We have a question that relates to differences between MongoDB versions 4.0 and 4.4.
We have observed this difference in our production MongoDB cluster (managed service) with three nodes, running in 3 different regions of one of the major cloud providers (CentralUS, EastUS, and CanadaCentral) with the preferred region of CentralUS. This preference is important to us as one of our applications is extremely latency sensitive and is also deployed in the CentralUS region.
Prior to the upgrade (4.0), after an election, the cluster would honor the preference (i.e. automatic failover to the preferred region post-election). After the upgrade (4.4, skipping 4.2), this preference is frequently not honored, and requires a manual failover (sometimes multiple times, as we cannot choose the new primary node due to being a managed service). We have successfully reproduced this behavior with an increased cluster size of 5 nodes (version 4.4).
Does anyone have information on differences between 4.0 and 4.4 that could result in this behavior or have additional context from their experience? If so, it would be very helpful.