On Fri, Sep 06, 2013 at 02:55:48PM +0200, Bolesław Tokarski wrote:
> Hello,
>
> Can somebody confirm me the behaviour of SSSD (we're currently on
> version 1.8.6, but will migrate to whatever comes in Ubuntu 14.04) with
> regards to Kerberos DNS records?
>
> I mean, sssd series 1.8 did not have any special handling for AD, so
> LDAP queries went to provided ldap_uri and Kerberos queries thanks to
> the dns_discover_domains are handled by the DNS SRV records for
>
_kerberos._udp.example.com. Correct me if I'm wrong.
You're right.
>
> The DNS SRV records have a preference, or a priority option. Is this
> taken into account, so that lower priority server is never accessed if
> the higher one answers?
>
Yes, we follow the RFC pretty much completely. Recently we added
additional logic that also prefers servers from the DNS domain that
matches the DNS domain of the client, but only on the same pririty
level.
> I am asking because or AD team implemented some "Disaster Recovery"
> domain controller which they only turn on for 2 hours a week, after work
> of course and I believe the logon time is much longer now. I don't have
> exact details for it, though. They claim that they cannot remove the SRV
> record for the "special" server as it will not replicate the AD
> structure to the server in such a case, so they offered to lower the
> priority, but I'm not sure if that's going to help.
>
> I was also informed that the client should be actually using SRV records
> for the particular site, which don't contain the "special" server.
Does
> the 1.9+ series AD backend solve this particular issue?
The site discovery functionality was added to 1.10
_______________________________________________
sssd-users mailing list
sssd-users(a)lists.fedorahosted.org
https://lists.fedorahosted.org/mailman/listinfo/sssd-users
Is your DNS SRV resolution code packaged in such a way that it could be
reusable by other programs? I was writing something up in python
recently that was going to use SRV records and found almost all of the
implementations to be flawed from the RFC. So a good reusable
implementation would be handy.
-Erinn