2 # ---------------------------------------------------------------
3 # Copyright (C) 2008-2013 Georgia Public Library Service
4 # Copyright (C) 2013 Equinox Software, Inc
5 # Bill Erickson <berick@esilibrary.com>
7 # This program is free software; you can redistribute it and/or
8 # modify it under the terms of the GNU General Public License
9 # as published by the Free Software Foundation; either version 2
10 # of the License, or (at your option) any later version.
12 # This program is distributed in the hope that it will be useful,
13 # but WITHOUT ANY WARRANTY; without even the implied warranty of
14 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 # GNU General Public License for more details.
16 # ---------------------------------------------------------------
17 use strict; use warnings;
19 use Net::Domain qw/hostfqdn/;
20 use POSIX qw/setsid :sys_wait_h/;
21 use OpenSRF::Utils::Logger q/$logger/;
23 use OpenSRF::Transport::PeerHandle;
24 use OpenSRF::Utils::SettingsClient;
25 use OpenSRF::Transport::Listener;
27 use OpenSRF::Utils::Config;
29 my $opt_service = undef;
30 my $opt_config = "@CONF_DIR@/opensrf_core.xml";
31 my $opt_pid_dir = "@PID_DIR@/run/opensrf";
32 my $opt_no_daemon = 0;
33 my $opt_settings_pause = 0;
34 my $opt_localhost = 0;
36 my $opt_shutdown_graceful = 0;
37 my $opt_shutdown_fast = 0;
38 my $opt_shutdown_immediate = 0;
39 my $opt_shutdown_graceful_all = 0;
40 my $opt_shutdown_fast_all = 0;
41 my $opt_shutdown_immediate_all = 0;
42 my $opt_kill_with_fire = 0;
43 my $opt_signal = ''; # signal name
44 my $opt_signal_all = 0;
45 my $opt_signal_timeout = 30;
49 my $opt_start_all = 0;
51 my $opt_restart_all = 0;
52 my $opt_start_services = 0;
53 my $opt_stop_services = 0;
54 my $opt_restart_services = 0;
55 my $opt_force_clean_process = 0;
56 my $opt_router_de_register = 0;
57 my $opt_router_de_register_all = 0;
58 my $opt_router_re_register = 0;
59 my $opt_router_re_register_all = 0;
61 my $opt_reload_all = 0;
63 my $opt_diagnostic = 0;
64 my $opt_ignore_orphans = 0;
68 my $hostname = $ENV{OSRF_HOSTNAME} || hostfqdn();
71 'service=s' => \$opt_service,
72 'config=s' => \$opt_config,
73 'pid-dir=s' => \$opt_pid_dir,
74 'no-daemon' => \$opt_no_daemon,
75 'settings-startup-pause=i' => \$opt_settings_pause,
76 'localhost' => \$opt_localhost,
78 'quiet' => \$opt_quiet,
79 'graceful-shutdown' => \$opt_shutdown_graceful,
80 'fast-shutdown' => \$opt_shutdown_fast,
81 'immediate-shutdown' => \$opt_shutdown_immediate,
82 'graceful-shutdown-all' => \$opt_shutdown_graceful_all,
83 'fast-shutdown-all' => \$opt_shutdown_fast_all,
84 'immediate-shutdown-all' => \$opt_shutdown_immediate_all,
85 'kill-with-fire' => \$opt_kill_with_fire,
86 'force-clean-process' => \$opt_force_clean_process,
87 'signal-timeout' => \$opt_signal_timeout,
88 'signal=s' => \$opt_signal,
89 'signal-all' => \$opt_signal_all,
90 'start' => \$opt_start,
92 'start-all' => \$opt_start_all,
93 'stop-all' => \$opt_stop_all,
94 'restart' => \$opt_restart,
95 'restart-all' => \$opt_restart_all,
96 'start-services' => \$opt_start_services,
97 'stop-services' => \$opt_stop_services,
98 'restart-services' => \$opt_restart_services,
99 'router-de-register' => \$opt_router_de_register,
100 'router-de-register-all' => \$opt_router_de_register_all,
101 'router-re-register' => \$opt_router_re_register,
102 'router-re-register-all' => \$opt_router_re_register_all,
103 'reload' => \$opt_reload,
104 'reload-all' => \$opt_reload_all,
105 'diagnostic' => \$opt_diagnostic,
106 'ignore-orphans' => \$opt_ignore_orphans
109 if ($opt_localhost) {
110 $hostname = 'localhost';
111 $ENV{OSRF_HOSTNAME} = $hostname;
114 my $C_COMMAND = "opensrf-c -c $opt_config -x opensrf -p $opt_pid_dir -h $hostname";
115 my $PY_COMMAND = "opensrf.py -f $opt_config -p $opt_pid_dir ". ($opt_localhost ? '-l' : '');
117 sub verify_services {
119 return 1 if $service and $service eq 'router';
120 my @services = (@perl_services, map {$_->{service}} @nonperl_services);
122 return 1 unless $service;
123 return 1 if grep { $_ eq $service } @services;
124 msg("$service is not configured to run on $hostname");
126 msg("No services are configured to run on $hostname");
128 msg("Perhaps you meant to use --localhost?") unless $opt_localhost;
136 my @pids = get_service_pids_from_file($service);
139 # no PID files exist. see if the service is running anyway
141 @pids = get_service_pids_from_ps($service);
143 msg("cannot signal $service : no pid file or running process");
148 for my $pid (@pids) {
149 if (kill($signal, $pid) == 0) { # no process was signaled.
