Production

This guide covers running Flynn in production.

Cluster Requirements

Storage

Flynn uses ZFS for all persistent storage. Use dedicated SSDs for the ZFS pool:

sudo zpool create -f flynn-default mirror /dev/sdb /dev/sdc

Mirror or raidz configurations are recommended for production data integrity.

Blobstore Backend

By default, the blobstore uses local ZFS storage. For production, configure an external object store:

Amazon S3

flynn -a blobstore env set \
  BACKEND=s3 \
  S3_BUCKET=my-flynn-blobstore \
  S3_REGION=us-east-1 \
  AWS_ACCESS_KEY_ID=... \
  AWS_SECRET_ACCESS_KEY=...

Google Cloud Storage

flynn -a blobstore env set \
  BACKEND=gcs \
  GCS_BUCKET=my-flynn-blobstore \
  GCS_CREDENTIALS='...'

Azure Blob Storage

flynn -a blobstore env set \
  BACKEND=azure \
  AZURE_CONTAINER=my-flynn-blobstore \
  AZURE_STORAGE_ACCOUNT=... \
  AZURE_STORAGE_KEY=...

DNS and Load Balancing

DNS

Configure a wildcard DNS record pointing to all cluster nodes:

*.flynn.example.com  A  10.0.0.1
*.flynn.example.com  A  10.0.0.2
*.flynn.example.com  A  10.0.0.3

External Load Balancer

For production, place a load balancer (HAProxy, nginx, AWS ALB) in front of the Flynn router:

• Forward ports 80 and 443 to all cluster nodes
• Use TCP mode (the Flynn router handles TLS termination)
• Enable health checks against the router’s health endpoint

Firewall

External Access

Inter-Node (Allow Between All Cluster Nodes)

Backups

Cluster Backup

flynn cluster backup --file cluster-backup.tar

This exports all apps, releases, formations, routes, and database contents.

Cluster Restore

flynn cluster restore --file cluster-backup.tar

Database Backups

Individual database backups:

flynn -a my-app pg dump > pg-backup.sql
flynn -a my-app pg restore < pg-backup.sql

Automated Backups

Schedule regular backups with cron:

0 */6 * * * flynn cluster backup --file /backups/flynn-$(date +\%Y\%m\%d-\%H\%M).tar

Monitoring

Health Checks

• Controller API: GET /status on the controller
• Router health: check port 80/443 response
• discoverd: GET /services on port 1111
• flynn-host: GET /host/status on port 1113

Logs

Flynn aggregates logs from all containers via the log aggregator service:

flynn -a my-app log -f

For external log shipping, configure a syslog drain.

Updating

Rolling Update

1. Back up the cluster
2. Update flynn-host binary on each node
3. Run flynn-host download to fetch new component images
4. Restart flynn-host on each node (one at a time for zero-downtime)
5. The scheduler will redeploy system apps with new images

Version Pinning

Flynn uses TUF for version management. Pin to a specific version:

flynn-host download --version v20260416.0

Adding/Removing Hosts

Adding a Host

1. Install Flynn on the new node
2. Start flynn-host daemon with --peer-ips pointing to existing nodes
3. The new host joins the cluster automatically

Removing a Host

# Drain jobs from the host
flynn-host drain <HOST_ID>

# Remove from cluster
flynn-host remove <HOST_ID>

Security Considerations

• Rotate the controller key periodically
• Use TLS for all external communication
• Keep ZFS pools encrypted at rest where possible
• Restrict SSH access to admin IPs only
• Use firewall rules to limit inter-node ports to cluster members only