maas-devel team mailing list archive

[Julian Edwards <julian.edwards@xxxxxxxxxxxxx>]

On Wednesday 05 Feb 2014 14:31:17 Raphaël Badin wrote:
> On 02/05/2014 02:54 AM, Julian Edwards wrote:
> > https://bugs.launchpad.net/maas/+bug/1250435
> > 
> > One solution for this is to simply ditch the CNAMEs and use the actual
> > chosen host name in the A records.  This would mean a slightly intrusive
> > change so that we don't fill the zone with the old N-N-N-N style first.
> > 
> > Can you guys remember why we did this in the first place, and the
> > ramifications of not doing it?
> 
> IIRC we did this to simplify the writing of reverse zone files: they are
> written once and for all.  See the following thread
> https://lists.launchpad.net/maas-devel/msg00355.html for details.
> 
> > Alternatively, the reason this bug crops up is because the node's hostname
> > is set to the CNAME rather than the A record.  We could do a quicker fix
> > if we use the A record as its hostname.  The CNAME would still exist for
> > humans, but this should fix picky apps like Postgres and SSH.
> 
> A less dramatic change is to hand out the generated hostname to clients
> like Juju.  The user-provided hostname would still be there for
> convenience but the charm would manipulate A records of the form
> "N-N-N-N.domain."

That is not enough to work, the actual hostname on the node needs to change as 
well.

I don't think changing the hostname (plus a change to hand that to juju) is a 
lot of work and would fix this problem.

As you say, the CNAME remains as a convenience.