dflockd-client 1.4.0

dflockd python client

dflockd-client

• dflockd-client
  • Installation
  • Quick start
    • Async client
    • Sync client
    • Manual acquire/release
    • Two-phase lock acquisition
    • Parameters
  • Semaphores
    • Parameters
  • Stats
  • Multi-server sharding

A Python client library for dflockd — a lightweight distributed lock server with FIFO ordering, automatic lease expiry, and background renewal.

Read the docs here

Installation

pip install dflockd-client

Or with uv:

uv add dflockd-client

Quick start

Async client

import asyncio
from dflockd_client.client import DistributedLock

async def main():
    async with DistributedLock("my-key", acquire_timeout_s=10) as lock:
        print(lock.token, lock.lease)
        # critical section — lease auto-renews in background

asyncio.run(main())

Sync client

from dflockd_client.sync_client import DistributedLock

with DistributedLock("my-key", acquire_timeout_s=10) as lock:
    print(lock.token, lock.lease)
    # critical section — lease auto-renews in background thread

Manual acquire/release

Both clients support explicit acquire() / release() outside of a context manager:

from dflockd_client.sync_client import DistributedLock

lock = DistributedLock("my-key")
if lock.acquire():
    try:
        pass  # critical section
    finally:
        lock.release()

Two-phase lock acquisition

The enqueue() / wait() methods split lock acquisition into two steps, allowing you to notify an external system after joining the queue but before blocking:

from dflockd_client.sync_client import DistributedLock

lock = DistributedLock("my-key")
status = lock.enqueue()       # join queue, returns "acquired" or "queued"
notify_external_system()      # your application logic here
if lock.wait(timeout_s=10):   # block until granted (no-op if already acquired)
    try:
        pass  # critical section
    finally:
        lock.release()

Async equivalent:

lock = DistributedLock("my-key")
status = await lock.enqueue()
await notify_external_system()
if await lock.wait(timeout_s=10):
    try:
        pass  # critical section
    finally:
        await lock.release()

Parameters

Parameter	Default	Description
key	(required)	Lock name
acquire_timeout_s	10	Seconds to wait for lock acquisition
lease_ttl_s	None (server default)	Lease duration in seconds
servers	[("127.0.0.1", 6388)]	List of (host, port) tuples
sharding_strategy	stable_hash_shard	Callable[[str, int], int] — maps (key, num_servers) to server index
renew_ratio	0.5	Renew at lease * ratio seconds

Semaphores

DistributedSemaphore allows up to N concurrent holders per key, using the same API patterns as DistributedLock:

from dflockd_client.sync_client import DistributedSemaphore

# Allow up to 3 concurrent workers on this key
with DistributedSemaphore("my-key", limit=3, acquire_timeout_s=10) as sem:
    print(sem.token, sem.lease)
    # critical section — up to 3 holders at once

Async equivalent:

from dflockd_client.client import DistributedSemaphore

async with DistributedSemaphore("my-key", limit=3, acquire_timeout_s=10) as sem:
    print(sem.token, sem.lease)

Manual acquire/release and two-phase (enqueue() / wait()) work the same as locks.

Parameters

Parameter	Default	Description
key	(required)	Semaphore name
limit	(required)	Maximum concurrent holders
acquire_timeout_s	10	Seconds to wait for acquisition
lease_ttl_s	None (server default)	Lease duration in seconds
servers	[("127.0.0.1", 6388)]	List of (host, port) tuples
sharding_strategy	stable_hash_shard	Callable[[str, int], int] — maps (key, num_servers) to server index
renew_ratio	0.5	Renew at lease * ratio seconds

Stats

Query server state (connections, held locks, active semaphores) using the low-level stats() function:

import asyncio
from dflockd_client.client import stats

async def main():
    reader, writer = await asyncio.open_connection("127.0.0.1", 6388)
    result = await stats(reader, writer)
    print(result)
    # {'connections': 1, 'locks': [], 'semaphores': [], 'idle_locks': [], 'idle_semaphores': []}
    writer.close()
    await writer.wait_closed()

asyncio.run(main())

Sync equivalent:

import socket
from dflockd_client.sync_client import stats

sock = socket.create_connection(("127.0.0.1", 6388))
rfile = sock.makefile("r", encoding="utf-8")
result = stats(sock, rfile)
print(result)
rfile.close()
sock.close()

Returns a dict with connections, locks, semaphores, idle_locks, and idle_semaphores.

Multi-server sharding

When running multiple dflockd instances, the client can distribute keys across servers using consistent hashing. Each key always routes to the same server.

from dflockd_client.sync_client import DistributedLock

servers = [("server1", 6388), ("server2", 6388), ("server3", 6388)]

with DistributedLock("my-key", servers=servers) as lock:
    print(lock.token, lock.lease)

The default strategy uses zlib.crc32 for stable, deterministic hashing. You can provide a custom strategy:

from dflockd_client.sync_client import DistributedLock

def my_strategy(key: str, num_servers: int) -> int:
    """Route all keys to the first server."""
    return 0

with DistributedLock("my-key", servers=servers, sharding_strategy=my_strategy) as lock:
    pass