150 msg("cannot signal $service: process $pid is not running");
151 my $pidfile = get_pid_file($service);
152 unlink $pidfile if $pidfile;
156 msg("sending $signal signal to pid=$pid $service");
162 # returns 2 if a process should have gone away but did not
163 # in the case of multiple PIDs (e.g. router), return the
164 # status of any failures, but not the successes.
167 my @pids = get_service_pids_from_file($service);
170 for my $pid (@pids) {
172 # to determine whether a process has died, we have to send
173 # a no-op signal to the PID and check the success of that signal
175 for my $i (1..$opt_signal_timeout) {
176 $sig_count = kill(0, $pid);
177 last unless $sig_count;
182 msg("timed out waiting on $service pid=$pid to die");
187 # cleanup successful. remove the PID file
188 my $pidfile = get_pid_file($service);
189 unlink $pidfile if $pidfile;
197 return "$opt_pid_dir/$service.pid";
200 # services usually only have 1 pid, but the router will have at least 2
201 sub get_service_pids_from_file {
203 my $pid_file = get_pid_file($service);
204 return () unless -e $pid_file;
205 my @pids = `cat $pid_file`;
206 s/^\s*|\n//g for @pids;
210 sub get_service_pids_from_ps {
213 my $ps = ($service eq 'router') ?
214 "ps ax | grep 'OpenSRF Router'" :
215 "ps ax | grep 'OpenSRF Listener \\[$service\\]'";
217 $ps .= " | grep -v grep | sed 's/^\\s*//' | cut -d' ' -f1";
219 s/^\s*|\n//g for @pids;
226 my $alive = do_init(1);
228 my @services = get_service_list_from_files(1);
231 @conf_services = (@perl_services,
232 map {$_->{service}} @nonperl_services);
233 push(@services, @conf_services);
238 for my $svc (@services) {
239 $len = length($svc) if length($svc) > $len;
243 for my $svc (sort keys %services) {
244 my @pf_pids = get_service_pids_from_file($svc);
245 my @ps_pids = get_service_pids_from_ps($svc);
246 my $svc_str = sprintf("%-${len}s ", $svc);
249 unless(@ps_pids or @pf_pids) {
250 msg("$svc_str is not running");
254 for my $pid (@ps_pids) {
257 my $str = "$svc_str [$pid] ";
258 my $times = `ps -o etime=,cputime= $pid`;
259 $times =~ s/^\s+|\s+$//g;
260 my @times = split(/ /, $times);
261 $str .= sprintf("uptime=%-11s cputime=%-11s ", $times[0], $times[1]);
263 if ($svc eq 'router') {
266 my @drones = `pgrep -f "Drone \\[$svc\\]"`;
267 $str .= "#drones=".scalar(@drones);
269 msg("\tERR $svc has no running drones.") unless @drones;
272 msg("\tERR $svc [$pid] NOT configured for this host.")
273 unless grep {$_ eq $svc} @conf_services
276 msg("\tERR $svc [$pid] NOT found in PID file.")
277 unless grep {$_ eq $pid} @pf_pids;
280 for my $pid (@pf_pids) {
282 msg("\tERR $svc Has PID file entry [$pid], ".
