We maintain a system that hold users repository on Win2008R2 Domain Controllers.
we use winbind to authenticate users with DC, to get groups information etc.
out Linux box is running RHEL5.3. we took off the stock samba (3.0.33) because of compatibility issues (PIPE_DISCONNECT) and installed 3.4.4.
in normal cases, all works well. but when there is a DC outage it takes winbind a very long time to regain functionality with another DC on the domain. this can take either 6 minutes, or sometime even longer.
during that time winbind is blocking.
if winbind is restarted - the alternative DC is found quickly, and all works well.
looks like there is some problems with the mechanisms that manage connections, detect dead server, and kill old connection and move to a new connection.
Thanks for your bug report!
Metze: can this be related to the winbindd reconnect bug you have recently investigated?
Yes, looks very likely
Can you try the patches attached to bug 7295?
Or use the current v3-5-test branch
I'm working on a backport for 3.4
Will try the patch on 3.5 and see if it helps.
Any results on this with 3.5.10?
3.4.14 should also have the related fixes
This should be in current 3.4.x, 3.5.x, 3.6.x and master