exporter: SIGTERM/SIGINT/SIGHUP management

Ceph exporter manages SIGINT, SIGTERM and SIGHUP signals

Fixes: https://tracker.ceph.com/issues/68721

Signed-off-by: Juan Miguel Olmo Martínez <jolmomar@redhat.com>

5 files changed, 97 insertions, 27 deletions

diff --git a/src/exporter/DaemonMetricCollector.cc b/src/exporter/DaemonMetricCollector.cc
index 4b8a8131bcf..d27b3ac43c5 100644
--- a/src/exporter/DaemonMetricCollector.cc
+++ b/src/exporter/DaemonMetricCollector.cc
@@ -29,9 +29,16 @@ using json_object = boost::json::object;
 using json_value = boost::json::value;
 using json_array = boost::json::array;
 
-void DaemonMetricCollector::request_loop(boost::asio::steady_timer &timer) {
-  timer.async_wait([&](const boost::system::error_code &e) {
-    std::cerr << e << std::endl;
+void DaemonMetricCollector::request_loop() {
+  timer.async_wait([this](const boost::system::error_code &e) {
+    if (shutdown_flag) {
+      dout(1) << "Metric collector request loop cancelled" << dendl;
+      return;
+    }
+
+    if (e) return; // Exit on error or cancellation
+
+    dout(10) << "Getting metrics loop..." << dendl;
     update_sockets();
 
     bool sort_metrics = g_conf().get_val<bool>("exporter_sort_metrics");
@@ -42,19 +49,24 @@ void DaemonMetricCollector::request_loop(boost::asio::steady_timer &timer) {
     auto stats_period = g_conf().get_val<int64_t>("exporter_stats_period");
     // time to wait before sending requests again
     timer.expires_from_now(std::chrono::seconds(stats_period));
-    request_loop(timer);
+    request_loop();
   });
 }
 
 void DaemonMetricCollector::main() {
-  // time to wait before sending requests again
-
-  boost::asio::io_context io;
-  boost::asio::steady_timer timer{io, std::chrono::seconds(0)};
-  request_loop(timer);
+  shutdown_flag = false;
+  timer.expires_from_now(std::chrono::seconds(0));
+  request_loop();
   io.run();
 }
 
+void DaemonMetricCollector::shutdown(){
+  shutdown_flag = true;
+  timer.cancel();  // Explicitly cancel the timer
+  dout(1) << "Collector shutdown initiated, timer canceled" << dendl;
+  io.stop();
+}
+
 std::string DaemonMetricCollector::get_metrics() {
   const std::lock_guard<std::mutex> lock(metrics_mutex);
   return metrics;
@@ -499,3 +511,4 @@ DaemonMetricCollector &collector_instance() {
   static DaemonMetricCollector instance;
   return instance;
 }
+
diff --git a/src/exporter/DaemonMetricCollector.h b/src/exporter/DaemonMetricCollector.h
index 3302e95df91..5831a0fa3b0 100644
--- a/src/exporter/DaemonMetricCollector.h
+++ b/src/exporter/DaemonMetricCollector.h
@@ -1,17 +1,20 @@
 #pragma once
 
 #include "common/admin_socket_client.h"
+#include <atomic>
 #include <map>
 #include <string>
 #include <vector>
 
 #include <boost/asio/steady_timer.hpp>
+#include <boost/thread.hpp>
 #include <boost/json/object.hpp>
 #include <filesystem>
 #include <map>
 #include <string>
 #include <vector>
 
+
 struct pstat {
   unsigned long utime;
   unsigned long stime;
@@ -43,11 +46,16 @@ public:
   std::string metrics;
   std::pair<labels_t, std::string> add_fixed_name_metrics(std::string metric_name);
   void update_sockets();
+  void shutdown();
 
 private:
   std::mutex metrics_mutex;
   std::unique_ptr<MetricsBuilder> builder;
-  void request_loop(boost::asio::steady_timer &timer);
+  boost::asio::io_context io;
+  boost::asio::steady_timer timer{io};
+  std::atomic<bool> shutdown_flag{false};
+
+  void request_loop();
 
   void dump_asok_metric(boost::json::object perf_info,
                         boost::json::value perf_values, std::string name,
@@ -108,3 +116,4 @@ public:
 };
 
