Kibana saved-object migration behavior during major version rolling upgrades

Question

When performing a rolling upgrade of Kibana instances behind a load balancer, the system triggers an automated saved-object migration on the .kibana system indices upon the first instance's startup. While Elasticsearch supports rolling minor-version upgrades, Kibana requires a strict version match between the application and the cluster to function.

The primary concern involves the persistence of custom objects—such as dashboards, index patterns, and alerting rules—when moving across major version boundaries. If a specific object type is deprecated or removed in the target version rather than transformed, the migration process may fail or result in data loss within the Kibana state.

How does Kibana handle saved objects that have been completely removed in the newer major version during a rolling cutover? Is there a mechanism to prevent the first instance from initiating an irreversible migration before the compatibility of all custom objects is validated?

Siddharth · Answer

Answer to the question
When a Kibana node starts in a rolling upgrade, it automatically runs the saved‑objects migration on the .kibana system index. The migration is applied once and is not reversible unless the index is restored from a snapshot. If a particular object type (for example, a legacy visualization or a deprecated alert rule) has been removed in the target major version, the migration deletes every instance of that type from the index. This deletion is irreversible in the running cluster.
What the migration does

Runs on the first upgraded node only.
Applies all schema changes defined for the new Kibana version.
Drops objects of types that no longer exist in the new release.
Leaves objects of unchanged types untouched.

No built‑in pause or validation hook
Kibana does not provide a mechanism to stop or defer this migration based on the presence of custom objects. The migration is triggered automatically on node startup, regardless of load‑balancer traffic or health‑checks. Therefore, an administrator cannot rely on Kibana to “wait until all nodes are upgraded” before performing the migration.
Recommended approach for a safe rolling cutover

Review release notes. Identify any object types that have been removed or majorly changed in the target major version.
Take a snapshot of the .kibana index. Use PUT /_snapshot/repo/name before starting the upgrade.
Verify the current state. In Dev Tools, run GET .kibana/_search?q=type:removed_type to see how many objects would be lost.
Upgrade the first node. Monitor its logs for a line similar to Saved objects migration completed.
Validate post‑migration. Check that critical dashboards, index patterns, and alert rules still exist.
Upgrade remaining nodes. Once all nodes are on the new version and the system is stable, you can restore the snapshot if you need to recover lost objects.

When you need to recover
If the migration has deleted objects you need, restore the pre‑upgrade snapshot to a separate test cluster and copy the missing objects back to production. This is the only way to recover data that was dropped during the migration.
Missing diagnostic detail
To fine‑tune the recommendation, it would help to know the exact Kibana major versions you are moving between (e.g., 7.x → 8.x). Different major releases may have different sets of removed object types.

Kibana saved-object migration behavior during major version rolling upgrades

1 answer

Answer to the question

What the migration does

No built‑in pause or validation hook

Recommended approach for a safe rolling cutover

When you need to recover

Missing diagnostic detail

Your perspective belongs here.

0 question comments