Guassp - GitLab User Access Synchronizer for SonarQube Projects

The project is intended to create a multi-user integration of SonarQube and GitLab. Project visibility and user permissions will be set in SonarQube in the same way as project permissions in GitLab.

This document is available in languages: eng 🇬🇧, ua 🇺🇦, rus 🇷🇺

• Implementation
• Requirements
• Components
• Container image
• Quick start
• Configuration
• API
• Pipeline
• Metrics
• Building and Debugging

Implementation

This utility consists of a front-end API that accepts requests to update permissions from the GitLab CI pipeline, where all trust is built around CI_JOB_TOKEN, trusted tasks are added to the RQ job queue (Redis Queue). Jobs are processed by the backend workers, and a separate prometheus metrics exporter is used to display statistics.

Permissions are divided by the access level of SonarQube roles according to the interests of group users in GitLab.
It takes into account how GitLab implements user access rights in the project, and users from invited groups, where access levels are limited in accordance with GitLab.

Requirements

Necessarily

• SonarQube must have configured [ALM][] integration with your GitLab, and you must specify the name of the integration key in SONARQUBE_ALM_KEY;
• Authentication in SonarQube must occur only through [ALM][] GitLab, otherwise the search for users will be violated, the rights are synchronized only for the explicitly corresponding account that came from GitLab;
• You must be able to provide an access token with administrative privileges to SonarQube in SONARQUBE_TOKEN for the worker to work correctly;
• You must create a service user in GitLab, which must have at least developer access in all projects that you want to analyze in SonarQube, with this user you must log in to SonarQube and give him global rights to perform analysis and create projects. Then you can use that user's token:
  • User token received in SonarQube as SONARQUBE_TOKEN used in CI pipeline to perform analysis and project management.
  • User token received in GitLab as GITLAB_TOKEN for worker, needed to get a list of all members of a group or project, including inherited and invited members
• The name of the project registered in SonarQube must be equal to $CI_PROJECT_NAMESPACE/$CI_PROJECT_NAME and the project key is recommended to be set to gitlab:$CI_PROJECT_ID. Only with these settings, synchronization will be successful, other states are possible but have not been tested.

Recommended

• The best experience will be gained by using the sonarqube-community-branch-plugin plugin which adds branch analysis and PR support to your projects. Also here you will need a previously created system user in GitLab with developer rights, which will allow you to decorate MR in your projects.
• For transparent integration, it is best to use the example script sq-integration-taks for the GitLab CI pipeline, otherwise other project registration options may not work correctly or even break the work. The example is also tailored for the presence of the sonarqube-community-branch-plugin plugin and the execution of OWASP DependencyCheck
• Use Prometheus and Grafana metrics to analyze synchronizer performance.

Components

The project is implemented on flask, for WSGI it is used bjoern, the work with the GitLab API is done through python-gitlab, and the SonarQube API through python-sonarqube-api. To process the message queue, RQ is used, whose metrics are returned by rq-exporter.

Container image

You can pull image from registries:

• ghcr.io/woozymasta/guassp:latest
• quay.io/woozymasta/guassp:latest
• docker.io/woozymasta/guassp:latest

Quick start

For a quick start, you can use the example from docker-compose:

1. docker-compose.env - change the settings in the environment file
2. docker-compose.yml - run with docker-compose up -d

Configuration

To launch the desired utility, transfer to the container or script guassp.sh argument:

• worker (default) - task handler from the queue
• api - API to control tasks
• exporter - Prometheus metric
• all-in-one - Launch worker,api and exporter immediately
• api-dev - Launching API via Flask DEV server

Application Options

• LISTEN_ADDRESS=0.0.0.0 - API listen address
• LISTEN_PORT=5000 - API listen port
• LOG_LEVEL=INFO - Logging level
• SECRET_KEY=secret - Secret key
• QUEUE_RESULT_TTL=7200 - Queue result storage time
• MORE_ACCURATE_SYNC=true - Search users in SonarQube by GitLab email else by username. It call one more API request per user is more accurate but slower x2

Options for working with GitLab

• GITLAB_URL=https://gitlab.com - GitLab server URL
• GITLAB_TOKEN - GitLab access token, must have permissions to view contributors and their permissions for maintained projects
• GITLAB_SKIP_USERS - comma-separated list of GitLab user IDs that will be skipped during synchronization

