Skip to main content
The Upgrade page allows you to upgrade your Ceph cluster to newer versions. Upgrades are performed in a controlled manner, updating one daemon at a time to minimize disruption. You can monitor progress, pause, resume, or stop upgrades as needed.

​
Key Concepts

Rolling Upgrade

Ceph upgrades one daemon at a time, maintaining cluster availability throughout the process.

Target Version

The Ceph version or container image you are upgrading to.

Daemon

A Ceph service process (mon, mgr, osd, mds, rgw) that gets upgraded individually.

Upgrade Blockers

Conditions that prevent an upgrade from starting, such as cluster health errors.

​
Required Permissions

ActionPermission
View Upgrade Statusiam:project:infrastructure:ceph:read
Start Upgradeiam:project:infrastructure:ceph:write
Pause Upgradeiam:project:infrastructure:ceph:write
Resume Upgradeiam:project:infrastructure:ceph:write
Stop Upgradeiam:project:infrastructure:ceph:write
View Historyiam:project:infrastructure:ceph:read

​
Upgrade Status

StatusDescription
Up to DateCluster is running the latest available version
AvailableNewer version is available for upgrade
In ProgressUpgrade is currently running
PausedUpgrade is paused and can be resumed

​
How to Check Available Upgrades

1

Select Cluster

Choose a Ceph cluster from the cluster dropdown.
2

View Current Status

Review the statistics cards showing:
  • Current Version: The version currently running
  • Target: Available upgrade target or “Up to Date”
  • Progress: Percentage complete (if upgrading)
  • Daemons: Number of daemons in the cluster
  • Health: Current cluster health status
3

Check for Blockers

If blockers exist, they will be displayed with warnings. Common blockers include:
  • HEALTH_ERR: Cluster has critical health errors
  • Active maintenance operations
  • Insufficient redundancy

​
How to Start an Upgrade

1

Click Start Upgrade

Click the Start Upgrade button to open the upgrade wizard.
2

Pre-flight Check

The wizard checks cluster health and readiness:
  • Cluster health status (HEALTH_OK, HEALTH_WARN, HEALTH_ERR)
  • Critical blockers are highlighted
Critical blockers (like HEALTH_ERR) must be resolved before upgrading. Non-critical warnings allow proceeding with caution.
3

Select Version

Choose your upgrade target:
  • Available Versions: Pre-defined versions from the registry
  • Custom Image: Specify a container image URL directly
The recommended version is highlighted if available.
4

Confirm Upgrade

Type UPGRADE to confirm you want to proceed. This prevents accidental upgrades.
5

Monitor Progress

The wizard shows real-time progress:
  • Overall progress percentage
  • Current daemon being upgraded
  • Daemon upgrade status (pending, upgrading, upgraded)
IMPORTANT: Upgrades can take significant time depending on cluster size. Do not interrupt power or network during the upgrade process.

​
How to Use Custom Image Upgrades

Custom image upgrades allow you to specify an exact container image URL, useful for:
  • Testing pre-release versions
  • Using custom-built images
  • Air-gapped environments with private registries
1

Start Upgrade Wizard

Click Start Upgrade and complete the pre-flight check.
2

Select Custom Image

Toggle the Use Custom Image option in the version selection step.
3

Enter Image URL

Enter the full container image URL, for example:
  • quay.io/ceph/ceph:v18.2.4
  • registry.example.com/ceph/ceph:v18.2.4
4

Confirm and Start

Complete the confirmation step to begin the upgrade.

​
How to Pause an Upgrade

Pausing stops the upgrade at the current daemon while maintaining cluster stability.
1

Locate Pause Button

During an active upgrade, the Pause button appears in the upgrade controls.
2

Click Pause

Click Pause to stop the upgrade process. The current daemon will complete before pausing.
3

Verify Paused State

The status changes to “Paused” and the cluster remains in a mixed-version state.
Pausing is useful when you need to investigate issues or perform other maintenance. The cluster remains operational during the pause.

​
How to Resume an Upgrade

1

Verify Paused State

Ensure the upgrade shows as “Paused” in the status.
2

Click Resume

Click the Resume button to continue the upgrade from where it stopped.
3

Monitor Progress

The upgrade continues with the next daemon in sequence.

