Skip to content
/ neosync Public
forked from nucleuscloud/neosync

A developer-first way to create safe, anonymized test data and sync it across all environments for high-quality local, stage and CI testing

License

Notifications You must be signed in to change notification settings

sibis/neosync

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

neosyncbanner

Open source Test Data Management

Neosync is a developer-first way to create anonymized, secure test data and sync it across all environments for high-quality local, stage and CI testing

PRs Welcome main tests Follow X Roadmap

Introduction

neosync-data-flow

Neosync is an open source platform that connects to a snapshot of your production database and allows teams to either generate synthetic data from their production schema or anonymize production-data and sync it across all of their environments for high-quality local, stage and CI testing.

Our mission is to help developers build better, more resilient applications while protecting sensitive data. To do that, we built Neosync to give teams three things:

  1. A world-class developer experience that fits into any workflow and follows modern developer best practices such as GitOps
  2. A platform that can anonymize sensitive data or automatically generate synthetic data from a schema and sync that across all environments
  3. An open source approach that allows you to keep your most sensitive data in your infrastructure

Features

  • Automatically generate synthetic data based on your schema
  • Anonymize existing production-data to protect data
  • Create subsets of your production database for local and CI testing by filtering on an object, id or custom query
  • Complete async pipeline that automatically handles job retries, failures and playback using an event-sourcing model
  • Referential integrity for your data automatically - never worry about broken foreign keys again
  • Use our declarative, GitOps based configs as a step in your CI pipeline to hydrate your CI DB
  • Pre-built transformers for all major data types
  • Define custom transformers
  • Pre-built integrations with Postgres, Mysql, S3

Getting started

You can also check out our Docs for more guides including a production-ready guide. Note: these are still a work in progress.

Run Neosync locally

To set up and run Neosync locally, make sure you have Git and Docker installed on your system.

The sections below detail the tools required to build and run the Neosync development environment. This can be circumvented by using our official devcontainer which comes pre-installed with all of the tools necessary.

There are then two ways to start Neosync:

  • Tilt
  • Docker Compose

Tilt is the method we currently use to development Neosync. This lets us develop as if we are running inside of a Kubernetes cluster. This isn't for everyone, which is we also offer a compose method for a simpler, kubernetesless approach.

Check out the sections below for which method applies to you.

Tools

Currently, the primary development environment is done by deploying the app and its dependent resources into a kind cluster using tilt. We utilize helm charts to wrap up deployable artifacts and have tilt install these to closely mimic a production environment. Due to this, there are a number of dependencies that must be installed on a host system prior to being able to run neosync locally.

Detailed below are the main dependencies are descriptions of how they are utilized:

Kubernetes

Kubernetes is used today as our primary development environment. Tilt is a great tool that lets you define your environment in code. This lets us develop quickly, locally, while closely mimicking a real production environment.

  • kind
    • Kubernetes in Docker. We use this to spin up a barebones kubernetes cluster that deploys all of the neosync resources.
  • tilt
    • Allows us to define our development environment as code.
  • ctlptl
    • CLI provided by the Tilt-team to make it easy to declaratively define the kind cluster that is used for development
  • kubectl
    • Allows for observability and management into the spun-up kind cluster.
  • kustomize
    • yaml templating tool for ad-hoc patches to kubernetes configurations
  • helm
    • Kubernetes package manager. All of our app deployables come with a helm-chart for easy installation into kubernetes
  • helmfile
    • Declaratively define a helmfile in code! We have all of our dev charts defined as a helmfile, of which Tilt points directly to.

Golang + Protobuf

  • Golang
    • The language of choice for our backend and worker packages
  • sqlc
    • Our tool of choice for the data-layer. This lets us write pure SQL and let sqlc generate the rest.
  • buf
    • Our tool of choice for interfacing with protobuf
  • golangci-ci
    • The golang linter of choice

Npm/Nodejs

  • Node/Npm

All of these tools can be easily installed with brew if on a Mac. Today, sqlc and buf don't need to be installed locally as we exec docker images for running them. This lets us declare the versions in code and docker takes care of the rest.

It's of course possible run everything on bare metal without Kuberentes or Tilt, but there will be more work getting everything up and running (at least today).

