activemq-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Antoine Toulme <antoine.tou...@gmail.com>
Subject Re: ARTEMIS-806 discussion WAS: Re: [VOTE] Apache Artemis 1.5.0
Date Mon, 07 Nov 2016 19:04:41 GMT
I am pinging you right now but it looks like you left the channel. Sorry I missed you.

I guess I might miss you again so maybe I'll try to do this on the dev list instead.

The problem exposed by ARTEMIS-806 is the following:
-We create a cluster of artemis servers running, as per the symmetric cluster example.
-We create a MQTT subscriber to "foo/#", on one of the nodes.
-We create a MQTT publisher to "foo/bar" to another node.

Expected:
-The subscriber should receive the message.

Actual:
-The subscriber does not receive the message. After a wait of 10s, we shut down the cluster.

Additional findings:
-If we change the subscriber subscription pattern to 'foo/bar', the message is received correctly.
-If we change the publisher to publish to the same node the subscriber is listening to, the
message is received correctly.
-If we use another protocol, the subscriber and publisher are able to communicate.

We have therefore a bug that is at the intersection of clustering and MQTT pattern subscription.

To go further, can you explain to me what the behavior of Artemis is when registering a pattern
subscription in a cluster? Is each node broadcasting to the cluster that it is subscribing
to those messages?
If yes, how is that logged, and is there a way for us to verify the subscription is broadcasted
to the rest of the cluster?

It also looks like all internal Artemis routing is done over the core protocol. Are the MQTT
subscriptions translated into an equivalent core subscription? Where is the code performing
this transformation?

