nydus 0.11.0

nydus.db.create_cluster

The create_cluster function is a simple helper to configure a Cluster based on a simple dict config.

There are two required keyword arguments:

• backend: full path to the backend class, which should extend nydus.db.backends.BaseConnection.

• hosts: a dictionary, where the key is an ordered numeric value, and the result is a dict of connection options.

  (the keys are numeric values simply for readability in configuration)

• defaults: a dictionary of default settings for all hosts to inherit from.

Optionally, you may also specify a value for router, which is the full path to the router class, which must extend nydus.db.routers.BaseRouter.

Distributed Queries

In some cases you may want to execute a query on many nodes (in parallel). Nydus has built-in support for this when any routing function returns a cluster of nodes:

from nydus.db import create_cluster
>>>
# by not specifying a router all queries are executed on all hosts
redis = create_cluster({
    'backend': 'nydus.db.backends.redis.Redis',
    'hosts': {
        0: {'db': 0},
        1: {'db': 1},
        2: {'db': 2},
    }
})
>>>
# map the query over all connections returned by the default router
res = redis.incr('foo')
>>>
assert type(res) == list
assert len(res) == 3

You can also map many queries (utilizing an internal queue) over connections (again, returned by the router):

with redis.map() as conn:
    results = [conn.incr(k) for k in keys]

As of release 0.5.0, the map() function now supports pipelines, and the included Redis backend will pipeline commands wherever possible.

Redis

Nydus was originally designed as a toolkit to expand on the usage of Redis at DISQUS. While it does provide a framework for building clusters that are not Redis, much of the support has gone into providing utilities for routing and querying on Redis clusters.

You can configure the Redis client for a connection by specifying it’s full path:

redis = create_cluster({
    'backend': 'nydus.db.backends.redis.Redis',
    'hosts': {
        0: {'db': 0},
    },
})

The available host options are:

• host

• port

• db

• timeout

• password

• identifier

The Redis client also supports pipelines via the map command. This means that all commands will hit servers at most as of once:

with redis.map() as conn:
    conn.set('a', 1)
    conn.incr('a')
    results = [conn.get('a'), conn.get('b')]
results['a'] == 2
results['b'] == None

redis = create_cluster({
    'backend': 'nydus.db.backends.redis.Redis',
    'router': 'nydus.db.routers.keyvalue.PartitionRouter',
    'hosts': {
        0: {'db': 0},
    },
})

redis = create_cluster({
    'backend': 'nydus.db.backends.redis.Redis',
    'router': 'nydus.db.routers.keyvalue.ConsistentHashingRouter',
    'hosts': {
        0: {'db': 0},
    },
})

redis = create_cluster({
    'backend': 'nydus.db.backends.redis.Redis',
    'router': 'nydus.db.routers.RoundRobinRouter',
    'hosts': {
        0: {'db': 0},
    },
})

Pycassa

Basic connection management for pycassa (Cassandra) clusters is supported, but we use a non-standard syntax for creating clusters as routing behavior and per-connection options are not useful in this context:

pycassa = create_cluster({
    'backend': 'nydus.db.backends.pycassa.Pycassa',
    'hosts': ['localhost'],
    'keyspace': 'test',
})