2018-03-31 18:51:45 +00:00
|
|
|
Overview
|
|
|
|
========
|
|
|
|
|
2018-03-31 23:50:40 +00:00
|
|
|
gRPC protocol is exclusively based on HTTP/2 (aka h2) protocol. Main concepts:
|
|
|
|
|
|
|
|
- each request in h2 connection is a bidirectional stream of frames_;
|
|
|
|
- streams give ability to do multiplexing_ - make requests in parallel using
|
|
|
|
single TCP connection;
|
|
|
|
- h2 has special `flow control`_ mechanism, which can help avoid network
|
|
|
|
congestion and fix problems with slow clients, slow servers and slow network;
|
|
|
|
- flow control only affects ``DATA`` frames, any other frame can be sent
|
|
|
|
without limitations;
|
|
|
|
- streams can be cancelled individually, all other streams in h2 connection
|
|
|
|
will continue work and there is no need to drop connection and reconnect;
|
|
|
|
- h2 is a binary protocol and allows headers compression using HPACK_ format,
|
|
|
|
so it is a very strict and efficient protocol.
|
|
|
|
|
|
|
|
h2 protocol is highly configurable, for example:
|
|
|
|
|
|
|
|
- `flow control`_ mechanism can use dynamically configurable
|
|
|
|
initial window size, to better match different use cases and conditions;
|
|
|
|
- you can set maximum frame size to control how much data you will
|
|
|
|
receive in each frame;
|
|
|
|
- you can limit number of concurrent streams for h2 connection.
|
|
|
|
|
|
|
|
gRPC protocol adds to h2 protocol messages encoding format and a notion
|
|
|
|
about metadata. gRPC metadata == additional h2 headers. So gRPC has the same
|
|
|
|
level of extensibility as HTTP has.
|
|
|
|
|
|
|
|
Messages are sent using one or several ``DATA`` frames, depending on maximum
|
|
|
|
frame size setting and message size. Messages are encoded using simple format:
|
|
|
|
prefix + data. Prefix contains length of the data and compression flag. You
|
2018-04-01 01:30:40 +00:00
|
|
|
can learn gRPC wire protocol in more details here: `gRPC format`_.
|
2018-03-31 23:50:40 +00:00
|
|
|
|
|
|
|
gRPC has 4 method types: unary-unary, unary-stream (e.g. download),
|
|
|
|
stream-unary (e.g. upload), stream-stream. They are all the same, the only
|
|
|
|
difference is how many messages are sent in each direction: exactly one (unary)
|
|
|
|
or any number of messages (stream).
|
|
|
|
|
2018-04-01 01:30:40 +00:00
|
|
|
Cancellation
|
|
|
|
~~~~~~~~~~~~
|
|
|
|
|
|
|
|
As it was said above, h2 allows you to cancel any stream without affecting other
|
|
|
|
streams, which are living in the same connection. And h2 protocol has special
|
|
|
|
frame to do this: RST_STREAM_. Both client and server can cancel streams.
|
|
|
|
This feature automatically gives you ability to proactively cancel gRPC method
|
|
|
|
calls in the same way. In ``grpclib`` you can cancel method calls immediately,
|
|
|
|
for example:
|
|
|
|
|
|
|
|
- client sends request to the server
|
|
|
|
- server spawns task to handle this request
|
|
|
|
- client wants to cancel this request and sends ``RST_STREAM`` frame
|
|
|
|
- server receives ``RST_STREAM`` frame and cancels task immediately
|
|
|
|
|
|
|
|
Most other protocols doesn't have this feature, so they have to terminate
|
|
|
|
whole TCP connection and perform reconnect for the next call. It is also not
|
|
|
|
obvious how to immediately detect terminated connections on the other side,
|
|
|
|
and this means that server most likely will continue result computations, when
|
|
|
|
this result is not needed anymore.
|
|
|
|
|
|
|
|
Deadlines
|
|
|
|
~~~~~~~~~
|
|
|
|
|
|
|
|
Deadlines are basically timeouts, which are propagated from service to service,
|
|
|
|
to meet initial timeout constrains. This is a simple and powerful idea.
|
|
|
|
|
|
|
|
Example:
|
|
|
|
|
|
|
|
- service X receives request with ``grpc-timeout: 100m`` in metadata
|
|
|
|
(100m means 100 milliseconds)
|
|
|
|
|
|
|
|
- service X immediately converts timeout into deadline:
|
|
|
|
|
|
|
|
.. code-block:: python
|
|
|
|
|
|
|
|
deadline = time.monotonic() + grpc_timeout
|
|
|
|
|
|
|
|
- service X spent 20ms doing some work
|
|
|
|
|
|
|
|
- now service X wants to make outgoing request to service Y, so it computes
|
|
|
|
how much time remains to perform this request:
|
|
|
|
|
|
|
|
.. code-block:: python
|
|
|
|
|
|
|
|
new_timeout = max(deadline - time.monotonic(), 0) # == 80ms
|
|
|
|
|
|
|
|
- service X performs request to service Y with metadata ``grpc-timeout: 80m``
|
|
|
|
|
|
|
|
- service Y uses the same logic to convert timeout -> deadline -> timeout.
|
|
|
|
|
|
|
|
With this feature it is possible to cancel whole call chain simultaneously,
|
|
|
|
even in case of network failures (broken connections).
|
|
|
|
|
2018-03-31 23:50:40 +00:00
|
|
|
grpclib
|
|
|
|
~~~~~~~
|
|
|
|
|
|
|
|
``grpclib`` tries to give you full control over these bidirectional h2 streams.
|
|
|
|
|
2018-03-31 18:51:45 +00:00
|
|
|
.. note:: *[auto]* mark below means that it is not necessary to explicitly call
|
|
|
|
these methods in your code, they will be called automatically behind the
|
|
|
|
scenes. They are exists to have more control.
|
|
|
|
|
|
|
|
.. raw:: html
|
|
|
|
:file: _static/diagram.html
|
2018-03-31 23:50:40 +00:00
|
|
|
|
|
|
|
.. _frames: http://httpwg.org/specs/rfc7540.html#FrameTypes
|
|
|
|
.. _multiplexing: http://httpwg.org/specs/rfc7540.html#StreamsLayer
|
|
|
|
.. _flow control: http://httpwg.org/specs/rfc7540.html#FlowControl
|
|
|
|
.. _HPACK: http://httpwg.org/specs/rfc7541.html
|
2018-04-01 01:30:40 +00:00
|
|
|
.. _gRPC format: https://github.com/grpc/grpc/blob/master/doc/PROTOCOL-HTTP2.md
|
|
|
|
.. _RST_STREAM: http://httpwg.org/specs/rfc7540.html#RST_STREAM
|