Another cause of Winbind consuming 100% of CPU

[Richard Sharpe]

On Tue, Jan 21, 2014 at 5:05 PM, Jeremy Allison <jra at samba.org> wrote:
> On Tue, Jan 21, 2014 at 04:28:59PM -0800, Richard Sharpe wrote:
>> Another customer hit this. Here is what the log shows:
>>
>> [2014/01/21 18:14:32.243610,  0]
>> winbindd/winbindd_util.c:499(rescan_forest_trusts)
>>   Entry[08]: name: RESORG, DNS: resorg.com, SID:
>> S-1-5-21-343818398-1202660629-839522115, trust_f
>> lags: 22, trust_type: 2
>> [2014/01/21 18:14:32.243610,  0]
>> winbindd/winbindd_util.c:499(rescan_forest_trusts)
>>   Entry[09]: name: BTEST2008, DNS: BXXXXYYYY.somesub.local, SID:
>> S-1-5-21-323814229-503867627-140192
>> 7092, trust_flags: 22, trust_type: 2
>> [2014/01/21 18:14:32.243610, 10]
>> winbindd/winbindd_util.c:527(rescan_forest_trusts)
>>   Following trust path for domain BXXXXYYYY (BXXXXYYYY.somsub.local)
>> [2014/01/21 18:14:32.243610,  0]
>> winbindd/winbindd_util.c:499(rescan_forest_trusts)
>>   Entry[10]: name: CXXXXYYYY, DNS: cxxxxyyyy.someothersub.us, SID:
>> S-1-5-21-323814229-503867627-140
>> 1927092, trust_flags: 22, trust_type: 2
>> [2014/01/21 18:14:32.243610, 10]
>> winbindd/winbindd_dual.c:1358(fork_domain_child)
>>
>> Somehow they got two domains with the same SID.
>>
>> How do we repro that?
>
> Not a clue :-).
>
>> The fix that Jeremy provided seems to fix this as well.
>
> So do you have to repro it ? I'm guessing
> the fix will cover all the cases where we
> could re-initialize a domain on the list,
> which should remove all opportunities for
> this kind of corruption.

Yes, we have a repro now. We think they might have cloned VMs from a
template and forgotten to run sysprep or new_sid etc.

We are going to see if we can setup something like that.


-- 
Regards,
Richard Sharpe
(何以解憂？唯有杜康。--曹操)