This sounds kind of like retroactive consumers (
http://activemq.apache.org/retroactive-consumer.html); are you using that
feature?
Are you able to reproduce this reliably in a test environment where you can
change things and see the effects? If so, does the problem occur with
non-durable subscriptions?
When the unsubscription doesn't work properly, can you confirm that the
subscription is gone from the broker? You can see that via JMX.
If two messages are sent while the consumer is unsubscribed, does it get
both or only one?
Does each subscription run through the same exact code?
Tim
On Nov 11, 2015 5:54 PM, "wuwufen" <1767549...@qq.com> wrote:
> 3 mqttClent （c1,c2,c3,c4,c5）;mqttclent setClindId,cleansession=false;
> activemq 5.12.0 (or befor)
> test topic(T)
>
>
>
> 1 c1,c2,c3 subscribe T
> 2 c5 send message m1 to T
> 3 c1,c2,c3 received m1
>
> .......sometimes (problem occurs)
>
> 4 c1 unsubscribe T
> 5 c5 send message m2 to T (b c received m2)
> 6 c1 subscribe T
> 7 c1 received m2
>
> 8 c1 unsubscribe T
> 9 c1 subscribe T
> 10 c1 received m2
> :
> :
> if(this problem occurs)
> {
> 11 d subscribe T
> 12 d received m2 //i don't know why
>
> }
>
>
> at last
> 13 b ,c ,d unsubscribe T
> 14 a unsubscribe T
> 15 a subscribe T
> 16 the result change good a din't received m2
>
>
>
>
>
>
>
>
>
>
>
>
>
> --
> View this message in context:
> http://activemq.2283324.n4.nabble.com/a-problem-about-mqtt-and-activemq-tp4703831p4703895.html
> Sent from the ActiveMQ - User mailing list archive at Nabble.com.
>
