A winbindd architecture question for the winbindd gurus out there...

Jeremy Allison jra at samba.org
Fri Jan 18 01:37:05 GMT 2008


On Thu, Jan 17, 2008 at 05:15:54PM -0800, Tim Prouty wrote:
> It was a bug in our code that doesn't exist in samba mainline.  Our  
> logic in rescan_trusted_domains() was broken such that it would rescan  
> 100 times every 5 minutes instead of once (ouch!).   Combine that with  
> some unreachable DCs, and this bug was causing us to flood the the  
> winbindd children such that the requests queue on the parent filled up  
> faster than the child could empty it.  Running in this state for any  
> length of time caused the parent process to reach the max process size  
> and subsequently core dump.  In a sense, it wasn't a memory leak,  
> since all of the memory was still accounted for, but it was still ugly.
> 
> Thanks again for your help in understanding the winbindd code!

No problem - thanks for that (I was worried we had an undiscovered bug
:-).

Jeremy.


More information about the samba-technical mailing list