#include "command.h"
#include "memory_vty.h"
#include "libfrr.h"
+#include "lib_errors.h"
#include <getopt.h>
#include <sys/un.h>
#include <systemd.h>
#include "watchfrr.h"
+#include "watchfrr_errors.h"
#ifndef MIN
#define MIN(X,Y) (((X) <= (Y)) ? (X) : (Y))
#define FUZZY(X) ((X)+JITTER((X)/20))
#define DEFAULT_PERIOD 5
-#define DEFAULT_TIMEOUT 10
+#define DEFAULT_TIMEOUT 90
#define DEFAULT_RESTART_TIMEOUT 20
#define DEFAULT_LOGLEVEL LOG_INFO
#define DEFAULT_MIN_RESTART 60
#define DEFAULT_MAX_RESTART 600
-#ifdef PATH_WATCHFRR_PID
-#define DEFAULT_PIDFILE PATH_WATCHFRR_PID
-#else
-#define DEFAULT_PIDFILE STATEDIR "/watchfrr.pid"
-#endif
-#ifdef DAEMON_VTY_DIR
-#define VTYDIR DAEMON_VTY_DIR
-#else
-#define VTYDIR STATEDIR
-#endif
#define PING_TOKEN "PING"
/* Needs to be global, referenced somewhere inside libfrr. */
struct thread_master *master;
+static char pidfile_default[256];
-typedef enum {
- MODE_MONITOR = 0,
- MODE_GLOBAL_RESTART,
- MODE_SEPARATE_RESTART,
- MODE_PHASED_ZEBRA_RESTART,
- MODE_PHASED_ALL_RESTART
-} watch_mode_t;
-
-static const char *mode_str[] = {
- "monitor",
- "global restart",
- "individual daemon restart",
- "phased zebra restart",
- "phased global restart for any failure",
-};
+static bool watch_only = false;
typedef enum {
PHASE_NONE = 0,
};
static struct global_state {
- watch_mode_t mode;
restart_phase_t phase;
struct thread *t_phase_hanging;
const char *vtydir;
long restart_timeout;
long min_restart_interval;
long max_restart_interval;
- int do_ping;
struct daemon *daemons;
const char *restart_command;
const char *start_command;
const char *stop_command;
struct restart_info restart;
- int unresponsive_restart;
int loglevel;
- struct daemon *special; /* points to zebra when doing phased restart */
+ struct daemon *special; /* points to zebra when doing phased restart */
int numdaemons;
int numpids;
- int numdown; /* # of daemons that are not UP or UNRESPONSIVE */
+ int numdown; /* # of daemons that are not UP or UNRESPONSIVE */
} gs = {
-.mode = MODE_MONITOR,.phase = PHASE_NONE,.vtydir = VTYDIR,.period =
- 1000 * DEFAULT_PERIOD,.timeout =
- DEFAULT_TIMEOUT,.restart_timeout =
- DEFAULT_RESTART_TIMEOUT,.loglevel =
- DEFAULT_LOGLEVEL,.min_restart_interval =
- DEFAULT_MIN_RESTART,.max_restart_interval =
- DEFAULT_MAX_RESTART,.do_ping = 1,};
+ .phase = PHASE_NONE,
+ .vtydir = frr_vtydir,
+ .period = 1000 * DEFAULT_PERIOD,
+ .timeout = DEFAULT_TIMEOUT,
+ .restart_timeout = DEFAULT_RESTART_TIMEOUT,
+ .loglevel = DEFAULT_LOGLEVEL,
+ .min_restart_interval = DEFAULT_MIN_RESTART,
+ .max_restart_interval = DEFAULT_MAX_RESTART,
+};
typedef enum {
DAEMON_INIT,
DAEMON_UNRESPONSIVE
} daemon_state_t;
-#define IS_UP(DMN) \
- (((DMN)->state == DAEMON_UP) || ((DMN)->state == DAEMON_UNRESPONSIVE))
+#define IS_UP(DMN) \
+ (((DMN)->state == DAEMON_UP) || ((DMN)->state == DAEMON_UNRESPONSIVE))
static const char *state_str[] = {
- "Init",
- "Down",
- "Connecting",
- "Up",
- "Unresponsive",
+ "Init", "Down", "Connecting", "Up", "Unresponsive",
};
struct daemon {
daemon_state_t state;
int fd;
struct timeval echo_sent;
- u_int connect_tries;
+ unsigned int connect_tries;
struct thread *t_wakeup;
struct thread *t_read;
struct thread *t_write;
#define OPTION_MINRESTART 2000
#define OPTION_MAXRESTART 2001
+#define OPTION_DRY 2002
static const struct option longopts[] = {
{"daemon", no_argument, NULL, 'd'},
{"statedir", required_argument, NULL, 'S'},
- {"no-echo", no_argument, NULL, 'e'},
{"loglevel", required_argument, NULL, 'l'},
{"interval", required_argument, NULL, 'i'},
{"timeout", required_argument, NULL, 't'},
{"restart", required_argument, NULL, 'r'},
{"start-command", required_argument, NULL, 's'},
{"kill-command", required_argument, NULL, 'k'},
- {"restart-all", required_argument, NULL, 'R'},
- {"all-restart", no_argument, NULL, 'a'},
- {"always-all-restart", no_argument, NULL, 'A'},
- {"unresponsive-restart", no_argument, NULL, 'z'},
- {"min-restart-interval", required_argument, NULL, OPTION_MINRESTART },
- {"max-restart-interval", required_argument, NULL, OPTION_MAXRESTART },
+ {"dry", no_argument, NULL, OPTION_DRY},
+ {"min-restart-interval", required_argument, NULL, OPTION_MINRESTART},
+ {"max-restart-interval", required_argument, NULL, OPTION_MAXRESTART},
{"pid-file", required_argument, NULL, 'p'},
{"blank-string", required_argument, NULL, 'b'},
{"help", no_argument, NULL, 'h'},
{"version", no_argument, NULL, 'v'},
- {NULL, 0, NULL, 0}
-};
+ {NULL, 0, NULL, 0}};
static int try_connect(struct daemon *dmn);
static int wakeup_send_echo(struct thread *t_wakeup);
static const char *progname;
