[Samba] replPropertyMetaData & KCC issues after updating to Samba 4.5.0

[Garming Sam]

On 23/09/16 00:59, lingpanda101 at gmail.com wrote:
> For clarification I'll add a few things.
>
> I initially deleted all the NTDS site links for each site and allowed
> the new KCC to create them. However it did not create them I believe
> correctly. By that I mean it defined what appeared to be a bridgehead
> server at each site. So I disabled the new KCC
> 'kccsrv:samba_kcc=false' in my smb.conf and allowed the full mesh to
> be used again. After all site links were recreated. I then switched
> the 'kccsrv:samba_kcc=true' in my smb.conf and that's what prompted
> the following errors above.
>
> To further expand on my Topology, I have 3 sites. I'll call them A,B
> and C. Each site contains 2 DC's. Sites use different subnets and are
> connected via. fiber. Sites B and C should not be replication
> partners. They should only replicate with Site
> A(Default-First-Site-Name). With the new KCC after deleting all the
> NTDS links, Sites B and C Domain Controller #1 becomes the bridgehead
> server for that site. Domain Controller #2 at sites B and C only
> replicates with Domain Controller #1 at it's respective site. So if
> the bridgehead server goes down, Domain Controller #2 at sites B and C
> will no longer receive changes.
>
> The new KCC does prevent sites B and C from replicating with each
> other. That is correct. This isn't a huge issue for me. I can continue
> using the old KCC for now. The full mesh isn't detrimental to my
> network. Don't want to take up too much of your time. Thanks
>
>
The KCC has been my pet project for the last little bit, so I am very
interested in how it functions in general. But as far as I can tell, the
KCC is doing what is expected of it. What should happen, and I say
should, is that if the bridgehead server dies, the bridgehead server
role will transfer to the other DC. There might be a brief period of
time before the KCC re-runs where the sites are disconnected, but in
general, the failovers should be relatively stable. With only a small
number of sites (and DCs), this might be more trouble than it's worth,
like you say. In either case, I appreciate your input.


Thanks,

Garming