-
Notifications
You must be signed in to change notification settings - Fork 2.3k
Description
Feature Description
We encountered an issue while attempting to upgrade Vitess from version 20 to 22: the replica upgrade failed to start correctly.
Upgrading both the primary and replica versions simultaneously is required for a successful upgrade.
After confirmation, the reason is that PR #17031 did not handle the situation where the primary was not upgraded. When the ServerUuid of PrimaryStatus cannot be obtained, the replica cannot start normally.
This is a disruptive upgrade. Could you please add handling for when ServerUuid is empty to ensure that replicas are compatible with the v20 primary during the upgrade?
Use Case(s)
We hope to upgrade the replica first, and then upgrade all components one by one.
Environment:
Kubernetes v1.32
vitess-operator v15
Current Vitess version: v20.0.4
Vitess v21.0.6 (verified)
Vitess v22.0.0 cannot be rolled over.