static void printhelp(FILE *target)
{
- fprintf(target, "Usage : %s [OPTION...] <daemon name> ...\n\n\
+ fprintf(target,
+ "Usage : %s [OPTION...] <daemon name> ...\n\n\
Watchdog program to monitor status of frr daemons and try to restart\n\
them if they are down or unresponsive. It determines whether a daemon is\n\
up based on whether it can connect to the daemon's vty unix stream socket.\n\
the daemon is responsive. If the daemon crashes, we will receive an EOF\n\
on the socket connection and know immediately that the daemon is down.\n\n\
The daemons to be monitored should be listed on the command line.\n\n\
-This program can run in one of 5 modes:\n\n\
-0. Mode: %s.\n\
- Just monitor and report on status changes. Example:\n\
- %s -d zebra ospfd bgpd\n\n\
-1. Mode: %s.\n\
- Whenever any daemon hangs or crashes, use the given command to restart\n\
- them all. Example:\n\
- %s -dz \\\n\
- -R '/sbin/service zebra restart; /sbin/service ospfd restart' \\\n\
- zebra ospfd\n\n\
-2. Mode: %s.\n\
- When any single daemon hangs or crashes, restart only the daemon that's\n\
- in trouble using the supplied restart command. Example:\n\
- %s -dz -r '/sbin/service %%s restart' zebra ospfd bgpd\n\n\
-3. Mode: %s.\n\
- The same as the previous mode, except that there is special treatment when\n\
- the zebra daemon is in trouble. In that case, a phased restart approach\n\
- is used: 1. stop all other daemons; 2. restart zebra; 3. start the other\n\
- daemons. Example:\n\
- %s -adz -r '/sbin/service %%s restart' \\\n\
- -s '/sbin/service %%s start' \\\n\
- -k '/sbin/service %%s stop' zebra ospfd bgpd\n\n\
-4. Mode: %s.\n\
- This is the same as the previous mode, except that the phased restart\n\
- procedure is used whenever any of the daemons hangs or crashes. Example:\n\
- %s -Adz -r '/sbin/service %%s restart' \\\n\
- -s '/sbin/service %%s start' \\\n\
- -k '/sbin/service %%s stop' zebra ospfd bgpd\n\n\
-As of this writing, it is believed that mode 2 [%s]\n\
-is not safe, and mode 3 [%s] may not be safe with some of the\n\
-routing daemons.\n\n\
In order to avoid attempting to restart the daemons in a fast loop,\n\
the -m and -M options allow you to control the minimum delay between\n\
restart commands. The minimum restart delay is recalculated each time\n\
a restart is attempted: if the time since the last restart attempt exceeds\n\
twice the -M value, then the restart delay is set to the -m value.\n\
-Otherwise, the interval is doubled (but capped at the -M value).\n\n", progname, mode_str[0], progname, mode_str[1], progname, mode_str[2], progname, mode_str[3], progname, mode_str[4], progname, mode_str[2], mode_str[3]);
+Otherwise, the interval is doubled (but capped at the -M value).\n\n",
+ progname);
- fprintf(target, "Options:\n\
+ fprintf(target,
+ "Options:\n\
-d, --daemon Run in daemon mode. In this mode, error messages are sent\n\
to syslog instead of stdout.\n\
-S, --statedir Set the vty socket directory (default is %s)\n\
--e, --no-echo Do not ping the daemons to test responsiveness (this\n\
- option is necessary if the daemons do not support the\n\
- echo command)\n\
-l, --loglevel Set the logging level (default is %d).\n\
The value should range from %d (LOG_EMERG) to %d (LOG_DEBUG),\n\
but it can be set higher than %d if extra-verbose debugging\n\
-r, --restart Supply a Bourne shell command to use to restart a single\n\
daemon. The command string should include '%%s' where the\n\
name of the daemon should be substituted.\n\
- Note that -r and -R are incompatible.\n\
-s, --start-command\n\
Supply a Bourne shell to command to use to start a single\n\
daemon. The command string should include '%%s' where the\n\
Supply a Bourne shell to command to use to stop a single\n\
daemon. The command string should include '%%s' where the\n\
name of the daemon should be substituted.\n\
--R, --restart-all\n\
- When one or more daemons is down, try to restart everything\n\
- using the Bourne shell command supplied as the argument.\n\
- Note that -r and -R are incompatible.\n\
--z, --unresponsive-restart\n\
- When a daemon is unresponsive, treat it as being down for\n\
- restart purposes.\n\
--a, --all-restart\n\
- When zebra hangs or crashes, restart all daemons using\n\
- this phased approach: 1. stop all other daemons; 2. restart\n\
- zebra; 3. start other daemons. Requires -r, -s, and -k.\n\
--A, --always-all-restart\n\
- When any daemon (not just zebra) hangs or crashes, use the\n\
- same phased restart mechanism described above for -a.\n\
- Requires -r, -s, and -k.\n\
+ --dry Do not start or restart anything, just log.\n\
