Cyanide - Celery stress testing and integration test support.

Version:1.3.0
Web:https://cyanide.readthedocs.io/
Download:http://pypi.python.org/pypi/cyanide/
Source:http://github.com/celery/cyanide/
Keywords:celery, stress, integration, functional, testing

Introduction

This stress test suite will attempt to break the Celery worker in different ways, and can also be used to write new stress test suites for projects depending on Celery infrastructure.

The worker must currently be started separately, and it’s encouraged to repeat the suite using workers started with different configuration values.

Ideas include:

  1. Default, single process:

    $ celery -A cyanide worker -c 1
    
  2. Default, multiple processes:

    $ celery -A cyanide worker -c 8
    
  3. Frequent maxtasksperchild recycling, single child process:

    $ celery -A cyanide worker -c 1 --maxtasksperchild=1
    
  4. Frequent autoscale scale down & maxtasksperchild, single child process:

    $ AUTOSCALE_KEEPALIVE=0.01 celery -A cyanide worker \
    >   --autoscale=1,0 --maxtasksperchild=1
    
  5. Frequent maxtasksperchild, multiple child processes:

    $ celery -A cyanide worker -c 8 --maxtasksperchild=1
    
  6. Processes terminated by time limits:

    $ celery -A cyanide worker --time-limit=1
    
  7. Frequent maxtasksperchild, single child process with late ack:

    $ celery -A cyanide worker -c1 --maxtasksperchild=1 -Z acks_late
    
  8. Worker using the eventlet pool:

    Start the worker, here having a thousand green-threads:

    $ celery -A cyanide worker -c1000 -P eventlet
    

    You must activate the green test group when starting the test suite:

    $ celery cyanide -g green
    
  9. Worker using the gevent pool:

    Start the worker, here having a thousand green-threads:

    $ celery -A cyanide worker -c1000 -P gevent
    

    You must activate the green test group when starting the test suite:

    $ celery cyanide -g green
    

Tips

It’s a good idea to include the --purge argument to clear out tasks from previous runs.

Note that the stress client will probably hang if the test fails, so this test suite is currently not suited for automatic runs.

Configuration Templates

You can select a configuration template using the -Z command-line argument to any celery -A cyanide command or the celery cyanide command used to execute the test suite.

The templates available are:

  • default

    Using AMQP as a broker, RPC as a result backend, and using JSON serialization for task and result messages.

    Both broker and result store is expected to run at localhost.

  • vagrant1

    Use the VM started by celery vagrant up as the broker and result backend (RabbitMQ).

  • vagrant1_redis

    Use the VM started by celery vagrant up as the broker and result backend (Redis).

  • redis

    Using Redis as a broker and result backend.

  • redistore

    Using Redis as a result backend only.

  • acks_late

    Enables late ack globally.

  • pickle

    Using pickle as the serializer for tasks and results (also allowing the worker to receive and process pickled messages)

  • confirms

    Enables RabbitMQ publisher confirmations.

  • events

    Configure workers to send task events.

  • proto1

    Use version 1 of the task message protocol (pre 4.0)

You can see the resulting configuration from any template by running the command:

$ celery -A cyanide report -Z redis

Examples

Example running the stress test using the redis configuration template:

$ cyanide -Z redis

Example running the worker using the redis configuration template:

$ celery -A cyanide worker -Z redis

You can also mix several templates by providing a comma-separated list:

$ celery -A cyanide worker -Z redis,acks_late

In this example (redis,acks_late) the redis template will be used as main configuration, and then the additional keys from the acks_late template will be merged as changes.

Test Suite Options

After one or more worker instances are running, you can start executing the tests.

By default the complete test suite will be executed:

$ celery cyanide

You can also specify what test cases to run by providing one or more names as arguments:

$ celery cyanide revoketermfast revoketermslow

A full list of test case names can be retrieved with the -l switch:

$ celery cyanide -l
.> 1) chain,
.> 2) chaincomplex,
.> 3) parentids_chain,
.> 4) parentids_group,
.> 5) manyshort,
.> 6) unicodetask,
.> 7) always_timeout,
.> 8) termbysig,
.> 9) timelimits,
.> 10) timelimits_soft,
.> 11) alwayskilled,
.> 12) alwaysexits,
.> 13) bigtasksbigvalue,
.> 14) bigtasks,
.> 15) smalltasks,
.> 16) revoketermfast,
.> 17) revoketermslow

You can also start from an offset within this list, e.g. to skip the first two tests use --offset=2:

$ celery cyanide --offset=2

See celery cyanide --help for a list of all available command-line options.

Vagrant

Starting

Cyanide ships with a complete virtual machine solution to run your tests. The image ships with Celery, Cyanide, RabbitMQ and Redis and can be deployed simply by running the celery vagrant command:

$ celery vagrant up

The IP address of the new virtual machine will be 192.168.33.123, and you can easily tell both the worker and cyanide test suite to use it by specifying the vagrant1 (RabbitMQ) or vagrant1_redis templates:

$ celery -A worker -Z vagrant1
$ celery cyanide -Z vagrant1

SSH

To open an SSH session with the virtual machine after starting with celery vagrant up do:

$ ssh $(celery vagrant sshargs)

Stopping

To shutdown the virtual machine run the command:

$ celery vagrant halt

To destroy the instance run the command:

$ celery vagrant destroy

Note

To completely wipe your instance you need to remove the .vagrant directory.

The location of this directory can be retrieved by executing the following:

$ celery vagrant statedir
/opt/devel/cyanide/cyanide/vagrant/.vagrant

You can combine this with rm to force removal of this directory:

$ rm -rf $(celery vagrant statedir)

Environment Variables

CYANIDE_TRANS

If the CYANIDE_TRANS environment variable is set the stress test suite will use transient task messages instead of persisting messages to disk.

To avoid declaration collision the cstress.trans queue name will be used when this option is enabled.

CYANIDE_BROKER

You can set the CYANIDE_BROKER environment variable to change the default broker used:

$ CYANIDE_BROKER='amqp://' celery -A cyanide worker # …
$ CYANIDE_BROKER='amqp://' celery cyanide

CYANIDE_BACKEND

You can set the CYANIDE_BACKEND environment variable to change the result backend used:

$ CYANIDE_BACKEND='amqp://' celery -A cyanide worker # …
$ CYANIDE_BACKEND='amqp://' celery cyanide

CYANIDE_QUEUE

A queue named c.stress is created and used by default for all task communication.

You can change the name of this queue using the CYANIDE_QUEUE environment variable:

$ CYANIDE_QUEUE='cyanide' celery -A cyanide worker # …
$ CYANIDE_QUEUE='cyanide' celery cyanide

CYANIDE_PREFETCH

The CYANIDE_PREFETCH environment variable sets the default prefetch multiplier (default value is 10).

AWS_REGION

The AWS_REGION environment variable changes the Amazon AWS region to something other than the default us-east-1, to be used with the sqs template.

Custom Suites

You can define custom suites (look at source code of cyanide.suites.default for inspiration), and tell cyanide to use that suite by specifying the celery cyanide -S option:

$ celery cyanide -S proj.funtests:MySuite