 DaemonMetricCollector &collector_instance();
+
diff --git a/src/exporter/ceph_exporter.cc b/src/exporter/ceph_exporter.cc
index 2e2c16bb085..44b67c7e615 100644
--- a/src/exporter/ceph_exporter.cc
+++ b/src/exporter/ceph_exporter.cc
@@ -1,16 +1,31 @@
 #include "common/ceph_argparse.h"
 #include "common/config.h"
-#include "exporter/DaemonMetricCollector.h"
-#include "exporter/web_server.h"
+#include "common/debug.h"
 #include "global/global_init.h"
 #include "global/global_context.h"
-
+#include "global/signal_handler.h"
+#include "exporter/DaemonMetricCollector.h"
+#include "exporter/web_server.h"
 #include <boost/thread/thread.hpp>
 #include <iostream>
 #include <map>
 #include <string>
+#include <atomic>
+#include <chrono>
+#include <thread>
 
 #define dout_context g_ceph_context
+#define dout_subsys ceph_subsys_ceph_exporter
+
+DaemonMetricCollector &collector = collector_instance();
+
+static void handle_signal(int signum)
+{
+  ceph_assert(signum == SIGINT || signum == SIGTERM);
+  derr << "*** Got signal " << sig_str(signum) << " ***" << dendl;
+  // Finish the DaemonMetricCollector
+  collector.shutdown();
+}
 
 static void usage() {
   std::cout << "usage: ceph-exporter [options]\n"
@@ -27,7 +42,6 @@ static void usage() {
 }
 
 int main(int argc, char **argv) {
-
   auto args = argv_to_vec(argc, argv);
   if (args.empty()) {
     std::cerr << argv[0] << ": -h or --help for usage" << std::endl;
@@ -64,8 +78,30 @@ int main(int argc, char **argv) {
   }
   common_init_finish(g_ceph_context);
 
+  // Register signal handlers
+  init_async_signal_handler();
+  register_async_signal_handler(SIGHUP, sighup_handler);
+  register_async_signal_handler_oneshot(SIGINT, handle_signal);
+  register_async_signal_handler_oneshot(SIGTERM, handle_signal);
+
+  // Start the web server thread
   boost::thread server_thread(web_server_thread_entrypoint);
-  DaemonMetricCollector &collector = collector_instance();
+
+  // Start the DaemonMetricCollector
   collector.main();
+
+  // Interrupted. Time to terminate
+  unregister_async_signal_handler(SIGHUP, sighup_handler);
+  unregister_async_signal_handler(SIGINT, handle_signal);
+  unregister_async_signal_handler(SIGTERM, handle_signal);
+  shutdown_async_signal_handler();
+
+  // Stop the web server thread by interrupting it
+  stop_web_server();
+  server_thread.interrupt();  // Interrupt the web server thread
   server_thread.join();
+
+  dout(1) << "Ceph exporter stopped" << dendl;
+
+  return 0;
 }
diff --git a/src/exporter/web_server.cc b/src/exporter/web_server.cc
index 96cc02b389f..c01205f26bb 100644
--- a/src/exporter/web_server.cc
+++ b/src/exporter/web_server.cc
@@ -28,6 +28,9 @@ namespace net = boost::asio;      // from <boost/asio.hpp>
 namespace ssl = boost::asio::ssl; // from <boost/asio/ssl.hpp>
 using tcp = boost::asio::ip::tcp; // from <boost/asio/ip/tcp.hpp>
 
