Using Rover in CI/CD

You can use Rover in any CI/CD environment that uses a Rover-supported operating system (Linux, MacOS, or Windows). Most commonly, this is to run schema checks with rover graph check or rover subgraph check.

Rover's installation is similar to many other CLI tools, but the recommended method varies depending on which provider you're using. We've included instructions for some of the most common CI/CD providers:

• CircleCI

• GitHub Actions

• Bitbucket Pipelines

• Jenkins

• Gitlab CI/CD

If you're using Rover with a CI/CD provider not listed here, we'd love for you to share the steps by opening an issue or pull request.

CircleCI

Linux jobs using the curl installer

Normally when installing, Rover adds the path of its executable to your $PATH. However, CircleCI doesn't use the $PATH variable between run steps. This means that if you install Rover and try to run it in the next step, you get a command not found: rover error.

To fix this, you can modify the $PATH and append it to $BASH_ENV. $BASH_ENV is executed at the beginning of each step, enabling any changes to be maintained across steps. You can add Rover to your $PATH using $BASH_ENV like this:

1echo 'export PATH=$HOME/.rover/bin:$PATH' >> $BASH_ENV

After you install Rover and modify the $BASH_ENV as shown, Rover should work like normal.

Full example

1# Use the latest 2.1 version of CircleCI pipeline process engine. See: https://circleci.com/docs/2.0/configuration-reference
2version: 2.1
3
4jobs:
5  build:
6    docker:
7      - image: cimg/node:15.11.0
8    steps:
9      - run:
10          name: Install
11          command: |
12            # download and install Rover
13            curl -sSL https://rover.apollo.dev/nix/v0.26.2 | sh
14
15            # This allows the PATH changes to persist to the next `run` step
16            echo 'export PATH=$HOME/.rover/bin:$PATH' >> $BASH_ENV
17      - checkout
18      # after rover is installed, you can run it just like you would locally!
19      # only run this command with the `--background` flag if you have the Apollo Studio GitHub integration enabled on your repository
20      - run: rover graph check my-graph@prod --schema ./schema.graphql --background

GitHub Actions

Displaying schema check results on GitHub pull requests

If you use GitHub Actions to automatically run schema checks on every pull request (as shown below), you can install the Apollo Studio GitHub app to provide links to the results of those checks alongside your other pull request checks:

To display schema check results on pull requests correctly, you need to make sure Rover associates the schema check execution with the pull request's HEAD commit, as opposed to the merge commit that GitHub adds. To guarantee this, set the APOLLO_VCS_COMMIT environment variable in your action's configuration, like so:

1env:
2  APOLLO_VCS_COMMIT: ${{ github.event.pull_request.head.sha }}

Linux/MacOS jobs using the curl installer

Normally when installing, Rover adds the path of its executable to your $PATH. However, GitHub Actions doesn't use the $PATH variable between run steps. This means that if you install Rover and try to run it in the next step, you get a command not found: rover error.

To fix this, you can append Rover's location to the $GITHUB_PATH variable. $GITHUB_PATH is similar to your system's $PATH variable, and additions to $GITHUB_PATH can be used across multiple steps. You can modify it like this:

1echo "$HOME/.rover/bin" >> $GITHUB_PATH

The following is a full example script, showing how to choose an apollo environment and set an APOLLO_KEY variable:

Full example

1# .github/workflows/check.yml
2
3name: Check Schema
4
5# Controls when the action will run. Triggers the workflow on push or pull request events
6on: [push, pull_request]
7
8# A workflow run is made up of one or more jobs that can run sequentially or in parallel
9jobs:
10  # This workflow contains a single job called "build"
11  build:
12    # The type of runner that the job will run on
13    runs-on: ubuntu-latest
14
15    # https://docs.github.com/en/actions/reference/environments
16    environment: apollo
17
18    # https://docs.github.com/en/actions/reference/encrypted-secrets
19    # https://docs.github.com/en/actions/reference/workflow-syntax-for-github-actions#jobsjob_idstepsenv
20    env:
21      APOLLO_KEY: ${{ secrets.APOLLO_KEY }}
22      APOLLO_VCS_COMMIT: ${{ github.event.pull_request.head.sha }}
23
24    # Steps represent a sequence of tasks that will be executed as part of the job
25    steps:
26      # Checks-out your repository under $GITHUB_WORKSPACE, so your job can access it
27      - uses: actions/checkout@v2
28
29      - name: Install Rover
30        run: |
31          curl -sSL https://rover.apollo.dev/nix/v0.26.2 | sh
32
33          # Add Rover to the $GITHUB_PATH so it can be used in another step
34          # https://docs.github.com/en/actions/reference/workflow-commands-for-github-actions#adding-a-system-path
35          echo "$HOME/.rover/bin" >> $GITHUB_PATH
36      # only run this command with the `--background` flag if you have the Apollo Studio GitHub integration enabled on your repository
37      - name: Run check against prod
38        run: |
39          rover graph check my-graph@prod --schema ./test.graphql --background
40

You can also use this Apollo Solutions repository install Rover on GitHub action runners. Once installed, rover is added to PATH, so it can be used in subsequent steps.

Bitbucket Pipelines

The following is a full example configuration for Bitbucket Pipelines. It shows how to:

• Run rover subgraph check for each commit on all branches

• Run rover subgraph publish to keep the schema definition of your main branch in-sync with a base variant (@local in this case)

The example uses the following Pipeline Repository Variables to make the pipeline configuration portable across different repositories:

• APOLLO_KEY

• APOLLO_SUBGRAPH_NAME, which represents the name of the subgraph you're running schema checks for

• APOLLO_LOCAL_PORT, which represents the port number of the base variant

Full example

1# ./bitbucket-pipelines.yml
2
3image: node
4
5definitions:
6  steps:
7    - step: &rover-subgraph-check
8        name: "[Rover] Subgraph Check"
9        caches:
10          - node
11        script:
12          - 'echo "Subgraph name: $APOLLO_SUBGRAPH_NAME"'
13          - npx -p @apollo/rover@latest
14            rover subgraph check my-graph@prod
15            --name $APOLLO_SUBGRAPH_NAME
16            --schema ./schema.graphql
17
18    - step: &local-publish
19        name: "[Rover] @local publish (sync with main/master)"
20        caches:
21          - node
22        script:
23          - 'echo "Subgraph name: $APOLLO_SUBGRAPH_NAME"'
24          - 'echo "Local variant port: $APOLLO_LOCAL_PORT"'
25
26          - npx -p @apollo/rover@latest
27            rover subgraph publish my-graph@local
28            --name $APOLLO_SUBGRAPH_NAME
29            --schema ./schema.graphql
30            --routing-url http://localhost:$APOLLO_LOCAL_PORT/graphql
31
32pipelines:
33  default:
34    - step: *rover-subgraph-check
35
36  branches:
37    '{main,master}':
38      - step: *rover-subgraph-check
39      - step: *local-publish

Jenkins

To set up Rover for use with Jenkins, first consider which type of Jenkins agent you'll use in your pipelines. The samples below demonstrate a golang pipeline that uses Docker, but you can modify them to meet your specific needs.

Distributed builds via the node agent

If you're running a distributed build system using the node agent type, make sure that Rover is installed on all machines either as part of a baseline image or via a setup script. Also make sure it's available globally via the PATH environment variable.

Pipelines using Docker

If you're using Rover with a Docker-enabled pipeline, note the following additional considerations:

$PATH issues

Normally when installing, Rover adds the path of its executable to your $PATH. However, Jenkins doesn't persist the $PATH variable between runs of sh steps, because each sh block runs as its own process. This means that if you install Rover and try to run it in the next step, you get a command not found: rover error. This is functionally similar to the CircleCI note, but the resolution is different.

