I'm currently running 2.8.0-RC3. Now, I've set up a second server with RC7. I've read that the on-disk index format is different between these two version and I need to start with --repair. But I assume that this is not necessary (yet), because RC7 is empty and doesnt have any indexes yet. I thought the use of a replica set would be a good way to upgrade without service interruption. So, now I connect to RC3 (Primary) and say rs.add("otherserver"). At first this seemed to work and the other server is in "STARTUP2" mode as usual when it's starting sync. But on the primary (as I later found out) the log filled with "ReplExecNetThread-0, Failed to connect to x.x.x.x, errno:99 Cannot assign requested address" errors. After some minutes the primary suddenlty becomes SECONDARY and the database becomes unusable, because there is no PRIMARY I needed some time to find out you need to remove the bad server from the replica set config and do a reconfigure with

to recover from that situation. Is this a bug or am I doing something wrong? I've setup replica sets earlier with previous Mongo version and it worked flawlessly everytime. The ports are reachable in either direction (tested with telnet x.x.x.x 27017).