Upgrade questions
Find answers to frequently asked questions about IBM Storage Ceph upgrade.
What prerequisites must be met before upgrading?
Ensure that the cluster is in a healthy state (HEALTH_OK), at least one standby manager is available, valid backups exist, and all required IBM entitlements and subscriptions are active. Verify that the target Ceph version is supported for the current operating system and deployment method.
Can the upgrade be done online without taking down the cluster?
Yes. Orchestrated upgrades can be performed online by running the ceph orch upgrade command. Client access remains available during the upgrade, although temporary performance degradation can occur while services are restarted.
How should operating system upgrades be handled?
How do I monitor upgrade progress and manage issues during an upgrade?
Monitor upgrade progress by using the ceph orch upgrade status command and by reviewing cephadm logs. You can pause or resume the upgrade if issues are detected. Monitoring cluster health during the upgrade helps identify and address problems early.
What should I do about clients or client libraries after an upgrade?
After upgrading the cluster, plan to upgrade client packages and libraries to supported versions. Existing clients can continue to operate with older libraries for a limited time, but restarting or migrating clients is required to use newer features and ensure long-term compatibility.
How should licensing and entitlements be handled during or after an upgrade?
Ensure that all required IBM subscriptions and entitlements remain active before and after the upgrade. Verify repository access for the target version and enable any newly required repositories as documented.
What are the risks or failure modes, and how do I recover if an upgrade fails?
IBM Storage Ceph does not support software downgrades after an upgrade. If an upgrade fails, recovery typically involves restoring from a validated backup or redeploying the cluster. IBM strongly recommends testing upgrades in a non-production environment before upgrading production clusters.
The upgrade command does not start because the cluster health is not HEALTH_OK. What should I do?
Orchestrated upgrades require the cluster to be in a healthy state. Resolve any outstanding issues, such as degraded placement groups, down OSDs, or full pools, before starting the upgrade.
Do I need to upgrade clients such as radosgw or librbd immediately?
Client upgrades are not required immediately after a cluster upgrade. However, upgrading clients is recommended to maintain compatibility, receive fixes, and use new features introduced in the upgraded cluster.
The upgrade paused. What should I do?
If an upgrade pauses, review the upgrade status and relevant daemon logs to identify the cause. After resolving the issue, resume the upgrade by using the appropriate orchestration command.
The upgrade caused temporary performance degradation. Is this expected?
Yes. Temporary performance degradation can occur during upgrades as daemons restart and data rebalancing occurs. Staggered upgrades and appropriate tuning can help reduce the impact.
How do I monitor OSD and CRUSH map changes during an upgrade?
Monitor OSD and CRUSH map changes by reviewing cluster status and relevant Ceph commands before and after the upgrade. Detailed validation steps should follow IBM upgrade documentation.