Post on 22-Jan-2015
description
transcript
A Python Petting ZooPython and ZooKeeper
Devon JonesSenior Software Engineer
knewt.ly/pettingzoo-slides
Knewton is an education technology company with a goal of bringing adaptive education to the masses. Knewton makes it possible to break courses into tiny parts that are delivered to each student as personalized, real-time recommendations. Knewton recommends the best work for an individual learner by calculating data on what we know about a student, similar students, the learning objective, and the content itself at a given point in time. The Knewton platform today has tens of thousands of students, will have over 600k students starting Sept. 2012, and will soon have millions of students.
What is this about?
● What is ZooKeeper, how is it useful● State of ZooKeeper on python● The release of PettingZoo, Knewton's
ZooKeeper recipes for managing a distributed machine learning cluster
What is ZooKeeper?
ZooKeeper is a centralized service for maintaining configuration information, naming, providing distributed synchronization, and providing group services. All of these kinds of services are used in some form or another by distributed applications. Each time they are implemented there is a lot of work that goes into fixing the bugs and race conditions that are inevitable. Because of the difficulty of implementing these kinds of services, applications initially usually skimp on them, which make them brittle in the presence of change and difficult to manage. Even when done correctly, different implementations of these services lead to management complexity when the applications are deployed.
According to ZooKeeper's Apache Site:
Right, but what is it?
What is ZooKeeper?
ZooKeeper is a distributed filesystem based on the PAXOS algorithim with a few valuable features.● No single point of failure● Strictly ordered, observable state● Events● Sequential and ephemeral primitives● ACLs
PAXOShttp://groups.csail.mit.edu/tds/papers/DePrisco/WDAG97.pdf● Distributed Finite State Machine● Pairwise connections among hosts● Rounds of Consensus Propositions
conducted by leader● Time bound contracts on interactions● Useful to have internal concurrency
contracts
What is ZooKeeper for?
ZooKeeper is a platform for creating protocols for synchronization of distributed systems. Uses include:● Distributed configuration● Queues● Implementation of distributed concurrency
primitives such as locks, barriers, latches, counters, etc.
In short, it's a system for managing shared state between distributed systems.
Consistency Guarantees
ZooKeeper makes a number of promises:● Sequential Consistency● Atomicity● Single System Image● Reliability● Timeliness
Watches
● Can be set by any read operation● Will fire an event to the client who set them
once and only once● Can be set on nodes or on their data.● Will always be sent to you in a fixed order
Sequential Nodes
● Appends a monotonically increasing number to the end of the znode (file)
● Can be used directly for leader election● Provides communication of server's view of
ordering
Ephemeral Nodes
● Only exist for the life of a connection● If your connection does not respond to a
keepalive request, will disappear● Used to ensure reliability against service
disruption for most recipes● Used to trigger events, such as
reconfiguration if a service goes down that published discovery configs
Recipes
ZooKeeper's documentation explains how to implement a number of recipes:● Barriers● Locks● Queues● Counters● Two Phase Commit● Leader Election
Recipes
As a community, we need well tested versions of these recipes as well as other valuable protocols built on ZooKeeper
State of ZooKeeper
● Very low level● Coding for it is very complex● Lots of edge cases● ZooKeeper needs high level, well tested
libraries● Very few complete, high level solutions exist
ZooKeeper Libraries
The first high level library with significant recipe implementations has emerged from Netflix. It's name is Curator. Unfortunately for us, it's in Java, not Python.
Curator
https://github.com/Netfix/curator● State
○ High level api○ Non-Resilient Client○ Documentation○ Tests, Embeddable ZooKeeper for testing
● Recipes○ Leader Latch, Leader Election○ Multiple Locks and Semaphores○ Multiple Queues○ Barrier, Double Barrier○ Shared Counter/Distributed Atomic Long
State of ZooKeeper on Python
● Current state is in a lot of flux● Went from only low level bindings to a
number of incomplete bindings in first half of 2012
● So far nothing like Curator has emerged (but it appears to be brewing)
One of the top ranked results for 'Python ZooKeeper'
Summary of Python ZooKeeper bindings
● There are about 10 presently● Many suffer from not handling known edge
cases in ZooKeeper● Some suffer problems with resilient
connections● The following is derived from the python
ZooKeeper binding census of Ben Bangert
Official Bindings
● State○ Complete access to the ZooKeeper C bindings○ Full of sharp edges○ Not a resilient client○ No recipes○ Threads communication with ZooKeeper in a C
thread○ Foundation for most other libraries○ Very low level
Bindings
Twisted:● txzookeeper gevent:● gevent-zookeeper● kazoo
Bindings (Cont)
High level clients:● zc.zk● kazoo● twitter zookeeper Others:● zkpython (low level)● pykeeper (not resiliant)● zoop (not resiliant)
Bindings (Cont)
Recipes only:● zktools● PettingZoo
State of ZooKeeper on Python:Kind of a Mess
A project exists to merge some of the high level bindings in an attempt to create a python equivalent of Curator: https://github.com/python-zk Started by Ben Bangert to merge Kazoo & zc.zk with an attempt to implement all Curator recipes.
