We've seen a few problems lately, and I'm hoping someone can offer insight on resolving them. We are currently on 1151296 on machines that are definitely not overloaded on mem/CPU/IO/network.
1) When moving from build 1151296 from 1150478 the index format changed, or some other marker that the newer version choked on or would not accept. Was this expected? 2) When the master is restarted the slaves (currently) will sometimes declare that the entire index is invalid and will re-download it 3) Slaves will stop getting updates. It is not known if this is a push problem or a pull problem, but definitely a problem problem. We have allowed the slaves to replicate for a week without problems, but also without queries. Thanks for any insight or input you might have!