A small wrapper around Redis (using the "standard" Python redis client lib) that provides access to a FIFO queue and allows upstream code to mark pop'ed items as processed successfully (ACK'ing) or unsucessfully (FAIL'ing).
Failed items are automatically requeued. Addionally a backup is kept for items that were neither ack'ed nor fail'ed, i.e. in case the consumer crashes. The backup items will be requeued as soon as one of the consumer(s) comes up again.
All actions are atomic and it is safe to have multiple producers and consumers accessing the same queue concurrently.
redis.py | saferedisqueue |
---|---|
2.4.10 - 2.6.x | 1.x |
2.7.0 - 2.7.5 | no compatible version |
2.7.6 - 2.10.x | 2.x-3.x |
>>> queue = SafeRedisQueue(name='test')
>>> queue.put("Hello World")
'595d43b2-2e49-4e96-a1d2-0848d1c7f0d3'
>>> queue.put("Foo bar")
'1df060eb-b578-499d-bede-20db9da8184e'
>>> queue.get()
('595d43b2-2e49-4e96-a1d2-0848d1c7f0d3', 'Hello World')
>>> queue.get()
('1df060eb-b578-499d-bede-20db9da8184e', 'Foo bar')
Note: to be compatible with previous versions, 2 aliases push/pop
exist. Start using the new put/get
terminology as soon as possible since push/pop
will be deleted in a future version.
>>> queue.put("Good stuff")
>>> queue.put("Bad stuff")
>>> uid_good, payload_good = queue.get()
>>> uid_bad, payload_bad = queue.get()
...
# process the payloads...
...
>>> queue.ack(uid_good) # done with that one
>>> queue.fail(uid_bad) # something didn't work out with that one, let's requeue
>>> uid, payload = queue.get() # get again; we get the requeued payload again
>>> uid == uid_bad
True
...
# try again
...
>>> queue.ack(uid) # now it worked; ACK the stuff now
SafeRedisQueue.get accepts a timeout parameter:
- 0 (the default) blocks forever, waiting for items
- a positive number blocks that amount of seconds
- a negative timeout disables blocking
SafeRedisQueue accepts *args, **kwargs
and passes them to
redis.StrictRedis, so use whatever you need.
Three exceptions, use these in the keyword arguments to configure SafeRedisQueue itself:
- url
Shortcut to use instead of a host/port/db/password combinations. Accepts "redis URLs" just as the redis library does, for example:
- redis://[:password]@localhost:6379/0
- unix://[:password]@/path/to/socket.sock?db=0
When using this keyword parameter, all positional arguments (usually one the host) are ignored. Those two are equivalent:
SafeRedisQueue('localhost', port=6379, db=7)
SafeRedisQueue(url='redis://localhost:6379/7')
- name
A prefix used for the keys in Redis. Default: "0", which creates the following keys in your Redis DB:
- srq:0:items
- srq:0:queue
- srq:0:ackbuf
- srq:0:backup
- srq:0:backuplock
- autoclean_interval
An interval in seconds (default: 60) at which unacknowledged items are requeued automatically. (They are moved from the internal ackbuf and backup data structures to the queue again.)
Pass
None
to disable autocleaning. It's enabled by default!- serializer
An optional serializer to use on the items. Default: None
Feel free to write your own serializer. It only requires a
dumps
andloads
methods. Modules likepickle
,json
,simplejson
can be used out of the box.Note however, that when using Python 3 the
json
module has to be wrapped as it by default does not handle thebytes
properly that is emitted by the underlying redis.py networking code. This should work:class MyJSONSerializer(object): @staticmethod def loads(bytes): return json.loads(bytes.decode('utf-8')) @staticmethod def dumps(data): return json.dumps(data) queue = SafeRedisQueue(name='foobar',serializer=MyJSONSerializer)
Added in version 3.0.0
For quick'n'dirty testing, you can use the script from the command line to put stuff into the queue:
$ echo "Hello World" | python saferedisqueue.py producer
...and get it out again:
$ python saferedisqueue.py consumer cbdabbc8-1c0f-4eb0-8733-fdb62a9c0fa6 Hello World