Pyshard

Pyshard is a complete distributed key-value data storage written in Python using only standard library tools. Pyshard's using hash based sharding method. It means that shard of value you write will be selected in accordance to key hash (regards to lgiordani/pyshard). This project is experimental and should be used in another project pdx - distributed web indexing service.

Installation

pip install pyshard

Quick start

Bootstrap

To run 'hello world' service you need started up shard servers. For example:

# test_server.py

import sys
import asyncio
from pyshard import ShardServer

if __name__ == '__main__':
    loop = asyncio.get_event_loop()

    server = ShardServer(host=sys.argv[1], port=int(sys.argv[2]), start=.0, end=1.0)
    try:
        loop.run_until_complete(server._do_run())
    finally:
        loop.close()

python test_server.py localhost 5050 & \
python test_server.py localhost 5051

After servers started up you should start bootstrap server to map shards. Now bootstrap server needs config file with shard's markers:

{
  "shards": [
    {
      "name": "shard0-0.5",
      "start": 0.0,
      "end": 0.5,
      "size": 1024,
      "host": "127.0.0.1",
      "port": 5050
    },
    {
      "name": "shard0.5-1",
      "start": 0.5,
      "end": 1.0,
      "size": 1024,
      "host": "127.0.0.1",
      "port": 5051
    }
  ]
}

Every shard has next parameters: name - unique string name of shard, start and end - numeric limits of key hash, size - memory limit for this shard, host and port - shard address. start and end limit means that this shard will store values with key hash in range [start, end].

# test_bootstrap_server.py

import asyncio

from pyshard import BootstrapServer

from pyshard.settings import settings


if __name__ == '__main__':
    loop = asyncio.get_event_loop()
    server = BootstrapServer(*settings.BOOTSTRAP_SERVER, config_path='config_example.json',
                             buffer_size=1024, loop=loop)
    try:
        loop.run_until_complete(server._do_run())
    finally:
        loop.close()

python test_bootstrap_server.py

Now shards have got configurations from bootstrap service and ready.

App

>>> from pyshard import Pyshard
>>> from pyshard.settings import settings
>>> 
>>> app = Pyshard(bootstrap_server=settings.BOOTSTRAP_SERVER)
>>> app.create_index('test_index')
>>> app.write(index='test_index', key='test', doc='hello world')
60
>>> app.read(index='test_index', key='test')
{'hash_': 0.1671936, 'record': 'hello world'}
>>> app.write('test_index', 'test1', {'hello': 'world'})
54
>>> app.read('test_index', 'test')
{'hash_': 0.8204544, 'record': {'hello': 'world'}}
>>> app.pop('test_index', 'test1')
{'hash_': 0.8204544, 'record': {'hello': 'world'}}

Utilities

Since version 0.2.0 Pyshard has several console utilities. They are made to simplify some operations like cat or massive write.

Let's make file with data. Row format: {key}|{value}:

printf '1|test\n2|{"test": "test"}\n3|42\n4|0.9\n' > test_write.txt

We can add this rows to storage using pyshard write command.

cat test_write.txt | pyshard write test_index --force

--force oprion for creating index test_index if it does not exist

So let's cat storage with index test_index:

pyshard cat test_index

Command will log results to stdout:

2|{"hash_": 0.2258304, "record": {"test": "test"}}
3|{"hash_": 0.1904896, "record": 42}
1|{"hash_": 0.8102784, "record": "test"}
4|{"hash_": 0.7252864, "record": 0.9}

TODO

Index (data tables equivalent)
Connection id for shard servers (now it is an address)
App utils (pyshard read, pyshard write)
Nice run methods for services
Makefile