Core Issue
In the middle of an subscriber upgrade, the subscriber upgrade fails. The publisher is upgraded and currently the cluster operates in a stand alone configuration. These symptoms and behavior are reported:
- After a reset, IP phones receive a re-order tone.
- The subscriber thinks it is primary and has phones registered to it.
- The publisher thinks it is primary and has phones registered to it.
- Most phones think the subscriber is primary and the publisher is standby.
Resolution
To resolve this issue, perform these steps:
- Reboot the publisher.
- During the upgrade, reboot all the machines in the cluster so that they can be synchronized.
Start with fresh caches and have a common point of reference. In this particular case, the subscriber did not have a robust name resolution method. The HOSTS and LMHOSTS were not converged and were in conflict.
It is important that both of these files are synchronized with each other. Although Network Basic Input and Output System (NetBIOS) procedures are falling out of favor, they have not yet have reached the point where they can be dispensed. Important processes mostly related to database replication rely on NetBIOS services.
Note: The Microsoft Structured Query Language (SQL) server uses NetBIOS name resolution for some types of Remote Procedure Call (RPC) tasks. Because of this, the publisher and all subscribers should use the LMHOSTS file for NetBIOS name resolution. All servers in the cluster should use HOSTS files for proper name resolution between the servers.
- Edit a text file and reboot.
For more information about SQL servers and NetBIOS, refer to these documents: