GitLab CI/CD
GitLab runs service containers next to each job's container. The tests reach each service by its alias.
How services work in GitLab
- Hostname: the service's
alias(postgres), or a name derived from the image if you leave the alias out. Always set one. - Readiness: the runner waits up to 30 seconds for the service's port to open, then starts the job anyway. Slow services (MSSQL, Neo4j, Kafka) need a wait loop in the job. The example below includes one.
- Variables: job-level
variablesare passed to the job and to its services. Put service-only settings under the service's ownvariables:. CI=true: GitLab sets it for every job, so the example specs run without extra setup.
Complete .gitlab-ci.yml
All jobs are in one stage, so they run in parallel. The --project names are the ones in Propmaster's playwright.config.ts; use yours.
stages: [test]
workflow:
auto_cancel:
on_new_commit: interruptible # a new push cancels the older pipeline
default:
image: node:24
interruptible: true
variables:
npm_config_cache: "$CI_PROJECT_DIR/.npm"
PLAYWRIGHT_JUNIT_OUTPUT_FILE: results/junit.xml
# Shared setup: cached npm install, a wait helper, JUnit results in the MR widget.
.playwright:
cache:
key:
files: [package-lock.json]
paths: [.npm/]
before_script:
- |
wait_for() {
for i in $(seq 1 90); do
node -e "require('net').connect(+process.argv[2], process.argv[1]).on('connect', () => process.exit(0)).on('error', () => process.exit(1))" "$1" "$2" && return 0
sleep 2
done
echo "$1:$2 is not reachable"; return 1
}
- npm ci --prefer-offline --no-audit --no-fund
artifacts:
when: always
expire_in: 7 days
reports:
junit: results/junit.xml
paths:
- playwright-report/
- test-results/
checks:
extends: .playwright
artifacts: { when: on_failure, paths: [playwright-report/] }
script:
- npm run typecheck # tsc --noEmit
- npm run lint:tests # propmaster check: test IDs, tags, steps, imports
postgres:
extends: .playwright
services:
- name: postgres:15
alias: postgres
variables:
POSTGRES_DB: testdb
POSTGRES_USER: testuser
POSTGRES_PASSWORD: testpass
variables:
PW_DB_TYPE: postgresql
PW_DB_HOST: postgres
PW_DB_PORT: "5432"
PW_DB_NAME: testdb
PW_DB_USERNAME: testuser
PW_DB_PASSWORD: testpass
script:
- wait_for postgres 5432
- node scripts/create-test-schema.js # your schema setup
- npx playwright test --project=integration-database --reporter=line,junit
redis:
extends: .playwright
services:
- name: redis:7
alias: redis
variables:
PW_REDIS_HOST: redis
PW_REDIS_PORT: "6379"
PW_REDIS_KEY_PREFIX: "test:ci-$CI_PIPELINE_ID:"
script:
- wait_for redis 6379
- npx playwright test --project=integration-redis --reporter=line,junit
mongodb:
extends: .playwright
services:
- name: mongo:7
alias: mongodb
variables:
PW_MONGODB_HOST: mongodb
PW_MONGODB_DATABASE: testdb
script:
- wait_for mongodb 27017
- npx playwright test --project=integration-mongodb --reporter=line,junit
kafka:
extends: .playwright
services:
- name: apache/kafka:latest
alias: kafka
variables:
KAFKA_NODE_ID: "1"
KAFKA_PROCESS_ROLES: broker,controller
KAFKA_LISTENERS: PLAINTEXT://0.0.0.0:9092,CONTROLLER://0.0.0.0:9093
KAFKA_ADVERTISED_LISTENERS: PLAINTEXT://kafka:9092
KAFKA_CONTROLLER_LISTENER_NAMES: CONTROLLER
KAFKA_LISTENER_SECURITY_PROTOCOL_MAP: PLAINTEXT:PLAINTEXT,CONTROLLER:PLAINTEXT
KAFKA_CONTROLLER_QUORUM_VOTERS: 1@localhost:9093
KAFKA_AUTO_CREATE_TOPICS_ENABLE: "true"
KAFKA_NUM_PARTITIONS: "1"
KAFKA_OFFSETS_TOPIC_REPLICATION_FACTOR: "1"
CLUSTER_ID: MkU3OEVBNTcwNTJENDM2Qk
variables:
PW_KAFKA_BROKERS: kafka:9092
script:
- wait_for kafka 9092
- npx playwright test --project=integration-kafka --reporter=line,junit
browser:
extends: .playwright
image: mcr.microsoft.com/playwright:v1.65.1-noble # must match @playwright/test
before_script:
