I run a mongodb sharded cluster with 1 shard, 1 replica set (2 secondaries, 1 primary, 1 delayed slave).
Frequently, the following message appears in the logs:
[rsMgr] replSet total number of votes is even - add arbiter or give one member an extra vote
But why? The delayed slave is hidden and has a priority set to 0. Therefore, there should be 3 nodes voting.
Also, when I stepdown the primary for maintenance purposes for instance, the applications that query MongoDB are lightly interrupted. The cluster takes 5 to 10 seconds to find a new primary.