Yet another parallel test runner, built using a centralized queue to ensure optimal distribution of tests between workers.
Specifically optimized for CI environments: build statistics from each run are stored locally and used to sort the queue at the beginning of the next run.
test-queue uses a simple master + pre-fork worker model. The master exposes a unix domain socket server which workers use to grab tests off the queue.
─┬─ 21232 minitest-queue master
├─── 21571 minitest-queue worker [3] - AuthenticationTest
├─── 21568 minitest-queue worker [2] - ApiTest
├─── 21565 minitest-queue worker [1] - UsersControllerTest
└─── 21562 minitest-queue worker [0] - UserTest
test-queue also has a distributed mode, where additional masters can share the workload and relay results back to a central master.
TEST_QUEUE_WORKERS
: number of workers to use per master (default: all available cores)TEST_QUEUE_VERBOSE
: show results as they are available (default:0
)TEST_QUEUE_SOCKET
: unix socketpath
(or tcpaddress:port
pair) used for communication (default:/tmp/test_queue_XXXXX.sock
)TEST_QUEUE_RELAY
: relay results back to a central master, specified as tcpaddress:port
TEST_QUEUE_STATS
:path
to cache build stats in-build CI runs (default:.test_queue_stats
)TEST_QUEUE_FORCE
: comma separated list of suites to runTEST_QUEUE_RELAY_TIMEOUT
: when using remote workers, the amount of time a worker will try to reconnect to start workTEST_QUEUE_RELAY_TOKEN
: when using remote workers, this must be the same on both workers and the server for remote workers to run tests.TEST_QUEUE_SLAVE_MESSAGE
: when using remote workers, set this on a slave worker and it will appear on the slave's connection message on the master.TEST_QUEUE_SPLIT_GROUPS
: split tests up by example rather than example group. Faster for tests with short setup time such as selenium. RSpec only. Add the :no_split tag to ExampleGroups you don't want split.
test-queue bundles testunit-queue
, minitest-queue
and rspec-queue
binaries which can be used directly:
$ minitest-queue $(find test/ -name \*_test.rb)
$ rspec-queue --format progress spec
But the underlying TestQueue::Runner::TestUnit
, TestQueue::Runner::MiniTest
and TestQueue::Runner::RSpec
are
built to be subclassed by your application. I recommend checking a new
executable into your project using one of these superclasses.
$ vim script/test-queue
$ chmod +x script/test-queue
$ git add script/test-queue
Since test-queue uses fork(2)
to spawn off workers, you must ensure each worker
runs in an isolated environment. Use the after_fork
hook with a custom
runner to reset any global state.
#!/usr/bin/env ruby
class MyAppTestRunner < TestQueue::Runner::MiniTest
def after_fork(num)
# use separate mysql database (we assume it exists and has the right schema already)
ActiveRecord::Base.configurations['test']['database'] << num.to_s
ActiveRecord::Base.establish_connection(:test)
# use separate redis database
$redis.client.db = num
$redis.client.reconnect
end
def prepare(concurrency)
# create mysql databases exists with correct schema
concurrency.times do |i|
# ...
end
# If this is a remote slave, tell the master something about us
@slave_message = "Output for slave 123: https://myhost.com/build/123"
end
def around_filter(suite)
$stats.timing("test.#{suite}.runtime") do
yield
end
end
end
MyAppTestRunner.new.execute
To use distributed mode, the central master must listen on a tcp port. Additional masters can be booted in relay mode to connect to the central master. Workers must provide a TEST_QUEUE_RELAY_TOKEN to match the master's.
$ TEST_QUEUE_RELAY_TOKEN=123 TEST_QUEUE_SOCKET=0.0.0.0:12345 bundle exec minitest-queue ./test/sample_test.rb
$ TEST_QUEUE_RELAY_TOKEN=123 TEST_QUEUE_RELAY=0.0.0.0:12345 bundle exec minitest-queue ./test/sample_test.rb
$ TEST_QUEUE_RELAY_TOKEN=123 ./test-multi.sh
See the Parameterized Trigger Plugin for a simple way to do this with jenkins.