PettingZoo
https://github.com/Knewton/pettingzoo-python● State
○ Relies on zc.zk○ Documented, doc strings○ Tests (mock ZooKeeper)○ All recipes implemented in a Java version as well
● Recipes○ Distributed Config○ Distributed Bag○ Leader Queue○ Role Match
PettingZoo
● In heavy development● Distributed Discovery, Distributed Bag are
well tested and used in production● Leader Queue and Role Match are tested,
but undeployed● PettingZoo will be ported to or merged with
the kazoo effort when it is ready
Our Problem
Need to be able to do stream processing of observations of student interactions with course material. This involves multiple models that have interdependent parameters. This requires:● Sharding along different axes dependent
upon the models● Subscriptions between models for
parameters● Dynamic reconfiguration of the environment
to deal with current load
Distributed Discovery
Allows services in a dynamic, distributed environment to be able to be quickly alerted of service address changes.● Most service discovery recipes only contain
host:port, Distributed Discovery can share arbitrary data as well (using yaml)
● Can handle load balancing through random selection of config
● Handles rebalancing on pool change
How does this help us scale?
● Makes discovery of dependencies simple● Adds to reliability of system by quickly
removing dead resources● Makes dynamic reconfiguration simple as
additional resources become available
Distributed Discovery Example: Write
from pettingzoo.discovery import write_distributed_config
from pettingzoo.utils import connect_to_zk
conn = connect_to_zk( 'localhost:2181')
config = {
'header': {
'service_class': 'kestrel', 'metadata': {
'protocol': 'memcached', 'version': 1.0
}
},
'host': 'localhost',
'port': 22133
}
write_distributed_config(conn, 'kestrel', 'platform', config)
Distributed Discovery Example: Read
from pettingzoo.discovery import DistributedMultiDiscovery
from pettingzoo.utils import connect_to_zk
conn = connect_to_zk( 'localhost:2181')
dmd = DistributedMultiDiscovery(conn)
conf = None
def update_configs(path, updated_configs):
conf.update(updated_configs)
conf = dmd.load_config( 'kestrel', 'platform', callback=update_configs)
Distributed Bag
Recipe for a distributed bag (dbag) that allows processes to share a collection. Any participant can post or remove data, alerting all others.
● Used as a part of Role Match● Useful for any case where
processes need to share configuration determined at runtime
How does this help us scale?
● Can quickly alert processes as to who is subscribing to them
● Reduces load by quickly yanking dead subscriptions
● Provides event based subscriptions, making implementation simpler
Distributed Bag
Tokens
<bag>
Items Item1
Token3
Item2Item
3
● Sequential items contain the actual data
● Can be ephemeral● Clients set delete watch
on discrete items● Token is set to id of
highest item● Clients set a child
watch on the "Tokens" node
● Can determine exact adds and deletes with a constant number of messages per delta
Distributed Bag Exampleimport yaml
from pettingzoo.dbag import DistributedBag
from pettingzoo.utils import connect_to_zk
...
conn = connect_to_zk( 'localhost:2181')
dbag = DistributedBag(conn, '/some/data')
docs = {}
def acb(cb_dbag, cb_id):
docs[cb_id] = cb_dbag.get(cb_id)
def rcb(cb_dbag, cb_id):
docs.remove(cb.id)
dbag.add_listeners(add_callback=acb, remove_callback=rcb)
add_id = dbag.add(yaml.load(document), ephemeral=True)
docs[add_id] = document
Leader Queue
Recipe is similar to Leader Election, but makes it easy to monitor your spare capacity.● Used in Role Match● As services are ready to do work, they
create an ephemeral, sequential node in the queue.
● Any member always knows if either they are in the queue or at the front
● Watch lets leader know when it is elected
How does this help us scale?