​
How to Stop an Upgrade

Stopping an upgrade cancels the process entirely.
1

Click Stop

Click the Stop button during an active or paused upgrade.
2

Confirm Stop

Confirm that you want to stop the upgrade.
Stopping an upgrade leaves the cluster in a mixed-version state. While Ceph supports mixed versions, it’s recommended to either complete the upgrade or roll back manually.

​
How to View Upgrade History

1

Navigate to History Tab

Click the History tab on the Upgrade page.
2

Review Past Upgrades

The history table shows:
  • From/To versions
  • Status (success, failed, cancelled)
  • Start and completion times
  • Duration
  • Number of daemons upgraded
3

Filter Results

Use search and pagination to find specific upgrade records.

​
Statistics Cards

​
Current Version

The Ceph version currently running on the cluster (e.g., 18.2.2 reef).

​
Target/Available

Shows either:
  • Up to Date: No newer version available
  • Available: A newer version is available to upgrade to
  • Target: The version being upgraded to (during upgrade)

​
Progress

Percentage of daemons that have been upgraded. Only shown during active upgrades.

​
Daemons

Total number of Ceph daemons that will be upgraded.

​
Health

Current cluster health status:
  • HEALTH_OK: All healthy
  • HEALTH_WARN: Warnings present (upgrade can proceed with caution)
  • HEALTH_ERR: Critical issues (upgrade blocked)

​
Upgrade Order

Ceph upgrades daemons in a specific order to maintain cluster stability:
  1. Managers (mgr) - Updated first as they coordinate the upgrade
  2. Monitors (mon) - Quorum is maintained throughout
  3. OSDs - Updated one at a time to preserve data availability
  4. Metadata Servers (mds) - For CephFS clusters
  5. RADOS Gateways (rgw) - For object storage clusters

​
Troubleshooting

  • Check cluster health with ceph health detail
  • Resolve any HEALTH_ERR conditions
  • Ensure all OSDs are up and in
  • Verify no other operations are in progress
  • Each daemon restart takes time
  • Large OSDs may take longer to restart
  • Check for recovery/backfill operations
  • Network bandwidth affects image download speed
  • Check the daemon’s host is accessible
  • Verify the container image is available
  • Review daemon logs on the host
  • Consider pausing and investigating
  • Mixed versions are supported but not ideal
  • Resume the upgrade to complete it
  • If issues persist, check version compatibility
  • Contact support for rollback procedures
  • Verify the image URL is correct
  • Check registry authentication
  • Ensure nodes can reach the registry
  • Test with podman pull <image> on a node
  • Versions are managed by the system administrator
  • Check if versions are marked as active in settings
  • Verify your cluster’s current version name matches version groups

​
FAQ

Upgrade duration depends on:
  • Number of daemons in the cluster
  • OSD sizes (larger OSDs take longer to restart)
  • Network speed for image downloads
  • Any recovery operations that occur
Small clusters may complete in under an hour; large clusters can take several hours.
Yes. Ceph’s rolling upgrade process maintains cluster availability:
  • Only one daemon is upgraded at a time
  • Monitor quorum is preserved
  • Data redundancy protects against OSD restarts
Clients may experience brief latency spikes during daemon restarts.
Ceph doesn’t support automatic rollback. Options include:
  • Restore from backup (if available)
  • Manually reinstall previous version (complex)
  • Complete the upgrade and address issues
Prevention is best: test upgrades in a staging environment first.
Generally, you should upgrade to the next major version only. For example:
  • Quincy → Reef (supported)
  • Pacific → Reef (not recommended, upgrade to Quincy first)
Check Ceph’s official upgrade documentation for supported paths.
It depends on the warning:
  • Minor warnings (clock skew, nearfull): Usually safe to proceed
  • Data-related warnings (degraded PGs): Resolve first
  • OSD warnings (down OSDs): Fix before upgrading
Review each warning to understand its impact.
Ceph is resilient to interruptions:
  • Completed daemons retain their new version
  • In-progress daemon may need manual recovery
  • The upgrade can be resumed once power is restored
  • Data is protected by replication
Signs of successful upgrade:
  • Progress shows 100%
  • Current Version matches Target Version
  • Cluster health returns to HEALTH_OK
  • All daemons show the new version
Check the History tab for the official completion record.