Refactor CI/CD workflow to improve database migration process and key generation
CI/CD / BackEnd (push) Successful in 1m18s
CI/CD / WebPlayer (push) Successful in 40s
CI/CD / Deploy (push) Skipped
CI/CD / BackEnd (pull_request) Successful in 1m27s
CI/CD / WebPlayer (pull_request) Successful in 39s
CI/CD / Deploy (pull_request) Skipped

This commit is contained in:
2026-09-23 23:59:45 +02:00
parent cfdc61a4b8
commit 430c711ff2
+14 -2
View File
@@ -86,11 +86,23 @@ jobs:
git fetch origin master
git reset --hard origin/master
- name: Migrate database, generate keys
- name: Generate stream-ticket keys
run: |
set -euo pipefail
cd "$DEPLOY_PATH"
$COMPOSE up -d wait-for-postgres keys-init migrate
$COMPOSE run --rm --no-deps keys-init
# The migrations are baked into the migrator image, so it is rebuilt on
# every deploy (plain `up` would reuse the stale image and re-apply the
# old migrations). `run --rm` starts wait-for-postgres, blocks until the
# migration finishes and propagates its exit code, so a failed migration
# aborts the deploy before the new API is started against an old schema.
- name: Migrate database
run: |
set -euo pipefail
cd "$DEPLOY_PATH"
$COMPOSE build migrate
$COMPOSE run --rm migrate
# Both jobs verify themselves through the public auth./s3. URLs, served
# by the shared edge nginx (Infrastructure's edge/), and are idempotent.