● Gives a convenient method of assigning work
● Makes monitoring current excess capacity easy
Leader Queue
● Candidates register with sequential, ephemeral nodes
● Candidate sets delete watch on predecessor
● Candidate is elected when it is the smallest node
● When elected, candidate takes over its new role
● When ready, candidate removes itself from the queue
● Only one candidate needs to call get_children upon any node exiting
<queue> C_3
C_1
C_4
Leader Queue Usage Examplefrom pettingzoo.leader_queue import LeaderQueue, Candidate
from pettingzoo.utils import connect_to_zk
class SomeCandidate(Candidate):
def on_elected(self):
<do something sexy>
conn = connect_to_zk( 'localhost:2181')
leaderq = LeaderQueue(conn)
leaderq.add_candidate(SomeCandidate())
Role Match
Allows systems to expose needed, long lived jobs, and for services to take over those jobs until all are filled.● Dbag used to expose jobs● Leader queue used to hold applicants● Records which jobs are presently held with
ephemeral node● Lets a new process take over if a worker
dies● We use it for sharding/segmentation to
dynamically adjust the shards as needed due to load
How does this help us scale?
● Core of our ability to dynamically adjust shards
● Lets the controlling process adjust problem spaces and have those tasks become automatically filled
● Monitoring is easy to identify who is working on what, when
Role Match
● Leader monitors for open jobs
● Job holder creates an ephemeral assignment
● Assignment id matches job id, indicating that it is claimed
assignment
<match>
Assgn1A_2
Leader Queueapplicant
job DistributedBag
Future: Distributed Config
Next project is Distributed Config.● Allows service config to be recorded and
changed with a yaml config● Every process that connects creates a child
node of the appropriate service● Any change in a child node's config
overrides the overall service config for that process
● Any change of the parent or child fires a watch to let the process know that it's config has changed
Questions?
devon@knewton.com
knewt.ly/pettingzoo-slides
knewt.ly/pettingzoo
Appendix: Official Bindings
● State○ Complete access to the ZooKeeper C bindings○ Full of sharp edges○ Not a resilient client○ No recipes○ Threads communication with ZooKeeper in a C
thread○ Foundation for most other libraries○ Very low level
Appendix: zkpython
https://github.com/duncf/zkpython/improvements to a fork of the official bindings● Status
○ Resilient Client○ No docs beyond the official bindings○ Good tests
● Recipes○ Basic Lock (Using unique id rather than UUID)
Appendix: pykeeper
https://github.com/nkvoll/pykeeper● State
○ Non-resilient Client (not resilient to errors)○ High level client○ Documented, No doc strings○ Tests (Requires running ZooKeeper)
Appendix: zoop
https://github.com/davidmiller/zoop● State
○ Doesn't handle node create edge-case○ Doesn't handle retryable exceptions○ Documented, doc strings○ Tests (Requires running ZooKeeper)
● Recipes○ Revocable Lock (Doesn't handle create node edge-
case, uses a permanent node instead of ephemeral)
Appendix: txzookeeper
https://launchpad.net/txzookeeper● State
○ Resilient Client○ Doc strings, no additional documentation○ Supports twisted (only)○ Well tested
● Recipes○ Basic Lock (Not using UUID)○ Queue○ ReliableQueue○ SerializedQueue
Appendix: twitter zookeeperhttps://github.com/twitter/commons/tree/master/src/python/twitter/common/zookeeper
● State○ Resilient Client○ Handles node create edge-case○ Some documentation○ Well Tested○ Tied to a lot of twitter commons code, difficult to
extract● Recipes
○ Service Registration/Discovery
Appendix: gevent-zookeeper
https://github.com/jrydberg/gevent-zookeeper/● State
○ Supports gevent○ No documentation○ No tests
Appendix: Kazoo
https://github.com/nimbusproject/kazoo● Status
○ Resilient Client○ Doc strings, no additional documentation○ Supports gevent○ Minimal tests
● Recipes○ Basic Lock (Uses UUID properly)
Appendix: zc.zk
https://github.com/python-zk/zc.zk● State
○ Resilient Client○ High level client○ Higher level automatic watch functionality○ Higher level restoration of ephemeral nodes○ Tests (Mock ZooKeeper)○ Documentation, no doc strings
● Recipes○ Service Registration/Discovery
Appendix: zktools
https://github.com/mozilla-services/zktools● State
○ Relies on zc.zk○ Documented, doc strings○ Tests (Requires running ZooKeepe)
● Recipes○ Shared Read/Write Locks○ AsyncLock○ Revocable Locks
Appendix: PettingZoo
https://github.com/Knewton/pettingzoo-python● State
○ Relies on zc.zk○ Documented, doc strings○ Tests (mock ZooKeeper)○ All Recipes implemented in a java version as well
● Recipes○ Distributed Config○ Distributed Bag○ Leader Queue○ Role Match