It looks as if the driver is unable to connect to secondary if the replica set member id's dont follow a continuous order from _id: 0. Sometimes I use a read preference of 'nearest' when calling the java driver.

Through removal and addition of replica members, I found myself with the following replicaset config:

"_id" : "rs0",
"version" : 29,
"members" : [

{ "_id" : 4, "host" : "54.209.160.30:27017" }

,

{ "_id" : 5, "host" : "54.186.49.236:27017", "priority" : 3 }

,

{ "_id" : 6, "host" : "54.187.243.162:27017", "arbiterOnly" : true }

]

I noticed that in the logs of the primary, the continuous repetition of the following:

2014-05-30T00:01:09.050+0000 [initandlisten] connection accepted from 54.209.160.30:59140 #523020 (18 connections now open)
2014-05-30T00:01:09.293+0000 [conn523020] authenticate db: local

{ authenticate: 1, nonce: "xxx", user: "__system", key: "xxx" }

2014-05-30T00:01:09.453+0000 [conn523020] replset couldn't find a slave with id 1, not tracking 53692f8a311b1b09a941a1c3
2014-05-30T00:01:09.533+0000 [conn523020] end connection 54.209.160.30:59140 (17 connections now open)

The logging statement are the only symptom of the issue. I have not noticed any connectivity problem from the java application. I dont knowif the driver is actually finding the 'nearest' replica member or just choosing primary due to this issue.