Last updated:
0 purchases
pytestcloudist 0.1.0
pytest-cloudist
pytest-cloudist is a pytest plugin to that distributes your tests to AWS EC2 machines with a minimum of fuss. It is a thin wrapper around Meadowrun, which does the heavy lifting of starting EC2 instances and synchronizing environment and code.
Installation
Simply pip install alongside pytest:
python -m pip install pytest-cloudist
That makes the following command line arguments on pytest available:
Distributed testing in the cloud using Meadowrun:
--cloudist={test,file,no}
Set mode for distributing tests to workers.
test: send each test to a worker separately.
file: send each test file to a worker separately.
(default) no: run tests inprocess, don't distribute.
--cd-tasks-per-worker-target=tasks_per_worker_target
The number of tasks to target per worker. This number determines whether individual tests or files are grouped and sent as a chunk to the test worker. Chunking is normally more efficient, but may affect load balancing and worsen the effect of stragglers.
--cd-num-workers=num_workers
Number of workers to use for distributed testing.
--cd-cpu-per-worker=logical_cpu_per_worker
The number of logical CPUs needed per worker.
--cd-memory-per-worker=memory_gb_per_worker
The amount of memory (in GiB) needed per worker.
--cd-interrupt-prob=INTERRUPTION_PROBABILITY_THRESHOLD
The estimated probability that spot instances are interrupted by AWS. Set to 0 for on-demand instances, which will be up to 10x more expensive.
--cd-extra-files=EXTRA_FILES
Extra files to copy as to the remote machines, if needed. .py files on sys.path are copied automatically.
--cd-init-command=INIT_COMMAND
Initialization command to run once per worker
Usage
By default, pytest-cloudist is not activated, i.e. your tests run locally as normal. To enable pytest-cloudist, pass either --cloudist test or --cloudist file with any other options.
Credits
The code and approach of pytest-cloudist, in terms of pytest integration, are heavily based on the code for pytest-xdist, and as a result it is also licensed as MIT.
For personal and professional use. You cannot resell or redistribute these repositories in their original state.
There are no reviews.