[EnMasse] Multicast address

Ulf Lilleengen lulf at redhat.com
Fri Jul 7 06:34:30 UTC 2017


On Thu, Jul 6, 2017 at 5:47 PM, Gordon Sim <gsim at redhat.com> wrote:

> On 06/07/17 16:35, Lohmann Carsten (INST/ECS4) wrote:
>
>>
>> How many routers do you have? Is it just one? Did you change the type of
>>> the address (i.e. delete and recreate with a different type) while there
>>> were links still attached?
>>>
>> Yes, just one router. We apply the standard "kubernetes/enmasse.yaml"
>> from EnMasse 0.11.2 and then register the addresses. There were no changes
>> to the addresses afterwards.
>>
>> I think I found the problem here:
>> Our deployment pipeline applies the "enmasse.yaml" and then waits until
>> the "messaging" service exists
>> (we added this so that resolution of ${MESSAGING_SERVICE_HOST} in our
>> hono deployment yaml works).
>> Directly afterwards the addresses get registered.
>>
>> And it seems that is too early - in the sense that the necessary EnMasse
>> initialization is not ready at that point yet.
>> When I put a delay before the address registration, everything works
>> later on.
>>
>> Two question from this:
>> - Do you have a hint on what the condition should be on which to wait
>> before registering the addresses?
>> Checking that "readyReplicas" of the "qdrouterd" deployment is greater
>> zero?
>>
>> - Could it be considered a bug that the address registration request got
>> on OK response at that point?
>> Maybe there should have been a 503 response instead?
>>
>
> Could it be that the receivers are connecting before the addresses have
> fully propagated? (And therefore are subject to the default semantics?)
>
> The actual configuration of the routers is an async process, so the fact
> that the definition of the addresses is accepted ok does not mean they are
> immediately propagated.
>
> I think the bug therefore may be that the readiness indicated in the
> address status does not (yet) cover propagation to the router(s).
>


Yes, the 'status': { isReady: true/false } in 0.11.2 does not cover router
address propagation. This is fixed on master, where the router address
configuration is checked as part of setting the isReady field as well.

Ulf
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/enmasse/attachments/20170707/10937bab/attachment.htm>


More information about the enmasse mailing list