[Freeipa-devel] Why do we require DNS record when service is being added?

[Alexander Bokovoy]

On Tue, 14 Jul 2015, Petr Spacek wrote:
>On 14.7.2015 10:08, Alexander Bokovoy wrote:
>> On Tue, 14 Jul 2015, Jan Pazdziora wrote:
>>> On Tue, Jul 14, 2015 at 08:31:19AM +0200, Petr Spacek wrote:
>>>> On 13.7.2015 19:37, Jan Pazdziora wrote:
>>>> >
>>>> > However -- what is the purpose of the DNS check when adding service?
>>>>
>>>> The service is typically a Kerberos service, which usually is not going to
>>>> work if the host does not have DNS record.
>>>
>>> So it's an error about existing *state* of the identity management
>>> system, not an error of the service-add operation itself or error
>>> about the result of that operation. IOW, the code tries to be smarter
>>> than necessary, hitting users who attempt to do things right,
>>> precreating host records. Plus it's an error about related object,
>>> not the object being manipulated / created which in itself is
>>> suspicious.
>>>
>>>> > Shouldn't that check be removed altogether?
>>>> I would rather relax the check so it can detect usage of host-add
>>>> --random/--password and emit a warning instead of hard error.
>>>>
>>>> What do you think about this approach?
>>>
>>> I guess you are then talking about not having that check in the
>>> host-add operation, not service-add:
>>>
>>>     # ipa host-add --random client56.example.test
>>>     ipa: ERROR: Host does not have corresponding DNS A/AAAA record
>>>
>>> Because to face the error during service-add, the user must already
>>> have overriden the error for the host itself.
>>>
>>> So how about:
>>>
>>>     No DNS check / error in host-add when --random is used.
>>>     No DNS check / error in service-add at all.
>> I would still add a warning in service-add "Host ... does not exist in
>> DNS, this service will not be accessible via Kerberos until A/AAAA
>> record for the host will be created".
>
>Yes, this is what I meant - host-add should do the DNS check and spit only
>warning if --random/--password is used. Service-add should require the host to
>exist (as it does now) but again the check should spit a warning instead of
>error if the host was created with --random.
Sounds good to me.

Will you make a ticket?
-- 
/ Alexander Bokovoy