Brew Install

Each tool above can be straightforwardly installed with brew if on Linux/MacOS

brew install kind tilt-dev/tap/tilt tilt-dev/tap/ctlptl kubernetes-cli kustomize helm helmfile go sqlc buf golangci-lint node

Devcontainer

Host machine setup can be skipped by developing inside of a vscode devcontainer. This container comes pre-baked with all of the tools we use to develop and work on neosync. This container also supports running neosync with compose or tilt.

Running Docker with Docker Desktop

When running with either Tilt or docker compose, we map volumes from these filesystems to the host machine for both neosync and Temporal's databases. We mount a container path locally in a .data folder. If on a Mac, ensure that you've allowed wherever this repository has been cloned into to the allow-list in Docker Desktop.

The allow list can be found by first opening Docker Desktop. Settings -> Resources -> File Sharing and add the path to the Neosync repository.

If you don't want to do this, you can remove the volume mappings in the compose file or remove the pvc for Tilt. This comes at a negative of the local database not surviving restarts, however.

Setup with Tilt

Step 1 is to ensure that the kind cluster is up and running along with its registry. This can be manually created, or done simply with ctlptl. The cluster is declaratively defined here

The below command invokes the cluster-create script that can be found here

make cluster-create

After the cluster has been successfully created, tilt up can be run to start up neosync. Refer to the top-level Tiltfile for a clear picture of everything that runs. Each dependency in the neosync repo is split into sub-Tiltfiles so that they can be run in isolation, or in combination with other sub-resources more easily.

Once everything is up and running, the app can be accessed at locally at https://localhost:3000.

Setup with Docker Compose

Neosync can be run with compose. This works pretty well, but is a bit more manual today than with Tilt. Not everything is hot-reload, but you can successfully run everything using just compose instead of having to manage a kubernetes cluster and running Tilt. To enable hot reloading, must run docker compose watch instead of up. Currently there is a limitation with devcontainers where this command must be run via sudo.

There are two compose files that need to be run today. The first is the Temporal compose, the second is the Neosync compose. It's suggested you run these separate (as of today) for a clean separation of concerns.

Building the backend and worker when using Docker Compose.

Prior to running docker compose up -d, the worker and api will need to be built.

When building the Go processes with the intention to run with docker compose, it's important to run make dbuild instead of the typical make build so that the correct GOOS is specified. This is only needed if your native OS is not Linux (or aren't running in a devcontainer). The make dbuild command ensures that the Go binary is compiled for Linux instead of the host os.

This will need to be done for both the worker and api processes prior to running compose up.

Running Compose

$ docker compose -f temporal/compose.yml up -d
$ docker compose -f compose.yml up -d

Once everything is up and running, the app can be accessed locally at https://localhost:3000.

Work to be done:

  • inherit the temporal compose inside of the neosync compose, separate with compose profiles.

Running Compose with Authentication

Note, a compose file with authentication pre-configured can be found here. This will stand up Keycloak with a pre-configured realm that will allow logging in to Neosync with a standard username and password, competely offline!

$ docker compose -f temporal/compose.yml up -d
$ docker compose -f compose/compose-auth.yml up -d

Resources

Some resources to help you along the way:

  • Docs for comprehensive documentation and guides: Note these are still a work in progress.
  • Discord for discussion with the community and Neosync team
  • X for the latest updates

Contributing

We love contributions big and small. Here are just a few ways that you can contribute to Neosync.

Licensing

We strongly believe in free and open source software and make this repo is available under the MIT expat license.

Triggering a Release

Triggering a release is done by cutting a git tag. This causes all artifacts for the various components in the system to build and publish.

Tag Format:

The tag format is a semver compliant tag that starts with v.

Examples:

  • v0.0.1
  • v0.0.1-nick.1

This is done by running the hack/tag.sh script like so:

$ ./hack/tag.sh <tag>

Example:

$ ./hack/tag.sh v0.0.1

About

A developer-first way to create safe, anonymized test data and sync it across all environments for high-quality local, stage and CI testing

Resources

License

Code of conduct

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • TypeScript 62.8%
  • Go 34.7%
  • MDX 1.4%
  • Shell 0.3%
  • CSS 0.3%
  • Starlark 0.2%
  • Other 0.3%