283 "which matches no running $svc processes");
290 sub do_start_router {
292 my $pidfile = get_pid_file('router');
293 `opensrf_router $opt_config routers $pidfile`;
295 sleep 2; # give the router time to fork (probably not need now but w/e)
298 # stop a specific service
300 my ($service, @signals) = @_;
301 @signals = qw/TERM INT KILL/ unless @signals;
302 for my $sig (@signals) {
303 last unless do_signal($service, $sig) == 2;
311 OpenSRF::System->bootstrap_client(config_file => $opt_config);
313 if (!OpenSRF::Transport::PeerHandle->retrieve) {
314 return 0 if $fail_ok;
315 die "Unable to bootstrap client for requests\n";
318 load_settings(); # load the settings config if we can
320 my $sclient = OpenSRF::Utils::SettingsClient->new;
321 my $apps = $sclient->config_value("activeapps", "appname");
323 # disconnect the top-level network handle
324 OpenSRF::Transport::PeerHandle->retrieve->disconnect;
327 $apps = [$apps] unless ref $apps;
328 for my $app (@$apps) {
329 if (!$sclient->config_value('apps', $app)) {
330 msg("Service '$app' is listed for this host, ".
331 "but there is no configuration for it in $opt_config");
334 my $lang = $sclient->config_value('apps', $app, 'language') || '';
335 if ($lang =~ /perl/i) {
336 push(@perl_services, $app);
338 push(@nonperl_services, {service => $app, lang => $lang});
345 # start a specific service
349 my @pf_pids = get_service_pids_from_file($service);
350 my @ps_pids = get_service_pids_from_ps($service);
352 if (@pf_pids) { # had pidfile
355 msg("service $service already running : @ps_pids");
358 } else { # stale pidfile
360 my $pidfile = get_pid_file($service);
361 msg("removing stale pid file $pidfile");
365 } elsif (@ps_pids and not $opt_ignore_orphans) { # orphan process
367 if ($opt_force_clean_process) {
368 msg("service $service pid=@ps_pids is running with no pidfile");
369 do_signal($service, 'KILL');
371 msg("service $service pid=@ps_pids is running with no pidfile! ".
372 "use --force-clean-process to automatically kill orphan processes");
377 return do_start_router() if $service eq 'router';
379 load_settings() if $service eq 'opensrf.settings';
381 if(grep { $_ eq $service } @perl_services) {
382 return unless do_daemon($service);
383 OpenSRF::System->run_service($service, $opt_pid_dir);
386 # note: we don't daemonize non-perl services, but instead
387 # assume the controller for other languages manages that.
388 my ($svc) = grep { $_->{service} eq $service } @nonperl_services;
390 if ($svc->{lang} =~ /c/i) {
391 system("$C_COMMAND -a start -s $service");
393 } elsif ($svc->{lang} =~ /python/i) {
394 system("$PY_COMMAND -a start -s $service");
400 # should not get here
406 msg("starting router and services for $hostname");
408 return do_start_services();
411 sub do_start_services {
412 msg("starting services for $hostname");
414 if(grep {$_ eq 'opensrf.settings'} @perl_services) {
415 do_start('opensrf.settings');
416 # in batch mode, give opensrf.settings plenty of time to start
417 # before any non-Perl services try to connect
418 sleep $opt_settings_pause if $opt_settings_pause;
421 # start Perl services
422 for my $service (@perl_services) {
423 do_start($service) unless $service eq 'opensrf.settings';
426 # start each non-perl service individually instead of using the native
427 # start-all command. this allows us to test for existing pid files
428 # and/or running processes on each service before starting.
429 # it also means each service has to connect-fetch_setting-disconnect
430 # from jabber, which makes startup slightly slower than native start-all
431 do_start($_->{service}) for @nonperl_services;
436 # signal a single service
440 return do_signal_all($signal, $service);
443 # returns the list of running services based on presence of PID files.
444 # the 'router' service is not included by deault, since it's
445 # usually treated special.
446 sub get_service_list_from_files {
447 my $include_router = shift;
448 my @services = `ls $opt_pid_dir/*.pid 2> /dev/null`;
449 s/^\s*|\n//g for @services;
450 s|.*/(.*)\.pid$|$1| for @services;
451 return @services if $include_router;
452 return grep { $_ ne 'router' } @services;
456 my ($signal, @services) = @_;
457 @services = get_service_list_from_files() unless @services;
459 do_signal_send($_, $signal) for @services;
461 # if user passed a know non-shutdown signal, we're done.
