[Freeipa-users] TTL in individual DNS records

[Stephen Ingram]

On Mon, Oct 21, 2013 at 9:37 AM, Petr Spacek <pspacek at redhat.com> wrote:

> On 21.10.2013 17:58, Stephen Ingram wrote:
>
>> On Sun, Oct 20, 2013 at 11:44 PM, Petr Spacek <pspacek at redhat.com> wrote:
>>
>>  On 18.10.2013 21:44, Stephen Ingram wrote:
>>>
>>>  I'm using IPA 3.0.x on RHEL 6.4 and trying to setup other zones in DNS.
>>>> I
>>>> notice that regardless of the TTL set in the SOA for the zone, the
>>>> individual records default to 86400. I see there has been previous
>>>> discussion on the list (
>>>> https://www.redhat.com/****archives/freeipa-users/2012-**<https://www.redhat.com/**archives/freeipa-users/2012-**>
>>>> November/msg00158.html<https:/**/www.redhat.com/archives/**
>>>> freeipa-users/2012-November/**msg00158.html<https://www.redhat.com/archives/freeipa-users/2012-November/msg00158.html>
>>>> >
>>>>
>>>> )
>>>> saying that the 86400 value is hard-coded into bind-dyndb-ldap. It
>>>> appears
>>>> as though it might be rectified sometime in the 3.3.x series, however,
>>>> I'm
>>>> wondering if there is a workaround for now. Is there a way to just
>>>> delete
>>>> this value such that the individual records will default to the value in
>>>> the SOA until a real fix comes along?
>>>>
>>>>
>>> For now, the only workaround is to set TTL explicitly for all affected
>>> DNS
>>> names. Sorry!
>>>
>>> $ ipa dnsrecord-mod --help | grep ttl
>>>    --ttl=INT             Time to live
>>>
>>> The most important thing is that SOA TTL is not related to default record
>>> TTL by definition.
>>>
>>> Some details are described here:
>>> https://www.redhat.com/****archives/freeipa-users/2012-**<https://www.redhat.com/**archives/freeipa-users/2012-**>
>>> November/msg00160.html<https:/**/www.redhat.com/archives/**
>>> freeipa-users/2012-November/**msg00160.html<https://www.redhat.com/archives/freeipa-users/2012-November/msg00160.html>
>>> >
>>>
>>
>>
>> Am I reading this correctly then that you have to set for each *record* vs
>>
> I really meant *name*. "ipa dnsrecord-mod" operates on whole DNS name. (It
> also means that all records under single *name* share the same TTL value.)


That's what I thought. I was referring to a name as a record.


>  the *zone*. If so, this makes the DNS part of IPA almost unusable except
>> to
>> those willing to stick with the default 86400 or write a script to handle
>> each record in the zone. I've been following the list for some time, but
>> haven't heard much about usage of the DNS component. And, among the users
>> I
>> speak with no one uses the DNS component. Perhaps this is the reason why?
>> I
>>
> Up to now, nobody have told us that 'DNS part of IPA almost unusable'
> without configuration option for default TTL, so it simply didn't get the
> priority. We have seen stroger demand for DNS views, for example :-)


Understood. Perhaps my use case is different than most. If I were using
scripts, I don't think this would be much of an issue, however, with
several UI users with varying levels of experience, it is difficult if you
want to vary TTL per zone instead of per name. After reading the RFC
referenced in the ticket I see now that the default TTL I was thinking was
part of the SOA is actually a separate entity. And, thus, I now see why IPA
needs to also make this distinction.

>
>  haven't looked at the code yet, but would this be that difficult to fix?
>>  I
>>
> If you are okay with statically configured TTL for all zones, then it is
> five-minute fix. (Simply change the current value and recompile or add a
> new parameter to /etc/named.conf.)
>
Could you please point me to the code where this static definition happens?

>
> If you want to define default TTL per-zone in LDAP, then you have to
> define new attribute in LDAP schema, store the default TTL value in
> zone_register and push it to record parser as necessary.
>
> In https://fedorahosted.org/bind-**dyndb-ldap/ticket/70#comment:7<https://fedorahosted.org/bind-dyndb-ldap/ticket/70#comment:7>you can see that we are trying to cooperate with schema/OID space owner to
> find & standardize some solution.
>
> Any contribution is more than welcome! Join us in the ticket and we can
> discuss various propsals.


I see now why this is not a quick solution. I was unaware that the
attribute to handle this default TTL didn't exist. It looks there are two
ideas on the table (JHogarth and JCholasta) right now. But, from the ticket
discussion, it looks like maybe the new attribute is being added instead
and already in progress?

Steve
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/freeipa-users/attachments/20131021/e78447d1/attachment.htm>