| Age | Commit message (Collapse) | Author |
|---|
|
The approach taken here is very similar to what browsers are
doing. A separate pool is created for each host/port/scope.
The authority (host header) is used to determine which pool
will execute requests. A connection process is semi-randomly
chosen, from the connections that have capacity. Maximum
capacity is determined by the protocol (the HTTP/2 setting
set by the server is used, for example). Multiple processes
can process requests/responses on the same connection
concurrently. There is no need to "give back" the response
to the pool, the number of ongoing streams is maintained via
an event handler.
The implementation is currently not strict, there may be
more attempts to create requests than there is capacity.
I'm not sure if it should be made strict or if Gun should
just wait before sending requests (it only matters in the
HTTP/2 case at the moment).
When there is no connection with capacity available in the
pool (because they have too many streams, or are reconnecting,
or any other reason), checking out fails. There is no timeout
to wait for a connection to be available. On the other hand
the checkout_retry option allows setting multiple timeouts
to retry checking out a connection. Each retry attempt's
wait time can have a different value.
The initial implementation of this work was sponsored by
Kobil and made at the suggestion of Ilya Khaprov.
|
|
The test suite is 216 tests with a tunnel created via two
proxies leading to one origin server. The tests are for
example socks5_h2_https where socks5 identifies the first
SOCKS5 proxy, h2 the second HTTP/2 CONNECT proxy and https
the secure HTTP/1.1 origin server.
The test not only sets up the tunnel and does a request
(or sends/receives data in the case of raw origin servers)
but also confirms that the stream_info and info data is
correct.
|
|
|
|
|
|
|
|
The graceful shutdown is implemented through a new 'closing'
state. This state is entered under different circumstances
depending on the protocol.
The gun:shutdown/1 function is now implemented and documented.
It allows shutting down the connection gracefully regardless
of the current state of the connection and for all protocols.
The behavior is entirely dependent on the protocol.
For HTTP/1.1 the connection stays up only until after the
current stream is complete; other streams are immediately
canceled.
For HTTP/2 a GOAWAY frame is sent and existing streams
continue to be processed. The connection is closed after
all streams are processed and the server's GOAWAY frame
is received.
For Websocket a close frame is sent. The connection is
closed when receiving the server's close frame.
In all cases the closing_timeout option defines how long
we wait, as a maximum, before closing the connection after
the graceful shutdown was started.
The graceful shutdown is also initiated when the owner
process goes away; when sending an HTTP/1.1 request
with the connection: close header; when receiving an
HTTP/1.1 response with the connection: close header;
when receiving an HTTP/1.0 response without a connection
header; when the server sends a GOAWAY HTTP/2 frame;
or when we send or receive a Websocket close frame.
Along with these changes, the gun:ws_send/2 function
now accepts a list of frames as argument. Those frames
may include a close frame that initiates the graceful
shutdown.
|
|
Flow control is disabled by default. The initial flow value
must be set to enable it (either for the entire connection
or on a per-request basis). Flow applies to all HTTP streams
as well as Websocket. HTTP/2 pushed streams receive the same
value as their originating stream.
|
|
|
|
|
|
This covers many scenarios but more need to be added.
|
|
An invalid stream reference (the websocket tuple wrapper)
was sent in the gun_data message.
Also moves autobahn to its own test suite.
|
|
Some intermittent failures occurred because of trying to
connect to google.com or echo.websocket.org.
|
|
|
|
|
|
|
