| line |
stmt |
bran |
cond |
sub |
pod |
time |
code |
|
1
|
|
|
|
|
|
|
package Message::Passing::PSGI; |
|
2
|
1
|
|
|
1
|
|
812
|
use strict; |
|
|
1
|
|
|
|
|
3
|
|
|
|
1
|
|
|
|
|
49
|
|
|
3
|
1
|
|
|
1
|
|
7
|
use warnings; |
|
|
1
|
|
|
|
|
1
|
|
|
|
1
|
|
|
|
|
86
|
|
|
4
|
|
|
|
|
|
|
|
|
5
|
|
|
|
|
|
|
our $VERSION = '0.001'; |
|
6
|
|
|
|
|
|
|
$VERSION = eval $VERSION; |
|
7
|
|
|
|
|
|
|
|
|
8
|
|
|
|
|
|
|
1; |
|
9
|
|
|
|
|
|
|
|
|
10
|
|
|
|
|
|
|
=head1 NAME |
|
11
|
|
|
|
|
|
|
|
|
12
|
|
|
|
|
|
|
Message::Passing::PSGI - ALPHA QUALITY PSGI adaptor for Message::Passing |
|
13
|
|
|
|
|
|
|
|
|
14
|
|
|
|
|
|
|
=head1 SYNOPSIS |
|
15
|
|
|
|
|
|
|
|
|
16
|
|
|
|
|
|
|
# Run the server - note that the -e has to all be on one line! |
|
17
|
|
|
|
|
|
|
plackup -E production -s Twiggy -MPlack::App::Message::Passing -e'Plack::App::Message::Passing->new(return_address => "tcp://127.0.0.1:5555", send_address => "tcp://127.0.0.1:5556")->to_app' |
|
18
|
|
|
|
|
|
|
|
|
19
|
|
|
|
|
|
|
# Run your app with the handler |
|
20
|
|
|
|
|
|
|
plackup -E production -s Message::Passing testapp.psgi --host 127.0.0.1 --port 5556 |
|
21
|
|
|
|
|
|
|
|
|
22
|
|
|
|
|
|
|
# Browse to: |
|
23
|
|
|
|
|
|
|
http://localhost:5000/ |
|
24
|
|
|
|
|
|
|
|
|
25
|
|
|
|
|
|
|
=head1 DESCRIPTION |
|
26
|
|
|
|
|
|
|
|
|
27
|
|
|
|
|
|
|
B<ALPHA QUALITY EXPERIMENT - YOU HAVE BEEN WARNED!> |
|
28
|
|
|
|
|
|
|
|
|
29
|
|
|
|
|
|
|
This module implements a mongrel2 like strategy for web handlers, using |
|
30
|
|
|
|
|
|
|
L<Message::Passing::ZeroMQ>. |
|
31
|
|
|
|
|
|
|
|
|
32
|
|
|
|
|
|
|
=head1 WHY |
|
33
|
|
|
|
|
|
|
|
|
34
|
|
|
|
|
|
|
Because I could! It's a useful experiment to prove that L<Message::Passing> |
|
35
|
|
|
|
|
|
|
can be used for things entirely unlike my initial goals. |
|
36
|
|
|
|
|
|
|
|
|
37
|
|
|
|
|
|
|
=head2 NO, REALLY? |
|
38
|
|
|
|
|
|
|
|
|
39
|
|
|
|
|
|
|
B<Theoretically>, this is quite an interesting model - as you've totally split |
|
40
|
|
|
|
|
|
|
the front end connection acceptance and the back end request handling, |
|
41
|
|
|
|
|
|
|
you can do things which are harder in other server environments trivially. |
|
42
|
|
|
|
|
|
|
|
|
43
|
|
|
|
|
|
|
Examples of things that 'just work' include: |
|
44
|
|
|
|
|
|
|
|
|
45
|
|
|
|
|
|
|
=over |
|
46
|
|
|
|
|
|
|
|
|
47
|
|
|
|
|
|
|
=item Adding more handler processes |
|
48
|
|
|
|
|
|
|
|
|
49
|
|
|
|
|
|
|
Totally dynamic, run as many as you want |
|
50
|
|
|
|
|
|
|
|
|
51
|
|
|
|
|
|
|
=item Adding handler processes on other servers |
|
52
|
|
|
|
|
|
|
|
|
53
|
|
|
|
|
|
|
As long as your send/return sockets are bound to a host that's network |
|
54
|
|
|
|
|
|
|
accessible, you can spin up handlers wherever you want. |
|
55
|
|
|
|
|
|
|
|
|
56
|
|
|
|
|
|
|
=item Upgrade the application in production |
|
57
|
|
|
|
|
|
|
|
|
58
|
|
|
|
|
|
|
You can spin up a new version, verify it appears to be working correctly |