462 return if $signal =~ /HUP|USR1|USR2/;
464 do_signal_wait($_) for @services;
467 # pull all opensrf listener and drone PIDs from 'ps',
468 # kill them all, and remove all pid files
469 sub do_kill_with_fire {
470 msg("killing with fire");
472 my @pids = get_running_pids();
474 next unless $_ =~ /\d+/;
475 my $proc = `ps -p $_ -o cmd=`;
477 msg("killing with fire pid=$_ $proc");
481 # remove all of the pid files
482 my @files = `ls $opt_pid_dir/*.pid 2> /dev/null`;
483 s/^\s*|\n//g for @files;
485 msg("removing pid file $_");
490 sub get_running_pids {
493 # start with the listeners, then drones, then routers
495 "ps ax | grep 'OpenSRF Listener' ",
496 "ps ax | grep 'OpenSRF Drone' ",
497 "ps ax | grep 'OpenSRF Router' "
500 $_ .= "| grep -v grep | sed 's/^\\s*//' | cut -d' ' -f1" for @greps;
502 for my $grep (@greps) {
504 s/^\s*|\n//g for @spids;
505 push (@pids, @spids);
511 sub clear_stale_pids {
512 my @pidfile_services = get_service_list_from_files(1);
513 my @running_pids = get_running_pids();
515 for my $svc (@pidfile_services) {
516 my @pids = get_service_pids_from_file($svc);
517 for my $pid (@pids) {
518 next if grep { $_ eq $pid } @running_pids;
519 my $pidfile = get_pid_file($svc);
520 msg("removing stale pid file $pidfile");
526 sub do_stop_services {
528 @signals = qw/TERM INT KILL/ unless @signals;
530 msg("stopping services for $hostname");
531 my @services = get_service_list_from_files();
533 for my $signal (@signals) {
536 # send the signal to all PIDs
537 do_signal_send($_, $signal) for @services;
539 # then wait for them to go away
540 for my $service (@services) {
541 push(@redo, $service) if do_signal_wait($service) == 2;
545 last unless @services;
553 @signals = qw/TERM INT KILL/ unless @signals;
555 do_stop_services(@signals);
557 # graceful shutdown requires the presence of the router, so stop the
558 # router last. See if it's running first to avoid unnecessary warnings.
559 do_stop('router', $signals[0]) if get_service_pids_from_file('router');
564 # daemonize us. return true if we're the child, false if parent
566 return 1 if $opt_no_daemon;
568 my $pid_file = get_pid_file($service);
569 my $pid = OpenSRF::Utils::safe_fork();
571 msg("starting service pid=$pid $service");
579 open STDIN, '</dev/null';
580 open STDOUT, '>/dev/null';
581 open STDERR, '>/dev/null';
582 `echo $$ > $pid_file`;
586 # parses the local settings file
588 my $conf = OpenSRF::Utils::Config->current;
589 my $cfile = $conf->bootstrap->settings_config;
590 return unless $cfile;
591 my $parser = OpenSRF::Utils::SettingsParser->new();
592 $parser->initialize( $cfile );
593 $OpenSRF::Utils::SettingsClient::host_config =
594 $parser->get_server_config($conf->env->hostname);
599 print "* $m\n" unless $opt_quiet;
605 Usage: $0 --localhost --start-all
607 --config <file> [default: @CONF_DIR@/opensrf_core.xml]
608 OpenSRF configuration file
610 --pid-dir <dir> [default: @PID_DIR@/run/opensrf]
611 Directory where process-specific PID files are kept
613 --settings-startup-pause
614 How long to give the opensrf.settings server to start up when running
615 in batch mode (start_all). The purpose is to give plenty of time for
616 the settings server to be up and active before any non-Perl services
620 Force the hostname to be 'localhost', instead of the fully qualified
621 domain name for the machine.
624 Specifies which OpenSRF service to control
627 Do not print informational messages to STDOUT
630 Do not detach and run as a daemon process. Useful for debugging.
631 Only works for Perl services and only when starting a single service.
634 Print this help message
637 Print information about running services
639 ==== starting services =====
642 Start the router and all services
645 Start the service specified by --service
648 Start all services but do not start any routers
651 Restart the router and all services
654 Restart the service specified by --service
657 Restart all services but do not restart any routers
659 --force-clean-process
660 When starting a service, if a service process is already running
661 but no pidfile exists, kill the service process before starting
662 the new one. This applies to routers too.
665 When starting a service, if a service process is already running but
666 no pidfile exists, ignore the existing process and carry on starting
667 the new one (i.e., ignore orphans). This applies to routers too.