Thanks,

Antoine

> On Nov 7, 2016, at 6:45 AM, Martyn Taylor <mtaylor@redhat.com> wrote:
> 
> Hi Antoine,
> 
> I am on UTC time, usually around on IRC from 9am - 5pm.  Feel free to ping
> me.
> 
> Cheers
> 
> On Sat, Nov 5, 2016 at 12:05 AM, Antoine Toulme <antoine.toulme@gmail.com>
> wrote:
> 
>> Great! I am tmio on IRC, when is a good time to catch you? I tried to find
>> you when I filed the bug but was probably too late in your day.
>> 
>>> On Nov 4, 2016, at 12:23 PM, Martyn Taylor <mtaylor@redhat.com> wrote:
>>> 
>>> Hi Antoine,
>>> 
>>> I'm sorry your issue has not been addressed yet, I've been meaning to
>> look
>>> into this myself, but have been snowed under with other things recently.
>>> We should be able to get this resolved and into to a release soon.
>>> 
>>> I'm more than happy to help you get started with the code base, the IRC
>>> channels is great for this type of discussion.  My nick is martyntaylor
>>> feel free to ping me,
>>> 
>>> Thanks
>>> Martyn
>>> 
>>> 
>>> 
>>> On Fri, Nov 4, 2016 at 6:57 PM, Antoine Toulme <antoine.toulme@gmail.com
>>> 
>>> wrote:
>>> 
>>>> I would absolutely love to help with a fix. That’s why I created the bug
>>>> and provided reproduction steps.
>>>> 
>>>> However I was stomped with the code itself and have had no luck
>>>> reproducing this at smaller scale.
>>>> 
>>>> I did ask on the mailing list for help and steps to help move this
>> forward
>>>> further. Your help in unblocking me, from pointing which area of the
>> code
>>>> would be a likely culprit, to pointing me to better instructions to
>> debug
>>>> interactions in cluster with finer logging,
>>>> helping me understand better the messaging model of symmetric clusters,
>>>> or hinting at the status of MQTT support and what to reliably expect
>> would
>>>> be very appreciated.
>>>> 
>>>>> On Nov 4, 2016, at 11:25 AM, Clebert Suconic <
>> clebert.suconic@gmail.com>
>>>> wrote:
>>>>> 
>>>>> We had a lot of fixed done on 1.5.0…
>>>>> 
>>>>> I guess it wasn't clear if your issue was a configuration issue or an
>>>>> actual bug. But we try to fix as many issues we can between each
>>>>> releases. We will get there eventually, maybe you could even give us
>>>>> hand on the fix? that's the beauty of open source.
>>>>> 
>>>>> On Fri, Nov 4, 2016 at 1:17 PM, Antoine Toulme <
>> antoine.toulme@gmail.com>
>>>> wrote:
>>>>>> I have just tried the release bits with the bug I reported
>>>> (ARTEMIS-806) and I see the same incorrect behavior, where MQTT
>> subscribers
>>>> are not notified of messages published to another node in the cluster.
>>>>>> 
>>>>>> I therefore am casting +0 (non-binding).
>>>>>> 
>>>>>>> On Nov 4, 2016, at 10:13 AM, Christopher Shannon <
>>>> christopher.l.shannon@gmail.com> wrote:
>>>>>>> 
>>>>>>> +1 (binding)
>>>>>>> 
>>>>>>> On Fri, Nov 4, 2016 at 10:53 AM, Fabio Gomes dos Santos <
>>>>>>> supergrilo@gmail.com> wrote:
>>>>>>> 
>>>>>>>> +1
>>>>>>>> 
>>>>>>>> 2016-11-03 19:29 GMT-02:00 Clebert Suconic <
>> clebert.suconic@gmail.com
>>>>> :
>>>>>>>> 
>>>>>>>>> Hello all,
>>>>>>>>> 
>>>>>>>>> I would like to propose an Apache Artemis 1.5.0 release.
>>>>>>>>> 
>>>>>>>>> 1.5.0 has these new features as highlights.
>>>>>>>>> 
>>>>>>>>> [ARTEMIS-724] - Implement no-local consumer support in
AMQP
>>>>>>>>> [ARTEMIS-737] - Add JUnit Rules
>>>>>>>>> [ARTEMIS-743] - Default the queue address to the queue
name
>>>>>>>>> [ARTEMIS-753] - Persist pause status after restart of
the broker.
>>>>>>>>> [ARTEMIS-756] - Add basic CDI integration
>>>>>>>>> [ARTEMIS-824] - Management operations for creating and
destroying
>>>>>>>>> connector services
>>>>>>>>> 
>>>>>>>>> 
>>>>>>>>> 
>>>>>>>>> And many other fixes as part of this release, the complete
release
>>>>>>>>> notes can be found here:
>>>>>>>>> 
>>>>>>>>> https://issues.apache.org/jira/secure/ReleaseNote.jspa?
>>>>>>>>> version=12338118&styleName=Html&projectId=12315920&Create=Create
>>>>>>>>> 
>>>>>>>>> The binary distributions can be found here:
>>>>>>>>> https://repository.apache.org/content/repositories/
>>>>>>>>> orgapacheactivemq-1109/org/apache/activemq/apache-artemis/1.5.0/
>>>>>>>>> 
>>>>>>>>> The source archives can be found here:
>>>>>>>>> https://repository.apache.org/content/repositories/
>>>>>>>>> orgapacheactivemq-1109/org/apache/activemq/apache-artemis/1.5.0/
>>>>>>>>> 
>>>>>>>>> The Maven repository is here:
>>>>>>>>> https://repository.apache.org/content/repositories/
>>>>>>>> orgapacheactivemq-1109/
>>>>>>>>> 
>>>>>>>>> The source tag:
>>>>>>>>> https://git-wip-us.apache.org/repos/asf?p=activemq-artemis.
>>>>>>>>> git;a=tag;h=refs/tags/1.5.0
>>>>>>>>> 
>>>>>>>>> The project website for that version has been staged
to:
>>>>>>>>> http://home.apache.org/~clebertsuconic/
>>>>>>>>> 
>>>>>>>>> The vote will remain open for 72 hours.
>>>>>>>>> 
>>>>>>>>> [ ] +1 approve the release as Apache Artemis 1.4.0
>>>>>>>>> [ ] +0 no opinion
>>>>>>>>> [ ] -1 disapprove (and reason why)
>>>>>>>>> 
>>>>>>>>> Here's my (binding) +1 vote
>>>>>>>>> 
>>>>>>>> 
>>>>>>>> 
>>>>>>>> 
>>>>>>>> --
>>>>>>>> Fábio Santos
>>>>>>>> supergrilo@gmail.com
>>>>>>>> <http://br.linkedin.com/pub/f%C3%A1bio-santos/1b/20/422>
>>>>>>>> 
>>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> --
>>>>> Clebert Suconic
>>>> 
>>>> 
>> 
>> 


Mime
View raw message