-p, --pid-file Set process identifier file name\n\
(default is %s).\n\
-b, --blank-string\n\
When the supplied argument string is found in any of the\n\
- various shell command arguments (-r, -s, -k, or -R), replace\n\
+ various shell command arguments (-r, -s, or -k), replace\n\
it with a space. This is an ugly hack to circumvent problems\n\
passing command-line arguments with embedded spaces.\n\
-v, --version Print program version\n\
--h, --help Display this help and exit\n", VTYDIR, DEFAULT_LOGLEVEL, LOG_EMERG, LOG_DEBUG, LOG_DEBUG, DEFAULT_MIN_RESTART, DEFAULT_MAX_RESTART, DEFAULT_PERIOD, DEFAULT_TIMEOUT, DEFAULT_RESTART_TIMEOUT, DEFAULT_PIDFILE);
+-h, --help Display this help and exit\n",
+ frr_vtydir, DEFAULT_LOGLEVEL, LOG_EMERG, LOG_DEBUG, LOG_DEBUG,
+ DEFAULT_MIN_RESTART, DEFAULT_MAX_RESTART, DEFAULT_PERIOD,
+ DEFAULT_TIMEOUT, DEFAULT_RESTART_TIMEOUT, pidfile_default);
}
static pid_t run_background(char *shell_cmd)
switch (child = fork()) {
case -1:
- zlog_err("fork failed, cannot run command [%s]: %s",
- shell_cmd, safe_strerror(errno));
+ flog_err_sys(LIB_ERR_SYSTEM_CALL,
+ "fork failed, cannot run command [%s]: %s",
+ shell_cmd, safe_strerror(errno));
return -1;
case 0:
/* Child process. */
- /* Use separate process group so child processes can be killed easily. */
+ /* Use separate process group so child processes can be killed
+ * easily. */
if (setpgid(0, 0) < 0)
zlog_warn("warning: setpgid(0,0) failed: %s",
safe_strerror(errno));
{
char shell[] = "sh";
char dashc[] = "-c";
- char *const argv[4] = { shell, dashc, shell_cmd, NULL };
+ char *const argv[4] = {shell, dashc, shell_cmd, NULL};
execv("/bin/sh", argv);
- zlog_err("execv(/bin/sh -c '%s') failed: %s",
- shell_cmd, safe_strerror(errno));
+ flog_err_sys(LIB_ERR_SYSTEM_CALL,
+ "execv(/bin/sh -c '%s') failed: %s",
+ shell_cmd, safe_strerror(errno));
_exit(127);
}
default:
/* Parent process: we will reap the child later. */
- zlog_err("Forked background command [pid %d]: %s", (int)child,
- shell_cmd);
+ flog_err_sys(LIB_ERR_SYSTEM_CALL,
+ "Forked background command [pid %d]: %s",
+ (int)child, shell_cmd);
return child;
}
}
struct timeval delay;
time_elapsed(&delay, &restart->time);
- zlog_warn("Warning: %s %s child process %d still running after "
- "%ld seconds, sending signal %d",
- restart->what, restart->name, (int)restart->pid,
- (long)delay.tv_sec, (restart->kills ? SIGKILL : SIGTERM));
+ zlog_warn(
+ "Warning: %s %s child process %d still running after "
+ "%ld seconds, sending signal %d",
+ restart->what, restart->name, (int)restart->pid,
+ (long)delay.tv_sec, (restart->kills ? SIGKILL : SIGTERM));
kill(-restart->pid, (restart->kills ? SIGKILL : SIGTERM));
restart->kills++;
restart->t_kill = NULL;
static struct restart_info *find_child(pid_t child)
{
- if (gs.mode == MODE_GLOBAL_RESTART) {
- if (gs.restart.pid == child)
- return &gs.restart;
- } else {
- struct daemon *dmn;
- for (dmn = gs.daemons; dmn; dmn = dmn->next) {
- if (dmn->restart.pid == child)
- return &dmn->restart;
- }
+ struct daemon *dmn;
+ for (dmn = gs.daemons; dmn; dmn = dmn->next) {
+ if (dmn->restart.pid == child)
+ return &dmn->restart;
}
return NULL;
}
switch (child = waitpid(-1, &status, WNOHANG)) {
case -1:
- zlog_err("waitpid failed: %s", safe_strerror(errno));
+ flog_err_sys(LIB_ERR_SYSTEM_CALL, "waitpid failed: %s",
+ safe_strerror(errno));
return;
case 0:
zlog_warn("SIGCHLD received, but waitpid did not reap a child");
gs.numpids--;
thread_cancel(restart->t_kill);
restart->t_kill = NULL;
- /* Update restart time to reflect the time the command completed. */
+ /* Update restart time to reflect the time the command
+ * completed. */
gettimeofday(&restart->time, NULL);
} else {
- zlog_err
- ("waitpid returned status for an unknown child process %d",
- (int)child);
+ flog_err_sys(
+ LIB_ERR_SYSTEM_CALL,
+ "waitpid returned status for an unknown child process %d",
+ (int)child);
name = "(unknown)";
what = "background";
}
if (WIFSTOPPED(status))
- zlog_warn("warning: %s %s process %d is stopped",
- what, name, (int)child);
+ zlog_warn("warning: %s %s process %d is stopped", what, name,
+ (int)child);
else if (WIFSIGNALED(status))
- zlog_warn("%s %s process %d terminated due to signal %d",
- what, name, (int)child, WTERMSIG(status));
+ zlog_warn("%s %s process %d terminated due to signal %d", what,
+ name, (int)child, WTERMSIG(status));
else if (WIFEXITED(status)) {
if (WEXITSTATUS(status) != 0)
- zlog_warn
- ("%s %s process %d exited with non-zero status %d",
- what, name, (int)child, WEXITSTATUS(status));
+ zlog_warn(
+ "%s %s process %d exited with non-zero status %d",
