We are having a bizarre problem as well. I had a crash the other day (ended up being the Oracle SCAN listener). I was running 3.8.2 and decided to upgrade to 3.9.3 to test (before i realized it was the DB causing issues).
We have 3 rosters setup in LDAP groups. They have worked fine for an eternity (the 3 years I've been working here). After the upgrade, 3 of my Spark users are only seeing the roster that their account belongs to, not the other 2. We've tried everything, have them login to another PC, try Pidgin , no avail.
The bulk of my users are running Pidgin and are not experiencing the issue. I'm on Mac OS-X running Yosemite 10.10.1 beta using Adium and have no problems at all. I've bounced the openfire service a few times and it has not fixed the roster. One of the 3 users had her rosters just auto-magically re-appear 3 days later, the other 2 are still having issues.
One of my Pidgin users was getting a SSL handshake error and could not validate the self-0signed cert anymore, I had to put her on Spark and when i did, she lost her rosters too.
What can i do to fix this? I've seen talk about downgrading.
I run Centos 6.5 x64 and have installed openfire via the RPM bundle. How exactly does one downgrade? Can i simply stop the service, force install the rpm and restart? The last thing i need is this thing to break. We're in the midst of a heavy development cycle and the last thing i need is this to break.
I'm at a loss here, what can i do?