https://github.com/hubo1016/aiogrpc

asyncio wrapper for grpc.io
https://github.com/hubo1016/aiogrpc

Last synced: 29 days ago
JSON representation

asyncio wrapper for grpc.io

• Host: GitHub
• URL: https://github.com/hubo1016/aiogrpc
• Owner: hubo1016
• License: apache-2.0
• Created: 2017-08-11T09:23:51.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2020-08-24T12:53:31.000Z (over 4 years ago)
• Last Synced: 2024-11-02T22:50:39.448Z (about 1 month ago)
• Language: Python
• Size: 50.8 KB
• Stars: 115
• Watchers: 13
• Forks: 12
• Open Issues: 3
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# aiogrpc

[![Build Status](https://travis-ci.org/hubo1016/aiogrpc.svg?branch=master)](https://travis-ci.org/hubo1016/aiogrpc)

[![PyPI](https://img.shields.io/pypi/v/aiogrpc.svg)](https://pypi.python.org/pypi/aiogrpc)

asyncio wrapper for grpc.io. Currently, only client-side interfaces are supported.

# Interfaces

aiogrpc has exactly the same interface as grpc.io (https://grpc.io/grpc/python/grpc.html#module-grpc), except:

- All callbacks are called with `call_soon_threadsafe()` to run in the same thread with the event loop

- All futures are asyncio futures (i.e. awaitable)

- All blocking interfaces (`__call__` for unary methods, `code()`, `details()`, ...) become coroutines

- Stream input/output are async iterators

- `close()` method returns a asyncio future (use `await channel.close()`) (*since v1.4*)

# Usage

1. Compile your proto file with protoc to generate stub files

2. Use the stub files with aiogrpc.Channel, instead of grpc.Channel:

    ```python

    from aiogrpc import insecure_channel

    import asyncio

    from mystub import MyStub

    

    channel = insecure_channel('ipv4:///127.0.0.1:8080')

    mystub = MyStub(channel)

    

    async def test_call():

        return await mystub.mymethod(...)

    

    async def test_call_stream():

        async for v in mystub.my_stream_method(...):

            ...

    ```

# Balancing

Same as the original grpc.io, balancing (fail over) between multiple servers are supported:

1. Use multiple IPv4 or IPv6 addresses for balancing - use target like: `"ipv4:///1.2.3.4:9999,1.2.3.5:9999,1.2.3.6:9999"`

   or `"ipv6:///[1::2]:9999,[1::3]:9999,[1::4]:9999"`

2. Specify **ONE** DNS name which resolves to multiple addresses. GRPC will balance between the resolved addresses, and try

   to update the resolving result in interval. Example: `"dns:///myserver:9999"`

   

Notice that you cannot mix up IPv4, IPv6 and DNS addresses; only addresses with the same type

can be balanced. Also, when using DNS target, only one DNS name can be specified.

The default load balancing strategy is `"pick_first"` (fail-over). Set `"grpc.lb_policy_name"` option to `"round_robin"` for

round-robin load balancing: `channel = insecure_channel('ipv4:///1.2.3.4:9999,1.2.3.5:9999,1.2.3.6:9999', [('grpc.lb_policy_name', 'round_robin')])`

# Copyright Notice

All code are published under Apache 2.0 License

Some code and docstrings are modified from grpc.io (https://grpc.io/grpc/python/grpc.html#module-grpc)