Skip to main content

Migration to Yarn 3

NOTE: We do not yet recommend all projects to migrate to Yarn 3. Only do so if you have specific reasons for it.

While Backstage projects created with @backstage/create-app use Yarn 1 by default, it is possible to switch them to instead use Yarn 3. Tools like yarn backstage-cli versions:bump will still work, as they recognize both lockfile formats.


In addition to this guide, also be sure to check out the Yarn migration guide as well.

Ignore File Updates

First off, be sure to have the updated ignore entries in your app. These are included in all newly created projects, but might be missing in yours:

Add the following to .gitignore:

# Yarn 3 files

And this to .dockerignore:



Let's move on to the actual installation. We'd recommend making separate Git commits of most of these steps, in case you need to go back and debug anything. To install Yarn 3, run the following command in the project root:

yarn set version stable

We'll need the Yarn workspace tools plugin later on, so let's install that too:

yarn plugin import @yarnpkg/plugin-workspace-tools

Now we're ready to re-install all dependencies. This will update your yarn.lock and switch the project to use node-modules as the Yarn node linker.

In case you had a .yarnrc you can delete it now, but be sure to migrate over any options to .yarnrc.yml. See the Yarn configuration docs for a full list of options. For example, registry is now npmRegistryServer, and network-timeout is httpTimeout.

Migrate Usage

At this point you'll be all set up with Yarn 3! What remains is to migrate any usage of Yarn according to their migration guide. For example, any yarn install --frozen-lockfile commands should be replaced with yarn install --immutable.

You'll also need to update any Dockerfiles to add instructions to copy in your Yarn 3 installation into the image:

COPY .yarn ./.yarn
COPY .yarnrc.yml ./

In a multi-stage Dockerfile, each stage that runs a yarn command will also need the Yarn 3 installation. For example, in the final stage you may need to add the following:

COPY --from=build --chown=node:node /app/.yarn ./.yarn
COPY --from=build --chown=node:node /app/.yarnrc.yml ./

The --production flag to yarn install has been removed in Yarn 3, instead you need to use yarn workspaces focus --all --production to avoid installing development dependencies in your production deployment. A tradeoff of this is that yarn workspaces focus does not support the --immutable flag.

RUN yarn workspaces focus --all --production && rm -rf "$(yarn cache clean)"

Additionally, yarn config has been reworked from being able to store any arbitrary key-value pairs to only supporting a handful of predefined pairs. Previously, we would set our preferred python3 interpreter to work around any issues related to node-gyp so we need to provide an appropriate substitute.

FROM node:16-bullseye-slim

# Set Python interpreter for `node-gyp` to use
ENV PYTHON /usr/bin/python3

# Install sqlite3 dependencies. You can skip this if you don't use sqlite3 in the image,
# in which case you should also move better-sqlite3 to "devDependencies" in package.json.
RUN apt-get update && \
apt-get install -y --no-install-recommends libsqlite3-dev python3 build-essential && \
rm -rf /var/lib/apt/lists/* && \
yarn config set python /usr/bin/python3
rm -rf /var/lib/apt/lists/*

You'll want to make sure that the PYTHON environment variable is declared relatively early, before any instances of Yarn are invoked as node-gyp is indirectly triggered by some modules during installation.

If you have any internal CLI tools in your project that are exposed through "bin" entries in package.json, then you'll need to add these packages as dependencies in your project root package.json. This is to make sure Yarn picks up the executables and makes them available through yarn <executable>.