[Freeipa-devel] [PATCH 0011] check-for-existing-and-self-referential-segments

[Petr Vobornik]

On 06/10/2015 02:26 PM, Ludwig Krispenz wrote:
> Hi Petr,
> On 06/08/2015 04:50 PM, Ludwig Krispenz wrote:
>>
>> On 06/08/2015 04:47 PM, Petr Vobornik wrote:
>>> On 06/03/2015 06:20 PM, Simo Sorce wrote:
>>>> On Wed, 2015-06-03 at 14:53 +0200, Ludwig Krispenz wrote:
>>>>> Hi,
>>>>>
>>>>> this should prevent adding duplicate segments or segments with same
>>>>> start and end node
>>>>
>>>> LGTM!
>>>>
>>>> Simo.
>>>>
>>>
>>> The self referential check is done only in ipa_topo_pre_add. But it
>>> is still possible to create self referential in mod.
>>>
>>> Interesting thing is if I:
>>> - have segment (A, B)
>>> - modify it to (A, A) (success)
>>> - add (A, B), got: "Server is unwilling to perform: Segment already
>>> exists in topology or is self referential. Add rejected."
>>> - removal of (A, A): "Server is unwilling to perform: Removal of
>>> Segment disconnects topology.Deletion not allowed." note that, there
>>> are also: (A, D) and (A, C) segments.
>>>
>>> ACK if it will be addressed in separate patch.
> did you push this patch ?

Pushed to master: 777a9500ceba11e6adbd85306f460e8a320504cb

>> yes, it will be.
> but it will take more work, if we want to properly allow mods to change
> connectivity and endpoints, then we would need to check if the mod
> disconnects the topology, delete existing agreements, check if the new
> would be a duplicate and create new agmts. There could be some difficult
> scenarios,like having
>
> A <--> B <--> C <--> D,
>
> if you modify the segment B-C to A-D topology breaks and is then
> reconnected.
>
> So I think we should reject segment mods affecting endpoints of the
> segment, at least for alpha, beta ...
>> You find interesting scenarios :-)
>>

I think it's even more proper. IMHO, this operation should be done by 
"add" and "del" anyway.

So then, so we should add "no_update" flag in ipalib and reject it in 
topology plugin. I'll adjust the ipalib part.

-- 
Petr Vobornik