+//common io context for the web servers
+std::shared_ptr<net::io_context> global_ioc;
+
 // Base class for common functionality
 class web_connection {
 public:
@@ -43,7 +46,7 @@ protected:
   web_connection(net::any_io_executor executor, std::chrono::seconds timeout)
       : deadline_(executor, timeout) {}
 
-    // Common request processing logic
+  // Common request processing logic
   void process_request() {
     response_.version(request_.version());
     response_.keep_alive(request_.keep_alive());
@@ -64,7 +67,7 @@ protected:
     write_response();
   }
 
-    // Construct a response message based on the request target
+  // Construct a response message based on the request target
   void create_response() {
     if (request_.target() == "/") {
         response_.result(http::status::moved_permanently);
@@ -81,7 +84,7 @@ protected:
     }
   }
 
-    // Asynchronously transmit the response message
+  // Asynchronously transmit the response message
   virtual void write_response() = 0;
 
   // Check whether we have spent enough time on this connection
@@ -228,28 +231,33 @@ void https_server(tcp::acceptor &acceptor, ssl::context &ssl_ctx) {
 }
 
 void run_http_server(const std::string& exporter_addr, short unsigned int port) {
-  net::io_context ioc{1};
-  tcp::acceptor acceptor{ioc, {net::ip::make_address(exporter_addr), port}};
-  tcp::socket socket{ioc};
+  tcp::acceptor acceptor{*global_ioc, {net::ip::make_address(exporter_addr), port}};
+  tcp::socket socket{*global_ioc};
 
   http_server(acceptor, socket);
 
   dout(1) << "HTTP server running on " << exporter_addr << ":" << port << dendl;
-  ioc.run();
+  global_ioc->run();
 }
 
 void run_https_server(const std::string& exporter_addr, short unsigned int port, const std::string& cert_file, const std::string& key_file) {
-  net::io_context ioc{1};
   ssl::context ssl_ctx(ssl::context::tlsv13);
 
   ssl_ctx.use_certificate_chain_file(cert_file);
   ssl_ctx.use_private_key_file(key_file, ssl::context::pem);
 
-  tcp::acceptor acceptor{ioc, {net::ip::make_address(exporter_addr), port}};
+  tcp::acceptor acceptor{*global_ioc, {net::ip::make_address(exporter_addr), port}};
   https_server(acceptor, ssl_ctx);
 
   dout(1) << "HTTPS server running on " << exporter_addr << ":" << port << dendl;
-  ioc.run();
+  global_ioc->run();
+}
+
+void stop_web_server() {
+  if (global_ioc) {
+    global_ioc->stop();
+    dout(1) << "Ceph exporter web server stopped" << dendl;
+  }
 }
 
 void web_server_thread_entrypoint() {
@@ -259,18 +267,21 @@ void web_server_thread_entrypoint() {
     std::string cert_file = g_conf().get_val<std::string>("exporter_cert_file");
     std::string key_file = g_conf().get_val<std::string>("exporter_key_file");
 
+    // Initialize global_ioc
+    global_ioc = std::make_shared<net::io_context>(1);
+
     if (cert_file.empty() && key_file.empty()) {
       run_http_server(exporter_addr, port);
     } else {
       try {
           run_https_server(exporter_addr, port, cert_file, key_file);
       } catch (const std::exception &e) {
-          dout(1) << "Failed to start HTTPS server: " << e.what() << dendl;
+          derr << "Failed to start HTTPS server: " << e.what() << dendl;
           exit(EXIT_FAILURE);
       }
     }
   } catch (std::exception const &e) {
-      dout(1) << "Error: " << e.what() << dendl;
+      derr << "Error: " << e.what() << dendl;
       exit(EXIT_FAILURE);
   }
 }
diff --git a/src/exporter/web_server.h b/src/exporter/web_server.h
index c3339a8d43a..c6d4c54eca4 100644
--- a/src/exporter/web_server.h
+++ b/src/exporter/web_server.h
@@ -3,3 +3,4 @@
 #include <string>
 
 void web_server_thread_entrypoint();
+void stop_web_server();