README.md

🧞🐍 Genny Python 🧞🐍

Genny uses Python for glue with other systems and for doing pre/post-processing of workloads.

Quick-Start

Genny uses the .python-version file to determine which version of Python to use. Consider using pyenv.

brew update && brew install pyenv
pyenv install # uses .python-version to determine python version
pyenv rehash

Once a suitable python is installed, setup a virtual environment and install the genny package:

pip install virtualenv
virtualenv venv
source ./venv/bin/activate
pip install .

Or if you wish to install and develop the scripts locally at the same time, you can use setup.py develop to avoid having to install after every change:

python setup.py develop

Script: lint-yaml

lint-yaml

Lint most of the YAML files in the Genny repository, including workloads, phases, and evergreen.yml.

Make you run the script from the root of the Genny repository (so it can pick up on the .yamllint linter config file)

Script: genny-auto-tasks

Generates JSON that can be used as input to evergreen's generate.tasks, representing either evergreen tasks or buildvariants that run genny workloads.

If run with --generate-all-tasks, genny-auto-tasks outputs a JSON file of task definitions for all workloads in the /src/workloads genny directory. If run with --variants, this program outputs a JSON file of buildvariant definitions that run a subset of workload tasks based on the other arguments. These two invocations are separate to ensure that no task will be generated more than once when run on evergreen (which would cause generate.tasks to throw an error, even across different buildvariants).

Thus, running genny-auto-tasks is generally a two step process. First, one generates all tasks as follows, with the output JSON going to build/all_tasks.json in this case:

genny-auto-tasks --generate-all-tasks --output build/all_tasks.json

Upon using generate.tasks with the outputted JSON, one can then generate JSON representing buildvariants that run the proper subset of tasks, based either on git modification history or the evergreen environment (See "Patch-Testing Genny Changes with Sys-Perf / DSI" in the main README for more info). For example, these commands write JSON definitions of linux-1-node-replSet buildvariants that run modified/autorun genny workloads.

genny-auto-tasks --modified --variants linux-1-node-replSet --output build/variants.json
genny-auto-tasks --autorun --variants linux-1-node-replSet --output build/variants.json

Script: genny-metrics-legacy-report

Produces a summary in legacy JSON format for the Evergreen perf plugin.

./genny run -o metrics.csv -m cedar-csv InsertRemove.yml
genny-metrics-legacy-report metrics.csv

Script: genny-metrics-report

Send metrics to Evergreen's metrics collection service (Cedar).

This code snippet is provided for reference only as Cedar does not yet support local environments.

./genny run -o metrics.csv -m cedar-csv InsertRemove.yml
# requires expansions.yml in cwd for cedar parameters
genny-metrics-report metrics.csv

Running Self-Tests

pip install virtualenv
virtualenv venv
source ./venv/bin/activate
pip install .
python setup.py nosetests

Python Formatting

Genny uses yapf for python formatting.

pip install virtualenv
virtualenv venv
source ./venv/bin/activate
pip install .
yapf -i --recursive genny tests