[Samba] Samba Join as DC failed

[Daniel von Obernitz]

Hi Andrew,


Am 06.03.2014 22:47, schrieb Andrew Bartlett:
>>
>> The domain has about 20000 users, don't know if that can be considered=20
>> as large.
>> The machine has 2 cores and 6GB ram and both machines (AD-clone and=20
>> samba) are on the same esx-host in an isolated net structure...
>
> That is pretty large for Samba's AD DC code, but folks have deployed
> sites that large.  Do you have any idea where it spent the time?  Was it
> CPU busy, or swapping, or something else?

I just retested the join with "time", but it aborted because of not 
enough disk space, which was correct. But until then it ran about 
244minutes.. I will retest it again now to geht the full duration.

I didn't watch the process the whole time, but htop showed the process 
using 1 CPU at 100% all the time.


> We know we need to improve some issues at the large end, it would be
> very interesting to run that join under 'perf record -g' on Linux (with
> a very, very large disk and much more memory) to see what we are hitting
> the worst, to see if we can improve it.
>

Would be glad to help, just give me some concrete values for very large 
disk (and where do I have to mount it - never worked with perf before) 
and how much memory and I'll see, what I can do here...


And I have another question: I joined the Samba using dns BIND9_DLZ .. I 
have to use an already existing bind9-server and the generated 
named-files are designed for a bind9 running on the same machine as 
samba. Are there any manuals how to work with an existing external 
bind9-DNS-Server? When I started samba with "samba -i" I got " Server 
not found in Kerberos database" and a RID-Error which is connected to 
that I think.

Thanks
Daniel