Skip to main content

A network mutex service for distributed environments.

Project description

Distributex

Distributex is a simple mutex service for coordinating certain cluster operations.

Note: Distributex is not designed for tasks which require high performance or fair lock acquisition. It is a very simple Busy-waiting lock with very slow acquisition. Do not use it for extremely large clusters either as there’s a good chance a requester might never obtain a lock.

Distributex server

The Distributex server provides a simple HTTP service. It is written using Twisted, and provides a Twisted plugin. It also requires PyYAML for its configuration.

You can start it as follows, or wrap it in supervisor, or pass -d, or whatever:

$ twistd -n distributex -c distributex.yaml

The configuration file is a simple YAML structure which defines lock pools. A lock pool is a resource you want to allow things to fight over.:

backend: memcache
pools:
    - name: pool1
      expire: 300

    - name: pool2
      expire: 300
      servers: acme1, acme2

This will create two pools whose lock expires after 5 minutes. It’s generally a good idea to set an expiry to ensure something, otherwise it will default to 30 minutes. If you don’t want it to expire then set it to 0, but I don’t recommend that.

You can specify either the ‘memcache’ backend or ‘inmemory’, there are pros and cons to both. Memcache will be slower, but state is retained away from the Distributex server and you can scale out workers - however since the inmemory backend can handle about 5000 waiting locks on a single machine redundancy is the only real concern. Lock expiry is also a bit more reliable and simpler in the memcache backend.

A comma separated list of servers can also be provided to prevent accidental pool incursions. This isn’t secure, nor are lock releases, since anyone can just forge their hostname. The distributex client will pass the FQDN of the host.

It is also possible to set ‘maxlocks’ which allows the pool to behave like a semaphore.

You can test the service as follows:

$ curl "http://localhost:9889/get/?host=me&pool=pool1"
YES
$ curl "http://localhost:9889/get/?host=them&pool=pool1"
NO
$ curl "http://localhost:9889/release/?host=me&pool=pool1"
OK
$ curl "http://localhost:9889/get/?host=them&pool=pool1"
YES
$ curl "http://localhost:9889/release/?host=them&pool=pool1"
OK

The service also provides a ‘wait’ command which will leave the connection open until a lock is obtained.

Distributex client

Not included in this package is a simple Python script is provided to wrap commands. You can get it from https://raw.github.com/calston/distributex/master/client/distributex

usage: distributex [-h] -H HOST -r POOL [-p PORT] [command]

Distributex client

positional arguments:
  command     Command to execute when lock is obtained

optional arguments:
  -h, --help  show this help message and exit
  -H HOST     Server hostname
  -r POOL     Resource pool
  -p PORT     Server port (default 9889)
  -l          Use local locking as well

This is useful for blocking a cron job like Puppet.:

*/5 * * * * /usr/bin/distributex -H distributex.acme.com -r pool1 '/usr/bin/puppet agent --onetime --no-daemonize'

This will ensure that only one instance of Puppet in the cluster runs at any time. You might also want to pass -l to distributex to prevent local process overlap.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

distributex-0.7.tar.gz (4.4 kB view details)

Uploaded Source

File details

Details for the file distributex-0.7.tar.gz.

File metadata

  • Download URL: distributex-0.7.tar.gz
  • Upload date:
  • Size: 4.4 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No

File hashes

Hashes for distributex-0.7.tar.gz
Algorithm Hash digest
SHA256 73fda756b1829ff41b1fd5eb0009e24b5b5a1d55460565b6a9ba4283f83c65d2
MD5 d236950b3082e759c46b0c0107d8844e
BLAKE2b-256 120ea6636a0a6b04b3db049aac3ba373eb2472ccc2fd8f1c3ae0c3ae2202d6fc

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page