1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
|
[[overview]]
== Request overview
This chapter explains the different steps a request
goes through until a response is sent, along with
details of the Cowboy implementation.
=== Request/response
As you already know, HTTP clients connect to the server and
send a request for a resource; the server then sends a
response containing the resource if it could obtain it.
Before the server can send the resource, however, it
needs to perform many different operations to read the
request, find the resource, prepare the response being
sent and often other related operations the user can
add like writing logs.
Requests take the following route in Cowboy:
image::http_req_resp.png[HTTP request/response flowchart]
This shows the default middlewares, but they may be
configured differently in your setup. The dark green
indicates the points where you can hook your own code,
the light green is the Cowboy code that you can of
course configure as needed.
The `acceptor` is the part of the server that accepts
the connection and create an Erlang process to handle
it. The `parser` then starts reading from the socket
and handling requests as they come until the socket
is closed.
A response may be sent at many different points in the
life of the request. If Cowboy can't parse the request,
it gives up with an error response. If the router can't
find the resource, it sends a not found error. Your
own code can of course send a response at any time.
When a response is sent, you can optionally modify it
or act upon it by enabling the `onresponse` hook. By
default the response is sent directly to the client.
=== And then?
Behavior depends on what protocol is in use.
HTTP/1.0 can only process one request per connection,
so Cowboy will close the connection immediately after
it sends the response.
HTTP/1.1 allows the client to request that the server
keeps the connection alive. This mechanism is described
in the next section.
HTTP/2 is designed to allow sending multiple requests
asynchronously on the same connection. Details on what
this means for your application is described in this
chapter.
=== Keep-alive (HTTP/1.1)
With HTTP/1.1, the connection may be left open for
subsequent requests to come. This mechanism is called
`keep-alive`.
When the client sends a request to the server, it includes
a header indicating whether it would like to leave the
socket open. The server may or may not accept, indicating
its choice by sending the same header in the response.
Cowboy will include this header automatically in all
responses to HTTP/1.1 requests. You can however force
the closing of the socket if you want. When Cowboy sees
you want to send a `connection: close` header, it will
not override it and will close the connection as soon
as the reply is sent.
This snippet will force Cowboy to close the connection.
[source,erlang]
----
Req2 = cowboy_req:reply(200, [
{<<"connection">>, <<"close">>},
], <<"Closing the socket in 3.. 2.. 1..">>, Req).
----
Cowboy will only accept a certain number of new requests
on the same connection. By default it will run up to 100
requests. This number can be changed by setting the
`max_keepalive` configuration value when starting an
HTTP listener.
[source,erlang]
----
cowboy:start_http(my_http_listener, 100, [{port, 8080}], [
{env, [{dispatch, Dispatch}]},
{max_keepalive, 5}
]).
----
Cowboy implements the keep-alive mechanism by reusing
the same process for all requests. This allows Cowboy
to save memory. This works well because most code will
not have any side effect impacting subsequent requests.
But it also means you need to clean up if you do have
code with side effects. The `terminate/3` function can
be used for this purpose.
=== Pipelining (HTTP/1.1)
While HTTP is designed as a sequential protocol, with
the client sending a request and then waiting for the
response from the server, nothing prevents the client
from sending more requests to the server without waiting
for the response, due to how sockets work. The server
still handles the requests sequentially and sends the
responses in the same order.
This mechanism is called pipelining. It allows reducing
latency when a client needs to request many resources
at the same time. This is used by browsers when requesting
static files for example.
This is handled automatically by the server.
=== Asynchronous requests (HTTP/2)
In HTTP/2, the client can send a request at any time.
And the server can send a response at any time too.
This means for example that the client does not need
to wait for a request to be fully sent to send another,
it is possible to interleave a request with the request
body of another request. The same is true with responses.
Responses may also be sent in a different order.
Because requests and responses are fully asynchronous,
Cowboy creates a new process for each request, and these
processes are managed by another process that handles the
connection itself.
HTTP/2 servers may also decide to send resources to the
client before the client requests them. This is especially
useful for sending static files associated with the HTML
page requested, as this reduces the latency of the overall
response. Cowboy does not support this particular mechanism
at this point, however.
|
