From: Valentine Krasnobaeva <vkrasnobaeva@haproxy.com>
Date: Mon, 7 Oct 2024 09:37:33 +0000 (+0200)
Subject: MEDIUM: mworker: add support of master recovery mode
X-Git-Tag: v3.1-dev10~15
X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=5909d508bc8432a0995076524aa431d76ef7ed38;p=thirdparty%2Fhaproxy.git

MEDIUM: mworker: add support of master recovery mode

In this commit we add run_master_in_recovery_mode(), which groups all necessary
initialization steps, which master should perform to be able to enter in its
polling loop (run_master()), when it fails while parsing its new config.

As exit_on_failure() is now adapted for master recovery mode. Let's register
it as atexit handler, when master enters in this mode. And let's remove
atexit_flag variable for master, because we no longer use it.

We also slightly refactor here read_cfg_in_discovery_mode() in order to call
run_master_in_recovery_mode() for the case, described above. Warning messages
are mandatory before calling the run_master_in_recovery_mode() as this allows
to stop haproxy with error, if it was launched in zero-warning mode.

So, in recovery mode master does not launch any worker. It just performs its
necessary initialization routines and enters in its polling loop to continue
to monitor the existed worker process.
---

diff --git a/src/haproxy.c b/src/haproxy.c
index 1d969b0548..52b7fc06e4 100644
--- a/src/haproxy.c
+++ b/src/haproxy.c
@@ -2158,8 +2158,6 @@ static void apply_master_worker_mode()
 		/* in parent */
 		ha_notice("Initializing new worker (%d)\n", worker_pid);
 		master = 1;
-		atexit_flag = 1;
-		atexit(exit_on_failure);
 
 		/* in exec mode, there's always exactly one thread. Failure to
 		 * set these ones now will result in nbthread being detected
@@ -3003,27 +3001,83 @@ static void step_init_4(void)
 	ha_free(&global.pidfile);
 }
 
+static void run_master_in_recovery_mode(int argc, char **argv)
+{
+	struct mworker_proc *proc;
+
+	/* increment the number failed reloads */
+	list_for_each_entry(proc, &proc_list, list) {
+		proc->failedreloads++;
+	}
+
+	global.nbtgroups = 1;
+	global.nbthread = 1;
+	master = 1;
+	atexit(exit_on_failure);
+
+	/* creates MASTER proxy and attaches server to child->ipc_fd[0] */
+	if (mworker_cli_proxy_create() < 0) {
+		ha_alert("Can't create the master's CLI.\n");
+		exit(EXIT_FAILURE);
+	}
+
+	/* master CLI */
+	mworker_create_master_cli();
+	step_init_2(argc, argv);
+	step_init_3();
+	if (protocol_bind_all(1) != 0) {
+		ha_alert("Master failed to bind master CLI socket.\n");
+		exit(1);
+	}
+
+	step_init_4();
+	/* enter in master polling loop */
+	run_master();
+}
+
 /* parse conf in disovery mode and set modes from config */
 static void read_cfg_in_discovery_mode(int argc, char **argv)
 {
 	struct cfgfile *cfg, *cfg_tmp;
-	int ret;
 
-	/* load configs and read only the keywords with KW_DISCOVERY flag */
+	/* load configs in memory and parse only global section (MODE_DISCOVERY) */
 	global.mode |= MODE_DISCOVERY;
 
 	usermsgs_clr("config");
-	ret = load_cfg(progname);
-	if (ret == 0) {
-		/* read only global section in discovery mode */
-		ret = read_cfg(progname);
+	if (load_cfg(progname) < 0) {
+		if (getenv("HAPROXY_MWORKER_REEXEC") != NULL) {
+			ha_warning("Master failed to load new configuration and "
+				   "can't start a new worker. Already running worker "
+				   "will be kept. Please, check configuration file path "
+				   "and memory limits and reload %s.\n", progname);
+			/* failed to load new conf, so setup master CLI for master side,
+			 * do some init steps and just enter in mworker_loop
+			 * to monitor the existed worker from previous start
+			 */
+			run_master_in_recovery_mode(argc, argv);
+			/* never get there */
+		} else
+			exit(1);
 	}
-	if (ret < 0) {
+
+	if (read_cfg(progname) < 0) {
 		list_for_each_entry_safe(cfg, cfg_tmp, &cfg_cfgfiles, list) {
 			ha_free(&cfg->content);
 			ha_free(&cfg->filename);
 		}
-		exit(1);
+		if (getenv("HAPROXY_MWORKER_REEXEC") != NULL) {
+			ha_warning("Master failed to parse new configuration and "
+				   "can't start a new worker. Already running worker "
+				   "will be kept. Please, check global section settings "
+				   "and memory limits and reload %s.\n", progname);
+			/* failed to load new conf, so setup master CLI for master side,
+			 * do some init steps and just enter in mworker_loop
+			 * to monitor the existed worker from previous start
+			 */
+			run_master_in_recovery_mode(argc, argv);
+			/* never get there */
+		} else
+			exit(1);
 	}
 	usermsgs_clr(NULL);