Copyright The Linux Foundation and each contributor to CommunityBridge.
SPDX-License-Identifier: CC-BY-4.0
- Go 1.14.x
- Python 3
- Node 8/12+
NodeJS is used by serverless to facilitate deployment and location testing. In all but few occasions, we use NodeJS v12+.
The frontend UI still requires an older version of NodeJS - version 8. This is
due to compatibility issues with the Angular and Ionic toolchain for the legacy
UI. Use node version 8.x for the frontend source package installation
(typically the src
folder) and build/deploy commands.
For all other folders which use node, use version 12.x+.
In order to quickly switch between node versions, we recommend you use Node Version Manager - nvm. The CircleCI build configuration uses this approach to switch between node versions within the build andx deployment.
These are the steps for setting up a local dev environment for the python backend (work in progress). The legacy Python backend is used for the Gerrit and GitHub interaction, Docusign workflows, and a few other API calls.
Historical and Current Endpoints:
- /v1 (us-east-1)
- /v2 (us-east-1)
- /v1/salesforce/projects - handles project management console project listing queries (us-east-1)
- /v1/salesforce/project - handles project management console project queries (us-east-1)
- /v2/github/installation - handles github bot installation (us-east-1)
- /v2/github/activity - handles github activity callbacks to the product (us-east-1)
You will need to install a local Python 3.6+ runtime - Python 3.7 will work now
that we've updated the hug
library.
One easy way to install a Python runtime on a Mac is via the pipenv
tool.
Here, we'll use it to install a specific version of Python. There are other
approaches to installing Python. Feel free to add your preferred approach to the
information below.
brew install pipenv
To create a new Python installation using pipenv
, specify the specific version
using the --python VERSION
flag, like so:
pipenv --python 3.7
When given a Python version, like this, Pipenv will automatically scan your system for a Python that matches that given version and download one if necessary.
You can use pipenv
or any other tool to configure a local sandbox for testing.
We prefer to use virtualenv
.
pip3 install virtualenv
Now setup the virtual env using python 3.6+:
# Check your version and grab the appropriate path
python --version
which python
# Setup the virtual env
cd cla-backend
virtualenv --python=/usr/local/Cellar/python3/3.7.4/bin/python3 .venv
This will create a .venv
folder in the cla-backend
project folder. We'll
want to load the sandbox environment by using:
source .venv/bin/activate
If not already done, load the virtual environment. Then we can install the required modules:
source .venv/bin/activate
pip3 install -r requirements.txt
This will install the dependencies in the .venv
path.
We'll need a few run-time environment variables to communicate with AWS when running locally.
AWS_REGION
- the AWS region, normally this should be set tous-east-1
AWS_ACCESS_KEY_ID
- AWS key, used to authenticate to AWS for DynamoDB and SSMAWS_SECRET_ACCESS_KEY
- AWS secret key, used to authenticate to AWS for DynamoDB and SSMPRODUCT_DOMAIN
- e.g.export PRODUCT_DOMAIN=dev.lfcla.com
ROOT_DOMAIN
- e.g.export ROOT_DOMAIN=lfcla.dev.platform.linuxfoundation.org
Optional environment variables:
PORT
- optional, the HTTP port when running in the local mode. The default port is 5000.STAGE
- optional, specifies the environment stage. The default is typicallydev
.
For testing locally, you'll want to point to the dev environment or stand up a
local DynamoDB and S3 instance. Generally, we run the backend services and UI
locally and simply point to the DEV environment. The STAGE
environment
variable controls where we point. Make sure you export/provide/setup the AWS
properties in order to connect.
# ok to use node v12+
cd cla-backend
yarn install
This will install the serverless
tool and plugins.
# You can simply run this launcher:
# ok to use node v12+
yarn serve:dev
# Or run it using this approach - points to the DEV database/environment
node_modules/serverless/bin/serverless wsgi serve -s 'dev'
To test:
# Health Check
open http://localhost:5000/v2/health
# Get User by ID
open http://localhost:5000/v2/user/<some_uuid_from_users_table>
Current Endpoints:
- /v3 (us-east-1) - considered part of EasyCLA v1 which leverages the older (legacy) UI
- /v4 (us-east-2) - considered EasyCLA v2 which leverages the newer LFX UI and LFX Admin Consoles - includes integration with other platform services
- plus a number of support lambdas
Follow the Go Getting Started guide to install the tool.
