Upgrades

How to upgrade NodeFoundry and Ceph components.

NodeFoundry Updates

The Sentinel and worker node agents can be updated independently. The master node update process:

1. Download the latest release
2. Stop the Sentinel service
3. Replace the binary
4. Start the service

Worker nodes pull updates automatically from the master on their next check-in.

Ceph Version Upgrades

Ceph upgrades should be performed as rolling upgrades — one daemon at a time to maintain cluster availability.

Upgrade Order

1. Monitors — upgrade one at a time, wait for quorum
2. Managers — upgrade standby first, then active
3. OSDs — upgrade one at a time, wait for PG recovery
4. MDS — upgrade standby first
5. RGW — upgrade one instance at a time

Best Practices

• Always read the Ceph release notes before upgrading
• Ensure the cluster is HEALTH_OK before starting
• Upgrade one daemon at a time and verify health between each
• Keep at least one version of rollback capability
• Schedule upgrades during low-traffic periods