|
59
|
|
|
|
|
|
|
etc before shutting down the old version |
|
60
|
|
|
|
|
|
|
|
|
61
|
|
|
|
|
|
|
=item Profile the application in production |
|
62
|
|
|
|
|
|
|
|
|
63
|
|
|
|
|
|
|
Just run a handler with NYTProf.. |
|
64
|
|
|
|
|
|
|
|
|
65
|
|
|
|
|
|
|
=back |
|
66
|
|
|
|
|
|
|
|
|
67
|
|
|
|
|
|
|
B<NOTE:> The properties above _do not_ exist in the current code - you |
|
68
|
|
|
|
|
|
|
B<will> drop requests in-flight if you shut handlers down!! (Patches to fix |
|
69
|
|
|
|
|
|
|
this should not be that hard, and would be welcome if anyone is interested) |
|
70
|
|
|
|
|
|
|
|
|
71
|
|
|
|
|
|
|
If you're actually interested in using this in production, I'd recommend |
|
72
|
|
|
|
|
|
|
you look at the real mongrel2, and L<Plack::Handler::Mongrel2>. |
|
73
|
|
|
|
|
|
|
|
|
74
|
|
|
|
|
|
|
=head1 BUGS |
|
75
|
|
|
|
|
|
|
|
|
76
|
|
|
|
|
|
|
Many, and varied. Please do not try to run this in production ;_) |
|
77
|
|
|
|
|
|
|
|
|
78
|
|
|
|
|
|
|
Issues include: |
|
79
|
|
|
|
|
|
|
|
|
80
|
|
|
|
|
|
|
=over |
|
81
|
|
|
|
|
|
|
|
|
82
|
|
|
|
|
|
|
=item Large responses will use SEVERAL times the response length in RAM |
|
83
|
|
|
|
|
|
|
|
|
84
|
|
|
|
|
|
|
=item Requests never timeout |
|
85
|
|
|
|
|
|
|
|
|
86
|
|
|
|
|
|
|
=item App Handler crashes / restarts will lost in-flight requests. |
|
87
|
|
|
|
|
|
|
|
|
88
|
|
|
|
|
|
|
=item Quite probably leaks RAM. |
|
89
|
|
|
|
|
|
|
|
|
90
|
|
|
|
|
|
|
This has not been tested, which means I quite probably got it wrong somewhere ;) |
|
91
|
|
|
|
|
|
|
|
|
92
|
|
|
|
|
|
|
=back |
|
93
|
|
|
|
|
|
|
|
|
94
|
|
|
|
|
|
|
=head1 SEE ALSO |
|
95
|
|
|
|
|
|
|
|
|
96
|
|
|
|
|
|
|
=over |
|
97
|
|
|
|
|
|
|
|
|
98
|
|
|
|
|
|
|
=item L<Plack::App::Message::Passing>. |
|
99
|
|
|
|
|
|
|
|
|
100
|
|
|
|
|
|
|
=item L<Plack::Handler::Message::Passing>. |
|
101
|
|
|
|
|
|
|
|
|
102
|
|
|
|
|
|
|
=item mongrel2 |
|
103
|
|
|
|
|
|
|
|
|
104
|
|
|
|
|
|
|
=item L<Message::Passing> |
|
105
|
|
|
|
|
|
|
|
|
106
|
|
|
|
|
|
|
=item L<Message::Passing::ZeroMQ>. |
|
107
|
|
|
|
|
|
|
|
|
108
|
|
|
|
|
|
|
=back |
|
109
|
|
|
|
|
|
|
|
|
110
|
|
|
|
|
|
|
=head1 AUTHOR |
|
111
|
|
|
|
|
|
|
|
|
112
|
|
|
|
|
|
|
Tomas Doran (t0m) C<< <bobtfish@bobtfish.net> >> |
|
113
|
|
|
|
|
|
|
|
|
114
|
|
|
|
|
|
|
=head1 COPYRIGHT |
|
115
|
|
|
|
|
|
|
|
|
116
|
|
|
|
|
|
|
Copyright the above author. |
|
117
|
|
|
|
|
|
|
|
|
118
|
|
|
|
|
|
|
=head1 LICENSE |
|
119
|
|
|
|
|
|
|
|
|
120
|
|
|
|
|
|
|
GNU Affero General Public License, Version 3 |
|
121
|
|
|
|
|
|
|
|
|
122
|
|
|
|
|
|
|
If you feel this is too restrictive to be able to use this software, |
|
123
|
|
|
|
|
|
|
please talk to us as we'd be willing to consider re-licensing under |
|
124
|
|
|
|
|
|
|
less restrictive terms. |
|
125
|
|
|
|
|
|
|
|
|
126
|
|
|
|
|
|
|
=cut |
|
127
|
|
|
|
|
|
|
|
|
128
|
|
|
|
|
|
|
1; |
|
129
|
|
|
|
|
|
|
|