After installation, you should have something similar to:
# After the GO installation, you should have GOROOT set
echo $GOROOT
/usr/local/opt/go/libexec
# And go should be in your path
go version
go version go1.14.6 darwin/amd64
To build the go stuff, the go tool is very specific on paths. If this isn't
correct, some of the build targets, such as swagger
will fail with a
nondescript error message. Generally:
# GOPATH also should be set, depends on where your code would generally be located
echo $GOPATH
# response would be something like this:
/Users/<my-os-user-id>/projects/go
# if `GOPATH` is not set, you will need to explictly set it:
export GOPATH=$HOME/projects/go # or wherever your go projects will be located
# most sane people set this in their `~/.bashrc` or `~/.zshrc` file so they
# don't have to remember to set it again
# If the `GOPATH` variable is NOT set, run: export GOPATH=<your_go_path>
# Inside of this folder you should traditionally have `src`, `bin`, and `pkg` folders
# To recreate them, run (needed for the first-time setup only):
pushd $GOPATH && mkdir -p src bin pkg && popd
# If you have code checked out in a different path than the GOPATH (as in the
# case of CLA, probably), you will need to establish a soft link to point where
# you checked out the public easycla source - this must follow the EasyCLA go path
# structure (e.g. the directory path of the source code on disk must match the
# code import paths). To create the path structure, run:
mkdir -p $GOPATH/src/github.com/communitybridge
pushd $GOPATH/src/github.com/communitybridge
ln -s <path_to_where_you_really_have_easycla_checked_out>/easycla easycla
popd
# Confirm everything:
ls -la $GOPATH/src/github.com/communitybridge/easycla
# Should see something like:
lrwxr-xr-x 1 ddeal staff 29 Jul 7 14:48 /Users/ddeal/projects/go/src/github.com/communitybridge/easycla@ -> /Users/ddeal/projects/easycla
EasyCLA currently relies on several private libraries for user authorization and common data models and other libraries:
- github.com/LF-Engineering/lfx-kit (authorization library)
- github.com/LF-Engineering/lfx-models (common data models)
- github.com/LF-Engineering/aws-lambda-go-api-proxy (forked library which includes query parameters fixes)
The near-term plan is to migrate some of these private libraries to the github.com/communitybridge organization as public repositories.
Until that time, each user needs to request access to these repositories
and create/update the following file: ~/.gitconfig
to include the following:
[url "ssh://git@github.com/"]
insteadOf = https://github.com/
This GoLang build file (Makefile) will handle pulling dependencies from the private repositories.
# Navigate to the go source code and build it - you must be in this path/directory
pushd $GOPATH/src/github.com/communitybridge/easycla/cla-backend-go
# First time only, you will need to install the dev tools. Simply run:
make setup
Once the tools are installed, you can run a clean build with all the bells and whistles. This will:
- remove the old binaries and support lambdas
- combine the swagger specification fragments into a single compiled swagger file (one for v1, one for v2) - python doesn't generate APIs from a swagger
- Runs swagger-go to generate the API and data models for both v1 and v2 APIs - validates swagger spec
- Downloads Org Service, Project Service, User Service and ACS Service swagger - generates the client stubs
- Downloads any external dependencies
- Builds API server and support lambdas
- formats the code (should be already formatted) - run this before you checkin code - the linter will catch violations
- build the code - use
build
orbuild-mac
based on your platform - test will run unit tests
- lint will run lint checks (do this before checking in code to avoid CI/CD catching the violations later)
Mac:
make all
# or
make all-mac
# or everything individually - including the extra lambdas
make clean swagger deps fmt build-mac build-aws-lambda-mac build-metrics-lambda-mac build-dynamo-events-lambda-mac build-zipbuilder-scheduler-lambda-mac build-zipbuilder-lambda-mac test lint
Linux:
make all-linux
# or everything individually - including the extra lambdas
make clean swagger deps fmt build-linux build-aws-lambda-linux build-metrics-lambda-linux build-dynamo-events-lambda-linux build-zipbuilder-scheduler-lambda-linux build-zipbuilder-lambda-linux test lint
After the above, you should have the binary now (Mac example):
ls -lhF cla
-rwxr-xr-x 1 ddeal staff 36M Jul 18 10:57 cla-mac*
The binary is based on your OS, Mac example:
file cla-mac
cla-mac: Mach-O 64-bit executable x86_64
Linux example:
file cla
cla: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), statically linked, Go BuildID=9KJXkKLbz8QXVJiIStug/13hRaqNdkbT0_CJAC_96/pqAvGfgbdkRS-xcMCFwk/PEhC2JMjFKBawWbErcth, stripped
To run the Go backend locally, it requires a few environment variables to be set:
AWS_REGION
- the AWS region, normally this should be set tous-east-1
AWS_ACCESS_KEY_ID
- AWS key, used to authenticate to AWS for DynamoDB and SSMAWS_SECRET_ACCESS_KEY
- AWS secret key, used to authenticate to AWS for DynamoDB and SSM
Optional environment settings:
PORT
- optional, the HTTP port when running in local mode. The default port is 8080.STAGE
- optional, specifies the environment stage. The default isdev
.GH_ORG_VALIDATION
- set tofalse
to test locally which will by-pass the GH auth checks and allow local functional tests (e.g. with cURL or Postman) - default is enabled/true
First build and setup the environment. Then simply run it:
# Mac
./cla-mac
# or linux
./cla
You should see the typical diagnostic details on startup indicating that it started without errors including build information with successful load of configuration parameters. To confirm the service is up and running locally, connect to the health service using a browser, curl or PostMan:
# Endpoints
open http://localhost:8080/v3/ops/health
open http://localhost:8080/v4/ops/health
If testing in local mode, set the USE_LOCAL_SERVICES=true
environment variable
and review the localhost values in the cla.service.ts
implementation for each
console. Update the ports as necessary.
For example, to test the corporate console:
cd cla-frontend-corporate-console
# use node 8
yarn install-frontend
# move to the source folder
cd src
# Ensure the AWS environment is setup
export AWS_PROFILE=cla-dev-profile
export AWS_REGION=us-east-1
# To use local services, set the following value.
# You will need to have the go and python backend up and running locally,
# otherwise it will use the DEV, STAGING, or PROD environment services based on
# the STAGE environment variable and the specific yarn target that you invoke.
export USE_LOCAL_SERVICES=true
# Run locally (code changes auto-deploy) - dev environment settings
# use node 8
yarn serve:dev