Profile of /usr/bin/starman

Filename	/usr/bin/starman
Statements	Executed 0 statements in 0s

Subroutines
Calls	P	F	Exclusive Time	Inclusive Time	Subroutine
1	1	1	24µs	1.49ms	CGI::Compile::ROOT::home_vagrant_kohaclone_circ_ysearch_2epl::BEGIN@39

Call graph for these subroutines as a Graphviz dot language file.

Line	Calls	Time in subs	Code
0	1	1.49ms	Profile data that couldn't be associated with a specific line: # spent 1.49ms making 1 call to CGI::Compile::ROOT::home_vagrant_kohaclone_circ_ysearch_2epl::BEGIN@39
1			#!/usr/bin/perl
2			use strict;
3			use Plack::Runner;
4
5			sub version {
6			require Starman;
7			print "Starman $Starman::VERSION\n";
8			}
9
10			my $preload_app;
11
12			require Getopt::Long;
13			Getopt::Long::Configure("no_ignore_case", "no_auto_abbrev", "pass_through");
14			Getopt::Long::GetOptions(
15			"preload-app" => \$preload_app,
16			);
17
18			my @args = (server => 'Starman', env => 'deployment', version_cb => \&version);
19			if (!$preload_app) {
20			push @args, 'loader' => 'Delayed';
21			}
22
23			my @argv = @ARGV;
24
25			my $runner = Plack::Runner->new(@args);
26			$runner->parse_options(@argv);
27
28			if ($runner->{loader} eq 'Restarter') {
29			warn <<WARN;
30			!!
31			!! Using restarter with -r or -R options is known to cause issues with process supervisor
32			!! such as start_server, and it is STRONGLY discouraged to use with Starman. You have been warned.
33			!!
34			WARN
35			}
36
37			$runner->set_options(argv => \@argv);
38			$runner->run;
39			# spent 1.49ms (24µs+1.47) within CGI::Compile::ROOT::home_vagrant_kohaclone_circ_ysearch_2epl::BEGIN@39 which was called: # once (24µs+1.47ms) by CGI::Compile::ROOT::home_vagrant_kohaclone_circ_ysearch_2epl::CORE:binmode at line 0
40			__END__
41
42			=head1 NAME
43
44			starman - Starman launcher
45
46			=head1 SYNOPSIS
47
48			starman --listen :5001 --listen /tmp/starman.sock
49			starman --workers 32 --port 8080
50
51			=head1 OPTIONS
52
53			=over 4
54
55			=item -l, --listen
56
57			--listen HOST:PORT --listen :PORT --listen UNIX_SOCKET
58			--listen HOST:PORT:ssl
59
60			Specifies the TCP address, ports and UNIX domain sockets to bind to
61			wait for requests. You can repeat as many times as you want and mix
62			TCP and UNIX domain sockets.
63
64			For TCP sockets you can append C<:ssl> after the port to specify that
65			connections on that port should use SSL. Note that the SSL support is
66			experimental and hasn't been widely tested.
67
68			Defaults to any IP address and port 5000.
69
70			=item --host
71
72			--host 127.0.0.1
73
74			Specifies the address to bind.
75
76			This option is for a compatibility with L<plackup> and you're
77			recommended to use C<--listen> instead.
78
79			=item --port
80
81			--port 8080
82
83			Specifies the port to bind.
84
85			This option is for a compatibility with L<plackup> and you're
86			recommended to use C<--listen> instead.
87
88			=item -S, --socket
89
90			-S /tmp/starman.sock
91
92			Specifies the path to UNIX domain socket to bind.
93
94			This option is for a compatibility with L<plackup> and you're
95			recommended to use C<--listen> instead.
96
97			=item --workers
98
99			Specifies the number of worker pool. Defaults to 5.
100
101			Starman by default sets up other spare server configuration based on this
102			workers value, making sure there are B<always only> C<N> worker
103			processes running. So even if there're no idle workers, Starman won't
104			spawn off spare processes since that's mostly what you want to do by
105			fine tuning the memory usage etc. in the production environment.
106
107			=item --backlog
108
109			Specifies the number of backlog (listen queue size) of listener sockets. Defaults to 1024.
110
111			On production systems, setting a very low value can allow failover on
112			frontend proxy (like nginx) to happen more quickly, if you have
113			multiple Starman clusters.
114
115			If you're doing simple benchmarks and getting connection errors,
116			increasing this parameter can help avoid them. You should also
117			consider increasing C<net.core.somaxconn>. Note that this is not
118			recommended for real production system if you have another cluster to
119			failover (see above).
120
121			=item --max-requests
122
123			Number of the requests to process per one worker process. Defaults to 1000.
124
125			=item --preload-app
126
127			This option lets Starman preload the specified PSGI application in the
128			master parent process before preforking children. This allows memory
129			savings with copy-on-write memory management. When not set (default),
130			forked children loads the application in the initialization hook.
131
132			Enabling this option can cause bad things happen when resources like
133			sockets or database connections are opened at load time by the master
134			process and shared by multiple children.
135
136			Since Starman 0.2000, this option defaults to false, and you should
137			explicitly set this option to preload the application in the master
138			process.
139
140			Alternatively, you can use -M command line option (plackup's common
141			option) to preload the I<modules> rather than the <application>
142			itself.
143
144			starman -MCatalyst -MDBIx::Class myapp.psgi
145
146			will load the modules in the master process for memory savings with
147			CoW, but the actual loading of C<myapp.psgi> is done per children,
148			allowing resource managements such as database connection safer.
149
150			If you enable this option, sending C<HUP> signal to the master process
151			I<will not> pick up any code changes you make. See L</SIGNALS> for
152			details.
153
154			=item --disable-keepalive
155