Options for working with SonarQube

• SONARQUBE_URL - SonarQube server URL
• SONARQUBE_TOKEN - access token to SonarQube with administrative privileges
• SONARQUBE_ALM_KEY - [ALM][] key GitLab integration name
• SONARQUBE_SKIP_GROUPS - comma-separated list of SonarQube groups that will be skipped during synchronization

Options for working with Redis

• REDIS_URL=redis://localhost:6379/0 - Redis server URL

Prometheus Metrics Exporter Options

• EXPORTER_LISTEN_ADDRESS=0.0.0.0 - Exporter listen address
• EXPORTER_LISTEN_PORT=9726 - Exporter listen port

Also you can pass RQ args and environment variables

API

Job Registration

POST /task

{"job_token": str}

Headers: JOB-TOKEN or Authorization: Bearer

You need to pass the task token in any of the options:

curl -sL http://127.0.0.1:5000/task \
  -H "Content-Type: application/json" \
  -d '{"job_token": "'$CI_JOB_TOKEN'"}' | jq

curl -sL http://127.0.0.1:5000/task -X POST \
  -H "JOB-TOKEN: $CI_JOB_TOKEN" | jq

curl -sL http://127.0.0.1:5000/task -X POST \
  -H "Authorization: Bearer $CI_JOB_TOKEN" | jq

The task token can be passed in the JOB-TOKEN header, or Authorization: Bearer or be the value of the job_token key in JSON

Tasks Queue

GET /tasks

curl -sL http://127.0.0.1:5000/tasks | jq
curl -sL http://127.0.0.1:5000/tasks | jq -er '.tasks | keys'

Task Status

GET /task/<uuid:job_uuid>

curl -sL http://127.0.0.1:5000/task/8b155172-cfcf-4777-b9f4-bfce53b6eb0e | jq

Removing a task from the queue

DELETE /task/<uuid:job_uuid>

curl -sL http://127.0.0.1:5000/task/8b155172-cfcf-4777-b9f4-bfce53b6eb0e \
  -X DELETE | jq

Manual task registration by GitLab project ID

POST /task_manual/<int:prj_id>

curl -sL http://127.0.0.1:5000/task_manual/111 \
  -X POST | jq

Health

GET /health

API health, application availability and job queues.

Pipeline

In the GitLab CI pipeline, first of all, you must make sure that the [ALM][] setup is done and refers to your project, after which you can submit a task for synchronization in guassp. Now you can start the analysis.

For publish Guassp API behind Nginx as part of SonarQube API check Nginx config example

: "${SONARQUBE_PROJECT_KEY:=gitlab:$CI_PROJECT_ID}"

curl --location --fail --user "$SONARQUBE_TOKEN:" \
  "$SONARQUBE_URL/api/alm_settings/set_gitlab_binding" \
  -d "almSetting=$SONARQUBE_ALM_NAME" \
  -d "project=$SONARQUBE_PROJECT_KEY" \
  -d "repository=$CI_PROJECT_ID"

curl --location --fail -X POST -H "JOB-TOKEN: $CI_JOB_TOKEN" \
  "$SONARQUBE_URL/api/guassp/task"

You can see a more voluminous example of a script for executing SonarQube in the pipeline in the sq-integration-taks.sh file

Metrics

Metrics implemented using project rq-exporter

Dashboard ID 12196 is suitable for visualization in Grafana or use it adaptation which will output only metrics from guassp.

Building and Debugging

A set of commands for fast local debugging in container

# Build
podman build -t guassp .

# Redis
podman run --rm -d -p 6379:6379 --name redis redis
# API
podman run --rm -d -p 5000:5000 --env-file .env --name guassp-api localhost/guassp:latest api
# Workers
podman run --rm -d --env-file .env --name guassp-worker-1 localhost/guassp:latest worker
podman run --rm -d --env-file .env --name guassp-worker-2 localhost/guassp:latest worker
# Exporter
podman run --rm -d -p 9726:9726 --env-file .env --name guassp-exporter localhost/guassp:latest exporter

# Check
curl 0.0.0.0:9726 -s | grep -v '^#'
curl 0.0.0.0:5000/tasks -s | jq
curl 0.0.0.0:5000/task -s -X POST -H "JOB-TOKEN: $CI_JOB_TOKEN" | jq

Or run locally, for this you need to install dependencies

apt-get install -y libev-dev libevdev2
python -m venv .venv
./.venv/bin/activate
pip install requirements.txt

And for simplicity, run through a script guassp

./guassp.sh api
./guassp.sh worker
./guassp.sh exporter