A python client for etcd

etcd, raft, distributed, log, api, client
pip install python-etcd==0.4.5


python-etcd documentation

A python client for Etcd

Official documentation:



This version of python-etcd will only work correctly with the etcd server version 2.0.x or later. If you are running an older version of etcd, please use python-etcd 0.3.3 or earlier.

This client is known to work with python 2.7 and with python 3.3 or above. It is not tested or expected to work in more outdated versions of python.

From source

$ python install


The basic methods of the client have changed compared to previous versions, to reflect the new API structure; however a compatibility layer has been maintained so that you don't necessarily need to rewrite all your existing code.

Create a client object

import etcd

client = etcd.Client() # this will create a client against etcd server running on localhost on port 4001
client = etcd.Client(port=4002)
client = etcd.Client(host='', port=4003)
client = etcd.Client(host=(('', 4001), ('', 4002), ('', 4003)))
client = etcd.Client(host='', port=4003, allow_redirect=False) # wont let you run sensitive commands on non-leader machines, default is true
# If you have defined a SRV record for pointing to the clients
client = etcd.Client(srv_domain='', protocol="https")
# create a client against
client = etcd.Client(host='', protocol='https', port=443, version_prefix='/etcd')

Write a key

client.write('/nodes/n1', 1)
# with ttl
client.write('/nodes/n2', 2, ttl=4)  # sets the ttl to 4 seconds
client.set('/nodes/n2', 1) # Equivalent, for compatibility reasons.

Read a key'/nodes/n2').value'/nodes', recursive = True) #get all the values of a directory, recursively.

# raises etcd.EtcdKeyNotFound when key not found
except etcd.EtcdKeyNotFound:
    # do something
    print "error"

Delete a key


Atomic Compare and Swap

client.write('/nodes/n2', 2, prevValue = 4) # will set /nodes/n2 's value to 2 only if its previous value was 4 and
client.write('/nodes/n2', 2, prevExist = False) # will set /nodes/n2 's value to 2 only if the key did not exist before
client.write('/nodes/n2', 2, prevIndex = 30) # will set /nodes/n2 's value to 2 only if the key was last modified at index 30
client.test_and_set('/nodes/n2', 2, 4) #equivalent to client.write('/nodes/n2', 2, prevValue = 4)

You can also atomically update a result:

result ='/foo')
print(result.value) # bar
result.value += u'bar'
updated = client.update(result) # if any other client wrote '/foo' in the meantime this will fail
print(updated.value) # barbar

Watch a key'/nodes/n1', wait = True) # will wait till the key is changed, and return once its changed'/nodes/n1', wait = True, timeout=30) # will wait till the key is changed, and return once its changed, or exit with an exception after 30 seconds.'/nodes/n1', wait = True, waitIndex = 10) # get all changes on this key starting from index 10'/nodes/n1') #equivalent to'/nodes/n1', wait = True)'/nodes/n1', index = 10)

Refreshing key TTL

(Since etcd 2.3.0) Keys in etcd can be refreshed without notifying current watchers.

This can be achieved by setting the refresh to true when updating a TTL.

You cannot update the value of a key when refreshing it.

client.write('/nodes/n1', 'value', ttl=30)  # sets the ttl to 30 seconds
client.refresh('/nodes/n1', ttl=600)  # refresh ttl to 600 seconds, without notifying current watchers

Locking module

# Initialize the lock object:
# NOTE: this does not acquire a lock yet
client = etcd.Client()
# Or you can custom lock prefix, default is '/_locks/' if you are using HEAD
client = etcd.Client(lock_prefix='/my_etcd_root/_locks')
lock = etcd.Lock(client, 'my_lock_name')

# Use the lock object:
lock.acquire(blocking=True, # will block until the lock is acquired
      lock_ttl=None) # lock will live until we release it
lock.is_acquired  # True
lock.acquire(lock_ttl=60) # renew a lock
lock.release() # release an existing lock
lock.is_acquired  # False

# The lock object may also be used as a context manager:
client = etcd.Client()
with etcd.Lock(client, 'customer1') as my_lock:
    my_lock.is_acquired  # True
my_lock.is_acquired  # False

Get machines in the cluster


Get leader of the cluster


Generate a sequential key in a directory

x = client.write("/dir/name", "value", append=True)
print("generated key: " + x.key)
print("stored value: " + x.value)

List contents of a directory

#stick a couple values in the directory
client.write("/dir/name", "value1", append=True)
client.write("/dir/name", "value2", append=True)

directory = client.get("/dir/name")

# loop through directory children
for result in directory.children:
  print(result.key + ": " + result.value)

# or just get the first child value

Development setup

To create a buildout,

$ python
$ bin/buildout

to test you should have etcd available in your system path:

$ bin/test

to generate documentation,

$ cd docs
$ make

Release HOWTO

To make a release

  1. Update release date/version in NEWS.txt and
  2. Run 'python sdist'
  3. Test the generated source distribution in dist/
  4. Upload to PyPI: 'python sdist register upload'