid
 ExecReload=@CMAKE_INSTALL_PREFIX@/lib/icinga2/safe-reload @ICINGA2_SYSCONFIGFILE@
 TimeoutStartSec=30m

lib/base/application.cpp

@@ -44,10 +44,12 @@
 #ifdef __linux__
 #include <sys/prctl.h>
 #endif /* __linux__ */
-
 #ifdef _WIN32
 #include <windows.h>
-#endif /* _win32 */
+#endif /* _WIN32 */
+#ifdef HAVE_SYSTEMD
+#include <systemd/sd-daemon.h>
+#endif /* HAVE_SYSTEMD */
 
 using namespace icinga;
 
@@ -315,6 +317,11 @@ void Application::SetArgV(char **argv)
  */
 void Application::RunEventLoop()
 {
+
+#ifdef HAVE_SYSTEMD
+	sd_notify(0, "READY=1");
+#endif /* HAVE_SYSTEMD */
+
 	double lastLoop = Utility::GetTime();
 
 mainloop:
@@ -331,6 +338,10 @@ void Application::RunEventLoop()
 		double now = Utility::GetTime();
 		double timeDiff = lastLoop - now;
 
+#ifdef HAVE_SYSTEMD
+		sd_notify(0, "WATCHDOG=1");
+#endif /* HAVE_SYSTEMD */
+
 		if (std::fabs(timeDiff) > 15) {
 			/* We made a significant jump in time. */
 			Log(LogInformation, "Application")
@@ -347,6 +358,10 @@ void Application::RunEventLoop()
 	if (m_RequestRestart) {
 		m_RequestRestart = false;         // we are now handling the request, once is enough
 
+#ifdef HAVE_SYSTEMD
+		sd_notify(0, "RELOADING=1");
+#endif /* HAVE_SYSTEMD */
+
 		// are we already restarting? ignore request if we already are
 		if (l_Restarting)
 			goto mainloop;
@@ -357,6 +372,10 @@ void Application::RunEventLoop()
 		goto mainloop;
 	}
 
+#ifdef HAVE_SYSTEMD
+	sd_notify(0, "STOPPING=1");
+#endif /* HAVE_SYSTEMD */
+
 	Log(LogInformation, "Application", "Shutting down...");
 
 	ConfigObject::StopObjects();
@@ -712,6 +731,21 @@ void Application::SigUsr1Handler(int)
 	RequestReopenLogs();
 }
 
+/**
+ * Signal handler for SIGUSR2. Hands over PID to child and commits suicide
+ *
+ * @param - The signal number.
+ */
+void Application::SigUsr2Handler(int)
+{
+	Log(LogInformation, "Application", "Reload requested, letting new process take over.");
+#ifdef HAVE_SYSTEMD
+	sd_notifyf(0, "MAINPID=%lu", (unsigned long) m_ReloadProcess);
+#endif /* HAVE_SYSTEMD */
+
+	Exit(0);
+}
+
 /**
  * Signal handler for SIGABRT. Helps with debugging ASSERT()s.
  *
@@ -964,6 +998,9 @@ int Application::Run()
 
 	sa.sa_handler = &Application::SigUsr1Handler;
 	sigaction(SIGUSR1, &sa, nullptr);
+
+	sa.sa_handler = &Application::SigUsr2Handler;
+	sigaction(SIGUSR2, &sa, nullptr);
 #else /* _WIN32 */
 	SetConsoleCtrlHandler(&Application::CtrlHandler, TRUE);
 #endif /* _WIN32 */

lib/base/application.hpp

@@ -205,6 +205,7 @@ class Application : public ObjectImpl<Application> {
 
 	static void SigAbrtHandler(int signum);
 	static void SigUsr1Handler(int signum);
+	static void SigUsr2Handler(int signum);
 	static void ExceptionHandler();
 
 	static String GetCrashReportFilename();

lib/cli/daemoncommand.cpp

@@ -136,39 +136,6 @@ static bool SetDaemonIO(const String& stderrFile)
 	return true;
 }
 
-/**
- * Terminate another process and wait till it has ended
- *
- * @params target PID of the process to end
- */
-static void TerminateAndWaitForEnd(pid_t target)
-{
-#ifndef _WIN32
-	// allow 30 seconds timeout
-	double timeout = Utility::GetTime() + 30;
-
-	int ret = kill(target, SIGTERM);
-
-	while (Utility::GetTime() < timeout && (ret == 0 || errno != ESRCH)) {
-		Utility::Sleep(0.1);
-		ret = kill(target, 0);
-	}
-
-	// timeout and the process still seems to live: update pid and kill it
-	if (ret == 0 || errno != ESRCH) {
-		String pidFile = Application::GetPidPath();
-		std::ofstream fp(pidFile.CStr());
-		fp << Utility::GetPid();
-		fp.close();
-
-		kill(target, SIGKILL);
-	}
-
-#else
-	// TODO: implement this for Win32
-#endif /* _WIN32 */
-}
-
 String DaemonCommand::GetDescription() const
 {
 	return "Starts Icinga 2.";
@@ -251,11 +218,14 @@ int DaemonCommand::Run(const po::variables_map& vm, const std::vector<std::strin
 	}
 
 	if (vm.count("reload-internal")) {
-		int parentpid = vm["reload-internal"].as<int>();
-		Log(LogInformation, "cli")
-			<< "Terminating previous instance of Icinga (PID " << parentpid << ")";
-		TerminateAndWaitForEnd(parentpid);
-		Log(LogInformation, "cli", "Previous instance has ended, taking over now.");
+		/* We went through validation and now ask the old process kindly to die */
+		Log(LogInformation, "cli", "Requesting to take over.");
+		int rc = kill(vm["reload-internal"].as<int>(), SIGUSR2);
+		if (rc) {
+			Log(LogCritical, "Application")
+				<< "Failed to send signal to \"" << vm["reload-internal"].as<int>() <<  "\" with " << strerror(errno);
+			return EXIT_FAILURE;
+		}
 	}
 
 	if (vm.count("daemonize")) {