# Skips native builds: only needed if the project installs better-sqlite3 (SQLite), which this image can't compile.
- npm ci --prefer-offline --no-audit --no-fund --ignore-scripts
script:
- npx playwright test --project=browser-chromium --project=browser-firefox --project=browser-webkit --reporter=line,junitThe browser job replaces before_script, so it has no wait_for. It doesn't need one.
Other backends
Add a job like redis above with these services and variables:
services:
- name: mysql:8.0
alias: mysql
variables:
MYSQL_DATABASE: testdb
MYSQL_USER: testuser
MYSQL_PASSWORD: testpass
MYSQL_ROOT_PASSWORD: rootpass
variables:
PW_DB_TYPE: mysql
PW_DB_HOST: mysql
PW_DB_PORT: "3306"
PW_DB_NAME: testdb
PW_DB_USERNAME: testuser
PW_DB_PASSWORD: testpass
script:
- wait_for mysql 3306
- npx playwright test --project=integration-database --reporter=line,junitservices:
- name: mcr.microsoft.com/mssql/server:2022-latest
alias: mssql
variables:
ACCEPT_EULA: "Y"
MSSQL_SA_PASSWORD: "TestPass123!"
variables:
PW_DB_TYPE: mssql
PW_DB_HOST: mssql
PW_DB_PORT: "1433"
PW_DB_NAME: master
PW_DB_USERNAME: sa
PW_DB_PASSWORD: "TestPass123!"
PW_DB_ENCRYPT: "false"
PW_DB_TRUST_SERVER_CERTIFICATE: "true"
script:
- wait_for mssql 1433
- sleep 10 # the port opens a little before logins are accepted
- npx playwright test --project=integration-database --reporter=line,junitservices:
- name: neo4j:5
alias: neo4j
variables:
NEO4J_AUTH: neo4j/testpassword
variables:
PW_NEO4J_URI: bolt://neo4j:7687
PW_NEO4J_USERNAME: neo4j
PW_NEO4J_PASSWORD: testpassword
script:
- wait_for neo4j 7687
- npx playwright test --project=integration-neo4j --reporter=line,junitSecrets and environments
Keep real credentials in Settings → CI/CD → Variables, marked Masked and Protected. They reach the job as environment variables with the same name, so a variable named PW_DB_PASSWORD needs no mapping. To target a shared environment instead of service containers:
staging-api:
extends: .playwright
variables:
PW_ENVIRONMENT: stg # reads the "stg" block of environments.json
script:
- npx playwright test --project=openapi --reporter=line,junit
rules:
- if: $CI_COMMIT_BRANCH == $CI_DEFAULT_BRANCHFor Vault or OpenBao, Propmaster's secrets providers read VAULT_ADDR / BAO_ADDR and a token, or log in with a GitLab CI JWT (id_tokens: plus the jwt auth method).
Tips
- Merge request widget:
artifacts:reports:junitshows new failures in the MR.when: alwaysuploads results from failed runs too. - Docs-only changes: to skip integration jobs, add
rules: [{ changes: { paths: ["src/**/*", "tests/**/*", "package-lock.json"] } }]to.playwright. - Runner memory: MSSQL needs about 2 GB. On small shared runners, give MSSQL its own job, as above.
FF_NETWORK_PER_BUILD: only needed when services must reach each other, for example an API service that talks to the database service.