+ what, name, (int)child, WEXITSTATUS(status));
else
zlog_debug("%s %s process %d exited normally", what,
name, (int)child);
} else
- zlog_err("cannot interpret %s %s process %d wait status 0x%x",
- what, name, (int)child, status);
+ flog_err_sys(
+ LIB_ERR_SYSTEM_CALL,
+ "cannot interpret %s %s process %d wait status 0x%x",
+ what, name, (int)child, status);
phase_check();
}
-static int
-run_job(struct restart_info *restart, const char *cmdtype, const char *command,
- int force, int update_interval)
+static int run_job(struct restart_info *restart, const char *cmdtype,
+ const char *command, int force, int update_interval)
{
struct timeval delay;
if (restart->pid) {
if (gs.loglevel > LOG_DEBUG + 1)
- zlog_debug
- ("cannot %s %s, previous pid %d still running",
- cmdtype, restart->name, (int)restart->pid);
+ zlog_debug(
+ "cannot %s %s, previous pid %d still running",
+ cmdtype, restart->name, (int)restart->pid);
return -1;
}
- /* Note: time_elapsed test must come before the force test, since we need
+ /* Note: time_elapsed test must come before the force test, since we
+ need
to make sure that delay is initialized for use below in updating the
restart interval. */
if ((time_elapsed(&delay, &restart->time)->tv_sec < restart->interval)
&& !force) {
if (gs.loglevel > LOG_DEBUG + 1)
- zlog_debug("postponing %s %s: "
- "elapsed time %ld < retry interval %ld",
- cmdtype, restart->name, (long)delay.tv_sec,
- restart->interval);
+ zlog_debug(
+ "postponing %s %s: "
+ "elapsed time %ld < retry interval %ld",
+ cmdtype, restart->name, (long)delay.tv_sec,
+ restart->interval);
return -1;
}
snprintf(cmd, sizeof(cmd), command, restart->name);
if ((restart->pid = run_background(cmd)) > 0) {
restart->t_kill = NULL;
- thread_add_timer(master, restart_kill, restart, gs.restart_timeout,
- &restart->t_kill);
+ thread_add_timer(master, restart_kill, restart,
+ gs.restart_timeout, &restart->t_kill);
restart->what = cmdtype;
gs.numpids++;
} else
return restart->pid;
}
-#define SET_READ_HANDLER(DMN) \
- do { \
- (DMN)->t_read = NULL; \
- thread_add_read (master, handle_read, (DMN), (DMN)->fd, &(DMN)->t_read); \
- } while (0);
-
-#define SET_WAKEUP_DOWN(DMN) \
- do { \
- (DMN)->t_wakeup = NULL; \
- thread_add_timer_msec (master, wakeup_down, (DMN), FUZZY(gs.period), \
- &(DMN)->t_wakeup); \
- } while (0);
-
-#define SET_WAKEUP_UNRESPONSIVE(DMN) \
- do { \
- (DMN)->t_wakeup = NULL; \
- thread_add_timer_msec (master, wakeup_unresponsive, (DMN), \
- FUZZY(gs.period), &(DMN)->t_wakeup); \
- } while (0);
-
-#define SET_WAKEUP_ECHO(DMN) \
- do { \
- (DMN)->t_wakeup = NULL; \
- thread_add_timer_msec (master, wakeup_send_echo, (DMN), \
- FUZZY(gs.period), &(DMN)->t_wakeup); \
- } while (0);
+#define SET_READ_HANDLER(DMN) \
+ do { \
+ (DMN)->t_read = NULL; \
+ thread_add_read(master, handle_read, (DMN), (DMN)->fd, \
+ &(DMN)->t_read); \
+ } while (0);
+
+#define SET_WAKEUP_DOWN(DMN) \
+ do { \
+ (DMN)->t_wakeup = NULL; \
+ thread_add_timer_msec(master, wakeup_down, (DMN), \
+ FUZZY(gs.period), &(DMN)->t_wakeup); \
+ } while (0);
+
+#define SET_WAKEUP_UNRESPONSIVE(DMN) \
+ do { \
+ (DMN)->t_wakeup = NULL; \
+ thread_add_timer_msec(master, wakeup_unresponsive, (DMN), \
+ FUZZY(gs.period), &(DMN)->t_wakeup); \
+ } while (0);
+
+#define SET_WAKEUP_ECHO(DMN) \
+ do { \
+ (DMN)->t_wakeup = NULL; \
+ thread_add_timer_msec(master, wakeup_send_echo, (DMN), \
+ FUZZY(gs.period), &(DMN)->t_wakeup); \
+ } while (0);
static int wakeup_down(struct thread *t_wakeup)
{
dmn->t_wakeup = NULL;
if (try_connect(dmn) < 0) {
SET_WAKEUP_DOWN(dmn);
- zlog_err("%s state -> down : initial connection attempt failed",
- dmn->name);
+ flog_err(WATCHFRR_ERR_CONNECTION,
+ "%s state -> down : initial connection attempt failed",
+ dmn->name);
dmn->state = DAEMON_DOWN;
}
return 0;
static void daemon_down(struct daemon *dmn, const char *why)
{
if (IS_UP(dmn) || (dmn->state == DAEMON_INIT))
- zlog_err("%s state -> down : %s", dmn->name, why);
+ flog_err(WATCHFRR_ERR_CONNECTION,
+ "%s state -> down : %s", dmn->name, why);
else if (gs.loglevel > LOG_DEBUG)
zlog_debug("%s still down : %s", dmn->name, why);
if (IS_UP(dmn))
char why[100 + sizeof(buf)];
snprintf(why, sizeof(why),
"read returned bad echo response of %d bytes "
- "(expecting %u): %.*s", (int)rc, (u_int) sizeof(resp),
- (int)rc, buf);
+ "(expecting %u): %.*s",
+ (int)rc, (unsigned int)sizeof(resp), (int)rc, buf);
daemon_down(dmn, why);
return 0;
}
if (dmn->state == DAEMON_UNRESPONSIVE) {
if (delay.tv_sec < gs.timeout) {
dmn->state = DAEMON_UP;
- zlog_warn
- ("%s state -> up : echo response received after %ld.%06ld "