669 ==== stopping services =====
672 Stop the router and all services. Services are sent the TERM signal,
673 followed by the INT signal, followed by the KILL signal. With each
674 iteration, the script pauses up to --signal-timeout seconds waiting
675 for each process to die before sending the next signal.
678 Stop the service specified by --service. See also --stop-all.
679 If the requested service does not have a matching PID file, an
680 attempt to locate the PID via 'ps' will be made.
683 Stop all services but do not stop any routers. See also --stop-all.
685 --graceful-shutdown-all
686 Send TERM signal to all services + router
689 Send TERM signal to the service specified by --service
692 Send INT signal to all services + router
695 Send INT signal to the service specified by --service
697 --immediate-shutdown-all
698 Send KILL signal to all services + router
701 Send KILL signal to the service specified by --service
704 Send KILL signal to all running services + routers, regardless of
705 the presence of a PID file, and remove all PID files indiscriminately.
707 ==== signaling services =====
710 Send signal to all services
713 Name of signal to send. If --signal-all is not specified, the
714 signal will be sent to the service specified by --service.
717 Seconds to wait for a process to die after sending a shutdown signal.
718 All signals except HUP, USR1, and USR2 are assumed to be shutdown signals.
720 ==== special signals ====
723 --router-de-register-all
724 Sends a SIGUSR1 signal to the selected service(s), which causes each
725 service's listener process to send an "unregister" command to all
726 registered routers. The --all variant sends the signal to all
727 running listeners. The non-(--all) variant requires a --service.
730 --router-re-register-all
731 Sends a SIGUSR2 signal to the selected service(s), which causes each
732 service's listener process to send a "register" command to all
733 configured routers. The --all variant sends the signal to all
734 running listeners. The non-(--all) variant requires a --service.
738 Sends a SIGHUP signal to the selected service(s). SIGHUP causes
739 each listener process to reload its opensrf_core.xml config file
740 and gracefully re-launch drone processes. The -all variant sends
741 the signal to all services. The non-(-all) variant requires a
747 # we do not verify services for stop/signal actions, since those may
748 # legitimately be used against services not (or no longer) configured
749 # to run on the selected host.
750 do_init() and verify_services($opt_service) if
753 $opt_start_services or
756 $opt_restart_services) and (
757 not defined $opt_service or $opt_service ne 'router'
760 # starting services. do_init() handled above
761 do_start($opt_service) if $opt_start;
762 do_stop($opt_service) and do_start($opt_service) if $opt_restart;
763 do_start_all() if $opt_start_all;
764 do_start_services() if $opt_start_services;
765 do_stop_all() and do_start_all() if $opt_restart_all;
766 do_stop_services() and do_start_services() if $opt_restart_services;
769 do_stop($opt_service) if $opt_stop;
770 do_stop_all() if $opt_stop_all;
771 do_stop_services() if $opt_stop_services;
772 do_stop($opt_service, 'TERM') if $opt_shutdown_graceful;
773 do_stop($opt_service, 'INT') if $opt_shutdown_fast;
774 do_stop($opt_service, 'KILL') if $opt_shutdown_immediate;
775 do_stop_all('TERM') if $opt_shutdown_graceful_all;
776 do_stop_all('INT') if $opt_shutdown_fast_all;
777 do_stop_all('KILL') if $opt_shutdown_immediate_all;
778 do_kill_with_fire() if $opt_kill_with_fire;
781 $opt_signal = 'USR1' if $opt_router_de_register or $opt_router_de_register_all;
782 $opt_signal = 'USR2' if $opt_router_re_register or $opt_router_re_register_all;
783 $opt_signal = 'HUP' if $opt_reload or $opt_reload_all;
785 do_signal($opt_service, $opt_signal) if $opt_signal and $opt_service;
786 do_signal_all($opt_signal) if
789 $opt_router_de_register_all or
790 $opt_router_re_register_all;
793 do_diagnostic() if $opt_diagnostic;
796 # show help if no action was requested
797 do_help() if $opt_help or not (
800 $opt_start_services or
803 $opt_stop_services or
806 $opt_restart_services or
809 $opt_shutdown_graceful or
810 $opt_shutdown_graceful_all or
811 $opt_shutdown_fast or
812 $opt_shutdown_fast_all or
813 $opt_shutdown_immediate or
814 $opt_shutdown_immediate_all or
815 $opt_kill_with_fire or