156			Disable Keep-alive persistent connections. It is an useful workaround
157			if you run Starman behind a broken frontend proxy that tries to pool
158			connections more than a number of backend workers (i.e. Apache
159			mpm_prefork + mod_proxy).
160
161			=item --keepalive-timeout
162
163			The number of seconds Starman will wait for a subsequent request
164			before closing the connection if Keep-alive persistent connections
165			are enabled. Setting this to a high value may cause performance
166			problems in heavily loaded servers. The higher the timeout, the
167			more backend workers will be kept occupied waiting on connections
168			with idle clients.
169
170			Defaults to 1.
171
172			=item --read-timeout
173
174			The number of seconds Starman will wait for a request on a new connection
175			before closing it. Setting this to a high value may cause performance
176			problems in heavily loaded servers. The higher the timeout, the
177			more backend workers will be kept occupied waiting on connections
178			with idle clients. You may need this if your proxy / load balancer likes to
179			keep a pool of open connections while waiting for clients (eg. Amazon ELB).
180
181			Defaults to 5.
182
183			=item --user
184
185			To listen on a low-numbered (E<lt>1024) port, it will be necessary to
186			start the server as root. Use the C<--user> option to specify a userid
187			or username that the server process should switch to after binding to
188			the port.
189
190			Defaults to the current userid.
191
192			=item --group
193
194			Specify the group id or group name that the server should switch to after
195			binding to the port. This option is usually used with C<--user>.
196
197			Defaults to the current group id.
198
199			=item --pid
200
201			Specify the pid file path. Use it with C<-D\|--daemonize> option,
202			described in C<plackup -h>.
203
204			=item --error-log
205
206			Specify the pathname of a file where the error log should be written.
207			This enables you to still have access to the errors when using C<--daemonize>.
208
209			=item --ssl-cert
210
211			Specify the path to SSL certificate file.
212
213			=item --ssl-key
214
215			Specify the path to SSL key file.
216
217			=item --enable-ssl
218
219			Enable SSL on I<all> TCP sockets. This is an experimental feature.
220
221			=item --disable-proctitle
222
223			Disable the behavior to set proctitle to "starman (master)" and
224			"starman (worker)" respectively on master and workers.
225
226			=back
227
228			Starman passes through other options given to L<Plack::Runner>, the
229			common backend that L<plackup> uses, so the most options explained in
230			C<plackup -h> such as C<--access-log> or C<--daemonize> works fine in
231			starman too.
232
233			Setting the environment variable C<STARMAN_DEBUG> to 1 makes the
234			Starman server running in the debug mode.
235
236			=cut
237
238			=head1 SIGNALS
239
240			=over 4
241
242			=item HUP
243
244			Sending C<HUP> signal to the master process will restart all the workers
245			gracefully (meaning the currently running requests will shut down once
246			the request is complete), and by default, the workers will pick up the
247			code changes you make by reloading the application.
248
249			If you enable C<--preload-app> option, however, the code will be only
250			loaded in the startup process and will not pick up the code changes
251			you made. If you want to preload the app I<and> do graceful restarts
252			by reloading the code changes, you're recommended to use
253			L<Server::Starter>, configured to send C<QUIT> signal when superdaemon
254			received C<HUP>, i.e:
255
256			start_server --interval 5 --port 8080 --signal-on-hup=QUIT -- \
257			starman --preload-app myapp.psgi
258
259			You will then send the HUP signal to C<start_server> process to
260			gracefully reload the starman cluster (master and workers).
261
262			With Server::Starter 0.12 or later, you should also be able to set
263			C<--signal-on-term> to QUIT so that you can safely shutdown Starman
264			first and then stop the C<start_server> daemon process as well.
265
266			=item TTIN, TTOU
267
268			Sending C<TTIN> signal to the master process will dynamically increase
269			the number of workers, and C<TTOU> signal will decrease it.
270
271			=item INT, TERM
272
273			Sending C<INT> or C<TERM> signal to the master process will kill all
274			the workers immediately and shut down the server.
275
276			=item QUIT
277
278			Sending C<QUIT> signal to the master process will gracefully shutdown
279			the workers (meaning the currently running requests will shut down
280			once the request is complete).
281
282			=back
283
284			=head1 RELOADING THE APPLICATION
285
286			You're recommended to use signals (see above) to reload the
287			application, and are strongly discouraged to use C<-r> or C<-R>
288			(reloading flag) from plackup. These options will make a separate
289			directory watcher process, and makes your life difficult if you want to
290			combine with other process daemon tools such as L<Server::Starter>.
291
292			=head1 DIFFERENCES WITH PLACKUP
293
294			C<starman> executable is basically the equivalent of using C<plackup>
295			with C<Starman> server handler i.e. C<plackup -s Starman>, except that
296			C<starman> delay loads the application with the Delayed loader by
297			default, which can be disabled with C<--preload-app>.
298
299			C<starman> command also automatically sets the environment (C<-E>) to
300			the value of I<deployment>.
301
302			You're recommended to use C<starman> unless there's a reason to stick to
303			C<plackup> for compatibility.
304
305			=head1 SEE ALSO
306
307			L<Starman>
308
309			=cut