[EnMasse] Upcoming REST API changes

[Ulf Lilleengen]

On 05/08/2018 03:59 PM, Lohmann Carsten (INST/ECS4) wrote:
> Hi,
> 
>> To create address space:
>>
>> POST /apis/enmasse.io/v1/namespaces/myproject/addressspaces
>> Return: the created address space
>>
>> The 'namespace' field in the metadata previously decided in which namespace the
>> messaging infrastructure was created. This is now instead indicated by the
>> 'enmasse.io/namespace' annotation and should be regarded as an implementation
>> detail as we eventually move towards sharing the infrastructure between address
>> spaces.
> 
> I think, naming the annotation something like 'enmasse.io/messaging-infrastructure-namespace' instead of 'enmasse.io/namespace' could make that relationship clearer.
> That would make it clear, that the namespace containing address-controller, keycloak, etc. is not meant here.

Good point, agreed!

> 
> 
>>>      Presumably each CRD needs to have a unique name/identity though?
>>>      Using kubectl, how would you delete an address? I guess one approach
>>>      would be to prefix the address with the address-space name/id to get
>>>      the addresses own identity?
>>>
>>>
>>> such prefixing can't (easily) produce guaranteed uniqueness though,
>>> for simplicity, assume the separator character is - : address space
>>> a-b with address c would give the same compound address as space a,
>>> address b-c
>>
>> You could reserve some special character as the separator and not allow it in
>> address names (so one restriction, but much less than implied by kubernetes
>> names). (e.g. # @ $...)
>>
>> Not ideal though I agree. I'm not seriously proposing that either, just thinking
>> aloud.
> 
> Perhaps using "." as a separator for the address Kubernetes name (i.e. "addressSpaceName.addressName") and restricting address space names to not contain ".".
> There are anyway not many special separator chars to choose from (K8s names: "..maximum length of 253 characters and consist of lower case alphanumeric characters, -, and .").

If we go down this route, I think '.' is a good proposal as the 
separator. I'm starting to get convinced on embedding the addresses in 
the address space definition though :)

> 
> (Another idea in order to simplify the (K8s-) naming of the addresses would be to restrict 1 Namespace to only have 1 AddressSpace - kind of having a AddressSpace CRD singleton per Namespace. If messaging infrastructure is in another Namespace anyway, that could possibly be not too much of a restriction. But I guess it would feel unnatural in terms of the K8s API and could prevent some valid use cases.)
> 
> Best regards
> Carsten Lohmann
> 
> (INST/ECS4)
> Bosch Software Innovations GmbH | Ullsteinstr. 128 | 12109 Berlin | GERMANY | www.bosch-si.com
> 
> Sitz: Berlin, Registergericht: Amtsgericht Charlottenburg; HRB 148411 B
> Aufsichtsratsvorsitzender: Dr.-Ing. Thorsten Lücke; Geschäftsführung: Dr. Stefan Ferber, Michael Hahn
> 

-- 
Ulf