Deployment

The shipping image is ghcr.io/cyborgtests/playwright-reports-server:latest. Node 24 Alpine, runs as appuser, exposes port 3001, healthchecks GET /api/ping, data volume at /app/data/. Nothing fancy.

If you don't already have an opinion about how to host this, there are couple of requirements:

• if you want to merge reports in app - consider investing into more storage space, as it basically would need to maintain all blobs + produce a report at a time. Failed reports usually require a lot more than 1kb html for success run and could be up to 300-500Mb for decent set of tests (up to 100), or even more if something went really wrong.
• for the storage there are multiple options (local fs/s3/azure/other s3-compatible object storages)
• state lives in local SQLite. There's a replication to storage, but for fs you would need to share it between instances via volume/attached storage.
• it could work on a CPU-limited small 0.5GB RAM instance and that could be used as a starting point. Small "nuance" here is that node itself + dependencies eat up to ~200Mb.

Quick starts

Local filesystem, with a mounted volume:

docker run -d --name pw-reports \
  -p 3001:3001 \
  -v /path/on/host:/app/data \
  ghcr.io/cyborgtests/playwright-reports-server:latest

S3 storage:

docker run -d --name pw-reports \
  -p 3001:3001 \
  -e API_TOKEN=<long-random-string> \
  -e AUTH_SECRET=<another-long-random-string> \
  -e DATA_STORAGE=s3 \
  -e S3_ENDPOINT=s3.amazonaws.com \
  -e S3_REGION=us-east-1 \
  -e S3_ACCESS_KEY=<key> \
  -e S3_SECRET_KEY=<secret> \
  -e S3_BUCKET=my-pw-reports \
  ghcr.io/cyborgtests/playwright-reports-server:latest

Litestream replicates the metadata DB to the same bucket automatically.

Azure Blob storage. Litestream replicates the metadata DB to your Azure container, same shape as the S3 path:

docker run -d --name pw-reports \
  -p 3001:3001 \
  -e API_TOKEN=... \
  -e AUTH_SECRET=... \
  -e DATA_STORAGE=azure \
  -e AZURE_ACCOUNT_NAME=... \
  -e AZURE_ACCOUNT_KEY=... \
  -e AZURE_CONTAINER=playwright-reports-server \
  ghcr.io/cyborgtests/playwright-reports-server:latest

Kubernetes (Helm)

A chart lives in charts/playwright-reports-server. It runs the server as a single-replica StatefulSet (SQLite has one writer), as the image's non-root user with a read-only root filesystem, and takes secrets from Kubernetes Secrets only.

kubectl create secret generic pw-reports-auth \
  --from-literal=API_TOKEN="$(openssl rand -hex 32)" \
  --from-literal=AUTH_SECRET="$(openssl rand -base64 32)"

helm install pw-reports oci://ghcr.io/cyborgtests/charts/playwright-reports-server \
  --set auth.existingSecret=pw-reports-auth

• storage.type: fs (default): /app/data on a PersistentVolumeClaim from the StatefulSet. Prefer a block storage class: SQLite does not like network file systems.
• storage.type: s3: reports in the bucket, Litestream keeps the metadata DB there, the pod only needs a scratch volume. Keys come from an existing Secret (storage.s3.existingSecret).

All values are listed in the chart README.

Health and logs

docker logs -f pw-reports
curl http://localhost:3001/api/ping   # pong

The healthcheck is built into the image.

Updating

docker pull ghcr.io/cyborgtests/playwright-reports-server:latest
docker rm -f pw-reports
# re-run with the same flags

Backups

Backend	What to back up
fs	The whole data/ directory (the mounted host path)
s3	The bucket. Litestream keeps the metadata DB in there continuously.
azure	The container. Litestream keeps the metadata DB in it continuously.

The first backup test is the one before you need it. The second backup test is done after you need it, and usually almost always fails, especially when the first test wasn't done. Do the first test, ensure your data is shared across instances.

See also

• Storage
• Configuration