I previously have a database in Mongo 1.8 that was running master/slave replication.

The data store has three databases: local, admin, and appdb

I upgraded directly to Mongo 2.4.3, and I am starting to get sporadic master/slave replication failure. Specifically, I am seeing on the slaves:

Sat Jun 1 14:48:29.702 [replslave] Socket recv() timeout 127.0.0.1:37017
Sat Jun 1 14:48:29.703 [replslave] SocketException: remote: 127.0.0.1:37017 error: 9001 socket exception [3] server [127.0.0.1:37017]
Sat Jun 1 14:48:29.703 [replslave] DBClientCursor::init call() failed
Sat Jun 1 14:48:29.703 [replslave] repl: dbclient::query returns null (conn closed?)
Sat Jun 1 14:48:29.704 [replslave] repl: sleep 3 sec before next pass
Sat Jun 1 14:48:32.710 [replslave] repl: syncing from host:localhost:37017
Sat Jun 1 14:48:33.080 [replslave] replHandshake res not: 0 res:

{ ok: 0.0, errmsg: "unauthorized" }

And on the master:

Sat Jun 01 09:49:39.024 [conn1767] authenticate db: local

{ authenticate: 1, nonce: "XXXXXXXXXX", user: "repl", key: "XXXXXXXXX" }

Sat Jun 01 09:49:39.118 [conn1767] command denied:

{ handshake: ObjectId('XXXXXXXX-changes every time it outputs') }

The error is not always present. I have three slave instances. One is replicating fine with no error messages. The other two would start replication, but at some point encounter the "unauthorized" errmsg.

I tried deleting and re-creating the slave database, including adding the "repl" user to the local database. That works for a while, then the error comes up again.

I tried changing the "repl" user on the master local database, including removing the old "readOnly" field and adding in roles ['read', 'readWrite', 'dbAdmin'], but it still doesn't work.