To avoid this issue, do one of the following:

• Use the script, but reference rover by its full path ($HOME/.rover/bin/rover)

• Download the latest release via cURL and extract the binary like so (this downloads Rover 0.23.0 for Linux x86 architectures):

  Text

  copy

  1curl -L https://github.com/apollographql/rover/releases/download/v0.26.2/rover-v0.26.2-x86_64-unknown-linux-gnu.tar.gz | tar --strip-components=1 -zxv

Permission issues

If you run into permissions issues within Docker, you can resolve many of them by creating a user to run the install and build processes. The example Dockerfile below shows how to accomplish this with a specific Docker image for your Jenkins build pipeline:

1FROM golang:1.18
2RUN useradd -m rover && echo "rover:rover" | chpasswd
3USER rover
4RUN curl -sSL https://rover.apollo.dev/nix/latest | sh

Jenkinsfile Configuration

After you've installed Rover appropriately, you can execute the rover command within a sh step, as shown in the example configuration below. Because rover outputs logs via stderr and emits proper status codes, it generates build errors if the rover subgraph check command fails.

We recommend passing arguments to rover commands via environment variables. This enables you to reuse large portions of your pipeline, making it faster to onboard new subgraphs without rewriting code.

Additionally, we strongly recommend passing in the APOLLO_KEY by using a Jenkins credential and referencing it using credentials(key_name) within your jenkinsfile. An example of this is below.

1pipeline {
2  agent {
3    dockerfile {
4      filename './build_artifacts/Dockerfile'
5    }
6
7  }
8  stages {
9    stage('Rover Check') {
10      steps {
11        sh '''echo "Subgraph: $APOLLO_SUBGRAPH_NAME
12        $HOME/.rover/bin/rover subgraph check $APOLLO_GRAPH_REF --name $APOLLO_SUBGRAPH_NAME --schema $SCHEMA_PATH'''
13      }
14    }
15
16    stage('Build') {
17      steps {
18        sh 'go build .'
19      }
20    }
21
22    stage('Go Test') {
23      steps {
24        sh 'go test ./... -v'
25      }
26    }
27
28    stage('Schema Publish to Dev') {
29      when {
30        expression { env.BRANCH_NAME == 'main' }
31      }
32      steps {
33        sh '$HOME/.rover/bin/rover subgraph publish $APOLLO_GRAPH_REF --name $APOLLO_SUBGRAPH_NAME --schema $SCHEMA_PATH'
34      }
35    }
36
37  }
38  environment {
39    APOLLO_KEY = credentials('apollo_key')
40    APOLLO_SUBGRAPH_NAME = 'products'
41    APOLLO_CONFIG_HOME = '~/.config/rover'
42    SCHEMA_PATH = './graph/schema.graphqls'
43    APOLLO_GRAPH_REF = 'ApolloJenkins@dev'
44  }
45}

Gitlab

Since there isn't any official Docker image for Rover, we can use the debian:stable-slim as a base image. All you need to do is fetch the source via cURL, add the executable to the PATH variable, then publish your subgraphs.

1push_subgraphs:
2stages:
3  - publish_subgraphs
4  
5publish_subgraphs:
6  stage: publish_subgraphs
7  image: debian:stable-slim
8  retry: 1 # to retry if any connection issue or such happens
9  before_script:
10    - apt-get update && apt-get install curl -y
11  script:
12    - curl -sSL https://rover.apollo.dev/nix/latest | sh # Install the latest version of Rover
13    - export PATH="$HOME/.rover/bin:$PATH" # Manually add it to the ruuner PATH
14    - export APOLLO_KEY=$APOLLO_FEDERATION_KEY
15    - rover subgraph publish $APOLLO_GRAPH_REF --name $APOLLO_SUBGRAPH_NAME --schema $SCHEMA_PATH

Using With npm/npx

If you're running in a Node.js workflow, it might be easier to use the NPM distribution of Rover. This way, you don't need to adjust the PATH at all to run Rover, and it might fit better into your existing workflow.

You can use Rover by adding it to your package.json dependencies using these instructions and then execute it using npm scripts, similar to other workflows you might already have. If you don't want to install Rover as a dependency, you can run it with npx by using the -p flag:

1npx -p @apollo/rover rover graph check my-graph@prod --schema=./schema.graphql --background

Since most commands require you be authenticated, see the above sections for instructions on how to add environment variables for your CI/CD provider.