This issue occurs because the MAs continue to communicate with the Replica VA node which is still the old version. The upgrade script is triggered on the Primary VA node which is already the new version.
Resolution
Currently, there is no resolution.
Workaround: To work around this issue, use one of these options:
Option 1: Ensure the Replica node (or all replica nodes if more than one) are upgraded before proceeding with the automatic IaaS upgrade. Note: If not explicitly excluded (by executing touch /tmp/disable-iaas-upgrade), the IaaS upgrade will start automatically after rebooting the primary VA (when the upgrade has completed).
Option 2:
Stop all Management Agent services on all IaaS nodes.
Open their respective ManagementAgent.exe.config files and ensure the list of VAs in the <managemetEndpoints> section starts with the current Primary VA.
Restart the Management Agents and trigger the automatic IaaS upgrade.
Option 3: Upgrade IaaS manually using the IaaS installer that can be downloaded from the VAMI installer download page.