- "seconds", dmn->name, (long)delay.tv_sec,
- (long)delay.tv_usec);
+ zlog_warn(
+ "%s state -> up : echo response received after %ld.%06ld "
+ "seconds",
+ dmn->name, (long)delay.tv_sec,
+ (long)delay.tv_usec);
} else
- zlog_warn
- ("%s: slow echo response finally received after %ld.%06ld "
- "seconds", dmn->name, (long)delay.tv_sec,
- (long)delay.tv_usec);
+ zlog_warn(
+ "%s: slow echo response finally received after %ld.%06ld "
+ "seconds",
+ dmn->name, (long)delay.tv_sec,
+ (long)delay.tv_usec);
} else if (gs.loglevel > LOG_DEBUG + 1)
zlog_debug("%s: echo response received after %ld.%06ld seconds",
dmn->name, (long)delay.tv_sec, (long)delay.tv_usec);
FILE *fp;
fp = fopen(DAEMON_VTY_DIR "/watchfrr.started", "w");
- fclose(fp);
- zlog_notice
- ("Watchfrr: Notifying Systemd we are up and running");
+ if (fp)
+ fclose(fp);
+#if defined HAVE_SYSTEMD
+ zlog_notice(
+ "Watchfrr: Notifying Systemd we are up and running");
systemd_send_started(master, 0);
+#endif
sent = 1;
}
}
dmn->connect_tries = 0;
zlog_notice("%s state -> up : %s", dmn->name, why);
daemon_send_ready();
- if (gs.do_ping)
- SET_WAKEUP_ECHO(dmn);
+ SET_WAKEUP_ECHO(dmn);
phase_check();
}
}
if ((reslen == sizeof(sockerr)) && sockerr) {
char why[100];
- snprintf(why, sizeof(why),
- "getsockopt reports that connection attempt failed: %s",
- safe_strerror(sockerr));
+ snprintf(
+ why, sizeof(why),
+ "getsockopt reports that connection attempt failed: %s",
+ safe_strerror(sockerr));
daemon_down(dmn, why);
return 0;
}
memset(&addr, 0, sizeof(struct sockaddr_un));
addr.sun_family = AF_UNIX;
- snprintf(addr.sun_path, sizeof(addr.sun_path), "%s/%s.vty",
- gs.vtydir, dmn->name);
+ snprintf(addr.sun_path, sizeof(addr.sun_path), "%s/%s.vty", gs.vtydir,
+ dmn->name);
#ifdef HAVE_STRUCT_SOCKADDR_UN_SUN_LEN
len = addr.sun_len = SUN_LEN(&addr);
#else
len = sizeof(addr.sun_family) + strlen(addr.sun_path);
-#endif /* HAVE_STRUCT_SOCKADDR_UN_SUN_LEN */
+#endif /* HAVE_STRUCT_SOCKADDR_UN_SUN_LEN */
/* Quick check to see if we might succeed before we go to the trouble
of creating a socket. */
if (access(addr.sun_path, W_OK) < 0) {
if (errno != ENOENT)
- zlog_err("%s: access to socket %s denied: %s",
- dmn->name, addr.sun_path,
- safe_strerror(errno));
+ flog_err_sys(LIB_ERR_SYSTEM_CALL,
+ "%s: access to socket %s denied: %s",
+ dmn->name, addr.sun_path,
+ safe_strerror(errno));
return -1;
}
if ((sock = socket(AF_UNIX, SOCK_STREAM, 0)) < 0) {
- zlog_err("%s(%s): cannot make socket: %s",
- __func__, addr.sun_path, safe_strerror(errno));
+ flog_err_sys(LIB_ERR_SOCKET, "%s(%s): cannot make socket: %s",
+ __func__, addr.sun_path, safe_strerror(errno));
return -1;
}
if (set_nonblocking(sock) < 0 || set_cloexec(sock) < 0) {
- zlog_err("%s(%s): set_nonblocking/cloexec(%d) failed",
- __func__, addr.sun_path, sock);
+ flog_err_sys(LIB_ERR_SYSTEM_CALL,
+ "%s(%s): set_nonblocking/cloexec(%d) failed",
+ __func__, addr.sun_path, sock);
close(sock);
return -1;
}
dmn->fd = sock;
dmn->t_write = NULL;
thread_add_write(master, check_connect, dmn, dmn->fd,
- &dmn->t_write);dmn->t_wakeup = NULL;
- thread_add_timer(master, wakeup_connect_hanging, dmn, gs.timeout,
- &dmn->t_wakeup);
+ &dmn->t_write);
+ dmn->t_wakeup = NULL;
+ thread_add_timer(master, wakeup_connect_hanging, dmn,
+ gs.timeout, &dmn->t_wakeup);
SET_READ_HANDLER(dmn);
return 0;
}
static int phase_hanging(struct thread *t_hanging)
{
gs.t_phase_hanging = NULL;
- zlog_err("Phase [%s] hanging for %ld seconds, aborting phased restart",
- phase_str[gs.phase], PHASE_TIMEOUT);
+ flog_err(WATCHFRR_ERR_CONNECTION,
+ "Phase [%s] hanging for %ld seconds, aborting phased restart",
+ phase_str[gs.phase], PHASE_TIMEOUT);
gs.phase = PHASE_NONE;
return 0;
}
case PHASE_STOPS_PENDING:
if (gs.numpids)
break;
- zlog_info
- ("Phased restart: all routing daemon stop jobs have completed.");
+ zlog_info(
+ "Phased restart: all routing daemon stop jobs have completed.");
set_phase(PHASE_WAITING_DOWN);
- /*FALLTHRU*/
+ /*FALLTHRU*/
case PHASE_WAITING_DOWN:
if (gs.numdown + IS_UP(gs.special) < gs.numdaemons)
break;
1);
set_phase(PHASE_ZEBRA_RESTART_PENDING);
- /*FALLTHRU*/
+ /*FALLTHRU*/
case PHASE_ZEBRA_RESTART_PENDING:
if (gs.special->restart.pid)
break;
gs.special->name);
set_phase(PHASE_WAITING_ZEBRA_UP);
- /*FALLTHRU*/
+ /*FALLTHRU*/
case PHASE_WAITING_ZEBRA_UP:
if (!IS_UP(gs.special))
break;
static void try_restart(struct daemon *dmn)
{
- switch (gs.mode) {
- case MODE_MONITOR:
+ if (watch_only)
return;
- case MODE_GLOBAL_RESTART:
- run_job(&gs.restart, "restart", gs.restart_command, 0, 1);
- break;
- case MODE_SEPARATE_RESTART:
- run_job(&dmn->restart, "restart", gs.restart_command, 0, 1);
- break;
- case MODE_PHASED_ZEBRA_RESTART:
- if (dmn != gs.special) {
- if ((gs.special->state == DAEMON_UP)
- && (gs.phase == PHASE_NONE))
- run_job(&dmn->restart, "restart",
- gs.restart_command, 0, 1);
- else
- zlog_debug
- ("%s: postponing restart attempt because master %s daemon "
- "not up [%s], or phased restart in progress",
- dmn->name, gs.special->name,
- state_str[gs.special->state]);
- break;
- }
- /*FALLTHRU*/
- case MODE_PHASED_ALL_RESTART:
- if ((gs.phase != PHASE_NONE) || gs.numpids) {
+ if (dmn != gs.special) {
+ if ((gs.special->state == DAEMON_UP)
+ && (gs.phase == PHASE_NONE))
+ run_job(&dmn->restart, "restart", gs.restart_command, 0,
+ 1);
+ else
+ zlog_debug(
+ "%s: postponing restart attempt because master %s daemon "
+ "not up [%s], or phased restart in progress",
+ dmn->name, gs.special->name,
+ state_str[gs.special->state]);
+ return;
+ }
+
+ if ((gs.phase != PHASE_NONE) || gs.numpids) {
+ if (gs.loglevel > LOG_DEBUG + 1)
+ zlog_debug(
+ "postponing phased global restart: restart already in "
+ "progress [%s], or outstanding child processes [%d]",
+ phase_str[gs.phase], gs.numpids);
+ return;
+ }
+ /* Is it too soon for a restart? */
+ {
+ struct timeval delay;
+ if (time_elapsed(&delay, &gs.special->restart.time)->tv_sec
+ < gs.special->restart.interval) {
if (gs.loglevel > LOG_DEBUG + 1)
- zlog_debug
- ("postponing phased global restart: restart already in "
- "progress [%s], or outstanding child processes [%d]",
- phase_str[gs.phase], gs.numpids);
- break;
- }
- /* Is it too soon for a restart? */
- {
- struct timeval delay;
- if (time_elapsed(&delay, &gs.special->restart.time)->
- tv_sec < gs.special->restart.interval) {
- if (gs.loglevel > LOG_DEBUG + 1)
- zlog_debug
- ("postponing phased global restart: "
- "elapsed time %ld < retry interval %ld",
- (long)delay.tv_sec,
- gs.special->restart.interval);
- break;
- }
+ zlog_debug(
+ "postponing phased global restart: "
+ "elapsed time %ld < retry interval %ld",
+ (long)delay.tv_sec,
+ gs.special->restart.interval);
+ return;
}
- run_job(&gs.restart, "restart", gs.restart_command, 0, 1);
- break;
- default:
- zlog_err("error: unknown restart mode %d", gs.mode);
- break;
}
+ run_job(&gs.restart, "restart", gs.restart_command, 0, 1);
}
static int wakeup_unresponsive(struct thread *t_wakeup)
dmn->t_wakeup = NULL;
if (dmn->state != DAEMON_UNRESPONSIVE)
- zlog_err("%s: no longer unresponsive (now %s), "
- "wakeup should have been cancelled!",
- dmn->name, state_str[dmn->state]);
+ flog_err(WATCHFRR_ERR_CONNECTION,
+ "%s: no longer unresponsive (now %s), "
+ "wakeup should have been cancelled!",
+ dmn->name, state_str[dmn->state]);
else {
SET_WAKEUP_UNRESPONSIVE(dmn);
try_restart(dmn);
dmn->t_wakeup = NULL;
dmn->state = DAEMON_UNRESPONSIVE;
- zlog_err("%s state -> unresponsive : no response yet to ping "
- "sent %ld seconds ago", dmn->name, gs.timeout);
- if (gs.unresponsive_restart) {
- SET_WAKEUP_UNRESPONSIVE(dmn);
- try_restart(dmn);
- }
+ flog_err(WATCHFRR_ERR_CONNECTION,
+ "%s state -> unresponsive : no response yet to ping "
+ "sent %ld seconds ago",
+ dmn->name, gs.timeout);
+ SET_WAKEUP_UNRESPONSIVE(dmn);
+ try_restart(dmn);
return 0;
}
struct daemon *dmn = THREAD_ARG(t_wakeup);
dmn->t_wakeup = NULL;
- if (((rc = write(dmn->fd, echocmd, sizeof(echocmd))) < 0) ||
- ((size_t) rc != sizeof(echocmd))) {
+ if (((rc = write(dmn->fd, echocmd, sizeof(echocmd))) < 0)
+ || ((size_t)rc != sizeof(echocmd))) {
char why[100 + sizeof(echocmd)];
snprintf(why, sizeof(why),
"write '%s' returned %d instead of %u", echocmd,
- (int)rc, (u_int) sizeof(echocmd));
+ (int)rc, (unsigned int)sizeof(echocmd));
daemon_down(dmn, why);
} else {
gettimeofday(&dmn->echo_sent, NULL);
return 0;
}
+bool check_all_up(void)
+{
+ struct daemon *dmn;
+
+ for (dmn = gs.daemons; dmn; dmn = dmn->next)
+ if (dmn->state != DAEMON_UP)
+ return false;
+ return true;
+}
+
static void sigint(void)
{
zlog_notice("Terminating on signal");
char *p;
return ((p = strchr(cmd, '%')) != NULL) && (*(p + 1) == 's')
- && !strchr(p + 1, '%');
+ && !strchr(p + 1, '%');
}
/* This is an ugly hack to circumvent problems with passing command-line
};
FRR_DAEMON_INFO(watchfrr, WATCHFRR,
- .flags = FRR_NO_PRIVSEP | FRR_NO_TCPVTY | FRR_LIMITED_CLI
- | FRR_NO_CFG_PID_DRY | FRR_NO_ZCLIENT,
+ .flags = FRR_NO_PRIVSEP | FRR_NO_TCPVTY | FRR_LIMITED_CLI
+ | FRR_NO_CFG_PID_DRY | FRR_NO_ZCLIENT,
- .printhelp = printhelp,
- .copyright = "Copyright 2004 Andrew J. Schorr",
+ .printhelp = printhelp,
+ .copyright = "Copyright 2004 Andrew J. Schorr",
- .signals = watchfrr_signals,
- .n_signals = array_size(watchfrr_signals),
+ .signals = watchfrr_signals,
+ .n_signals = array_size(watchfrr_signals),
- .privs = &watchfrr_privs,
-)
+ .privs = &watchfrr_privs, )
+
+#define DEPRECATED_OPTIONS "aAezR:"
int main(int argc, char **argv)
{
int opt;
- const char *pidfile = DEFAULT_PIDFILE;
+ const char *pidfile = pidfile_default;
const char *special = "zebra";
const char *blankstr = NULL;
+ snprintf(pidfile_default, sizeof(pidfile_default), "%s/watchfrr.pid",
+ frr_vtydir);
+
frr_preinit(&watchfrr_di, argc, argv);
progname = watchfrr_di.progname;
- frr_opt_add("aAb:dek:l:i:p:r:R:S:s:t:T:z", longopts, "");
+ frr_opt_add("b:dk:l:i:p:r:S:s:t:T:" DEPRECATED_OPTIONS, longopts, "");
gs.restart.name = "all";
while ((opt = frr_getopt(argc, argv, NULL)) != EOF) {
+ if (opt && opt < 128 && strchr(DEPRECATED_OPTIONS, opt)) {
+ fprintf(stderr,
+ "The -%c option no longer exists.\n"
+ "Please refer to the watchfrr(8) man page.\n",
+ opt);
+ exit(1);
+ }
+
switch (opt) {
case 0:
break;
- case 'a':
- if ((gs.mode != MODE_MONITOR)
- && (gs.mode != MODE_SEPARATE_RESTART)) {
- fputs("Ambiguous operating mode selected.\n",
- stderr);
- frr_help_exit(1);
- }
- gs.mode = MODE_PHASED_ZEBRA_RESTART;
- break;
- case 'A':
- if ((gs.mode != MODE_MONITOR)
- && (gs.mode != MODE_SEPARATE_RESTART)) {
- fputs("Ambiguous operating mode selected.\n",
- stderr);
- frr_help_exit(1);
- }
- gs.mode = MODE_PHASED_ALL_RESTART;
- break;
case 'b':
blankstr = optarg;
break;
- case 'e':
- gs.do_ping = 0;
+ case OPTION_DRY:
+ watch_only = true;
break;
case 'k':
if (!valid_command(optarg)) {
}
gs.stop_command = optarg;
break;
- case 'l':
- {
- char garbage[3];
- if ((sscanf
- (optarg, "%d%1s", &gs.loglevel,
- garbage) != 1)
- || (gs.loglevel < LOG_EMERG)) {
- fprintf(stderr,
- "Invalid loglevel argument: %s\n",
- optarg);
- frr_help_exit(1);
- }
+ case 'l': {
+ char garbage[3];
+ if ((sscanf(optarg, "%d%1s", &gs.loglevel, garbage)
+ != 1)
+ || (gs.loglevel < LOG_EMERG)) {
+ fprintf(stderr,
+ "Invalid loglevel argument: %s\n",
+ optarg);
+ frr_help_exit(1);
}
- break;
- case OPTION_MINRESTART:
- {
- char garbage[3];
- if ((sscanf(optarg, "%ld%1s",
- &gs.min_restart_interval,
- garbage) != 1)
- || (gs.min_restart_interval < 0)) {
- fprintf(stderr,
- "Invalid min_restart_interval argument: %s\n",
- optarg);
- frr_help_exit(1);
- }
+ } break;
+ case OPTION_MINRESTART: {
+ char garbage[3];
+ if ((sscanf(optarg, "%ld%1s", &gs.min_restart_interval,
+ garbage)
+ != 1)
+ || (gs.min_restart_interval < 0)) {
+ fprintf(stderr,
+ "Invalid min_restart_interval argument: %s\n",
+ optarg);
+ frr_help_exit(1);
}
- break;
- case OPTION_MAXRESTART:
- {
- char garbage[3];
- if ((sscanf(optarg, "%ld%1s",
- &gs.max_restart_interval,
- garbage) != 1)
- || (gs.max_restart_interval < 0)) {
- fprintf(stderr,
- "Invalid max_restart_interval argument: %s\n",
- optarg);
- frr_help_exit(1);
- }
+ } break;
+ case OPTION_MAXRESTART: {
+ char garbage[3];
+ if ((sscanf(optarg, "%ld%1s", &gs.max_restart_interval,
+ garbage)
+ != 1)
+ || (gs.max_restart_interval < 0)) {
+ fprintf(stderr,
+ "Invalid max_restart_interval argument: %s\n",
+ optarg);
+ frr_help_exit(1);
}
- break;
- case 'i':
- {
- char garbage[3];
- int period;
- if ((sscanf(optarg, "%d%1s", &period, garbage)
- != 1) || (gs.period < 1)) {
- fprintf(stderr,
- "Invalid interval argument: %s\n",
- optarg);
- frr_help_exit(1);
- }
- gs.period = 1000 * period;
+ } break;
+ case 'i': {
+ char garbage[3];
+ int period;
+ if ((sscanf(optarg, "%d%1s", &period, garbage) != 1)
+ || (gs.period < 1)) {
+ fprintf(stderr,
+ "Invalid interval argument: %s\n",
+ optarg);
+ frr_help_exit(1);
}
- break;
+ gs.period = 1000 * period;
+ } break;
case 'p':
pidfile = optarg;
break;
case 'r':
- if ((gs.mode == MODE_GLOBAL_RESTART) ||
- (gs.mode == MODE_SEPARATE_RESTART)) {
- fputs("Ambiguous operating mode selected.\n",
- stderr);
- frr_help_exit(1);
- }
if (!valid_command(optarg)) {
fprintf(stderr,
"Invalid restart command, must contain '%%s': %s\n",
frr_help_exit(1);
}
gs.restart_command = optarg;
- if (gs.mode == MODE_MONITOR)
- gs.mode = MODE_SEPARATE_RESTART;
- break;
- case 'R':
- if (gs.mode != MODE_MONITOR) {
- fputs("Ambiguous operating mode selected.\n",
- stderr);
- frr_help_exit(1);
- }
- if (strchr(optarg, '%')) {
- fprintf(stderr,
- "Invalid restart-all arg, must not contain '%%s': %s\n",
- optarg);
- frr_help_exit(1);
- }
- gs.restart_command = optarg;
- gs.mode = MODE_GLOBAL_RESTART;
break;
case 's':
if (!valid_command(optarg)) {
case 'S':
gs.vtydir = optarg;
break;
- case 't':
- {
- char garbage[3];
- if ((sscanf
- (optarg, "%ld%1s", &gs.timeout,
- garbage) != 1) || (gs.timeout < 1)) {
- fprintf(stderr,
- "Invalid timeout argument: %s\n",
- optarg);
- frr_help_exit(1);
- }
+ case 't': {
+ char garbage[3];
+ if ((sscanf(optarg, "%ld%1s", &gs.timeout, garbage)
+ != 1)
+ || (gs.timeout < 1)) {
+ fprintf(stderr,
+ "Invalid timeout argument: %s\n",
+ optarg);
+ frr_help_exit(1);
}
- break;
- case 'T':
- {
- char garbage[3];
- if ((sscanf
- (optarg, "%ld%1s", &gs.restart_timeout,
- garbage) != 1)
- || (gs.restart_timeout < 1)) {
- fprintf(stderr,
- "Invalid restart timeout argument: %s\n",
- optarg);
- frr_help_exit(1);
- }
+ } break;
+ case 'T': {
+ char garbage[3];
+ if ((sscanf(optarg, "%ld%1s", &gs.restart_timeout,
+ garbage)
+ != 1)
+ || (gs.restart_timeout < 1)) {
+ fprintf(stderr,
+ "Invalid restart timeout argument: %s\n",
+ optarg);
+ frr_help_exit(1);
}
- break;
- case 'z':
- gs.unresponsive_restart = 1;
- break;
+ } break;
default:
fputs("Invalid option.\n", stderr);
frr_help_exit(1);
}
}
- if (gs.unresponsive_restart && (gs.mode == MODE_MONITOR)) {
- fputs("Option -z requires a -r or -R restart option.\n",
+ if (watch_only
+ && (gs.start_command || gs.stop_command || gs.restart_command)) {
+ fputs("Options -r/-s/-k are not used when --dry is active.\n",
stderr);
- frr_help_exit(1);
}
- switch (gs.mode) {
- case MODE_MONITOR:
- if (gs.restart_command || gs.start_command || gs.stop_command) {
- fprintf(stderr,
- "No kill/(re)start commands needed for %s mode.\n",
- mode_str[gs.mode]);
- frr_help_exit(1);
- }
- break;
- case MODE_GLOBAL_RESTART:
- case MODE_SEPARATE_RESTART:
- if (!gs.restart_command || gs.start_command || gs.stop_command) {
- fprintf(stderr,
- "No start/kill commands needed in [%s] mode.\n",
- mode_str[gs.mode]);
- frr_help_exit(1);
- }
- break;
- case MODE_PHASED_ZEBRA_RESTART:
- case MODE_PHASED_ALL_RESTART:
- if (!gs.restart_command || !gs.start_command
- || !gs.stop_command) {
- fprintf(stderr,
- "Need start, kill, and restart commands in [%s] mode.\n",
- mode_str[gs.mode]);
- frr_help_exit(1);
- }
- break;
+ if (!watch_only
+ && (!gs.restart_command || !gs.start_command || !gs.stop_command)) {
+ fprintf(stderr,
+ "Options -s (start), -k (kill), and -r (restart) are required.\n");
+ frr_help_exit(1);
}
if (blankstr) {
if (gs.restart_command)
gs.restart_command =
- translate_blanks(gs.restart_command, blankstr);
+ translate_blanks(gs.restart_command, blankstr);
if (gs.start_command)
gs.start_command =
- translate_blanks(gs.start_command, blankstr);
+ translate_blanks(gs.start_command, blankstr);
if (gs.stop_command)
gs.stop_command =
- translate_blanks(gs.stop_command, blankstr);
+ translate_blanks(gs.stop_command, blankstr);
}
gs.restart.interval = gs.min_restart_interval;
master = frr_init();
+ watchfrr_error_init();
zlog_set_level(ZLOG_DEST_MONITOR, ZLOG_DISABLED);
if (watchfrr_di.daemon_mode) {
zlog_set_level(ZLOG_DEST_SYSLOG, MIN(gs.loglevel, LOG_DEBUG));
- if (daemon (0, 0) < 0) {
- fprintf(stderr, "Watchquagga daemon failed: %s",
- strerror(errno));
- exit (1);
+ if (daemon(0, 0) < 0) {
+ fprintf(stderr, "Watchfrr daemon failed: %s",
+ strerror(errno));
+ exit(1);
}
} else
zlog_set_level(ZLOG_DEST_STDOUT, MIN(gs.loglevel, LOG_DEBUG));
if (!(dmn = (struct daemon *)calloc(1, sizeof(*dmn)))) {
fprintf(stderr, "calloc(1,%u) failed: %s\n",
- (u_int) sizeof(*dmn),
+ (unsigned int)sizeof(*dmn),
safe_strerror(errno));
return 1;
}
gs.numdown++;
dmn->fd = -1;
dmn->t_wakeup = NULL;
- thread_add_timer_msec(master, wakeup_init, dmn, 100 + (random() % 900),
+ thread_add_timer_msec(master, wakeup_init, dmn,
+ 100 + (random() % 900),
&dmn->t_wakeup);
dmn->restart.interval = gs.min_restart_interval;
if (tail)
gs.daemons = dmn;
tail = dmn;
- if (((gs.mode == MODE_PHASED_ZEBRA_RESTART) ||
- (gs.mode == MODE_PHASED_ALL_RESTART)) &&
- !strcmp(dmn->name, special))
+ if (!strcmp(dmn->name, special))
gs.special = dmn;
}
}
fputs("Must specify one or more daemons to monitor.\n", stderr);
frr_help_exit(1);
}
- if (((gs.mode == MODE_PHASED_ZEBRA_RESTART) ||
- (gs.mode == MODE_PHASED_ALL_RESTART)) && !gs.special) {
- fprintf(stderr,
- "In mode [%s], but cannot find master daemon %s\n",
- mode_str[gs.mode], special);
+ if (!watch_only && !gs.special) {
+ fprintf(stderr, "\"%s\" daemon must be in daemon list\n",
+ special);
frr_help_exit(1);
}
strcpy(p, dmn->name);
p += strlen(p);
}
- zlog_notice("%s %s watching [%s], mode [%s]",
- progname, FRR_VERSION, buf,
- mode_str[gs.mode]);
+ zlog_notice("%s %s watching [%s]%s", progname,
+ FRR_VERSION, buf,
+ watch_only ? ", monitor mode" : "");
}
}