[jira] [Comment Edited] (ARTEMIS-3153) Broker doesn't respect prefixes from AMQP client

["Erwin Dondorp (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ARTEMIS-3153?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17368832#comment-17368832 ] 

Erwin Dondorp edited comment on ARTEMIS-3153 at 6/24/21, 2:48 PM:
------------------------------------------------------------------

Here are 8 test cases that assume that the addresses (and thus their queues) do not exist yet.
 Unique addresses are used for each test-case so that they do not interfere.
 The test cases each either perform one publish/send operation or a receive operation.

In all cases, the default-address-routing-type/default-queue-routing-type are set to the indicated type.

For MULTICAST:
 * producer without prefix: OK, address is created as MULTICAST, no queue is created as expected
 * producer with prefix: FAIL, address still created as MULTICAST, but also the exception "Peer did not create remote endpoint for link, target: multicast:address2m"
 * consumer without prefix: FAIL, address and queue are created, but as ANYCAST; should have been MULTICAST as per the default-routing-type
 * consumer with prefix: OK, address is created as MULTICAST, an anonymous queue is added

Due to the failures in the 2nd and 3rd case, there is a case failing when the prefix is used and also a test case failing when it is NOT used.
 Workaround is to use the consumer with prefix; and producer without it. But that sure feels like a workaround for everyone.

For ANYCAST:
 * producer without prefix: OK, address is created as ANYCAST, queue is also created as expected
 * producer with prefix: FAIL, address still created as ANYCAST, but also the exception "Peer did not create remote endpoint for link, target: anycast:address2a"
 * consumer without prefix: OK, address and queue are created, as ANYCAST
 * consumer with prefix: OK, address and queue are created, as ANYCAST

Almost same situation here, but the handling of the third test case now does not fail. that might be accidental, but it still counts.


was (Author: erwindon):
Here are 8 test cases that assume that the addresses (and thus their queues) do not exist yet.
 Unique addresses are used for each test-case so that they do not interfere.
 The test cases each perform one publish/send operation or a receive operation.

In all cases, the default-address-routing-type/default-queue-routing-type are set to the indicated type.

For MULTICAST:
 * producer without prefix: OK, address is created as MULTICAST, no queue is created as expected
 * producer with prefix: FAIL, address still created as MULTICAST, but also the exception "Peer did not create remote endpoint for link, target: multicast:address2m"
 * consumer without prefix: FAIL, address and queue are created, but as ANYCAST; should have been MULTICAST as per the default-routing-type
 * consumer with prefix: OK, address is created as MULTICAST, an anonymous queue is added

Due to the failures in the 2nd and 3rd case, there is a case failing when the prefix is used and also a test case failing when it is NOT used.
 Workaround is to use the consumer with prefix; and producer without it. But that sure feels like a workaround for everyone.

For ANYCAST:
 * producer without prefix: OK, address is created as ANYCAST, queue is also created as expected
 * producer with prefix: FAIL, address still created as ANYCAST, but also the exception "Peer did not create remote endpoint for link, target: anycast:address2a"
 * consumer without prefix: OK, address and queue are created, as ANYCAST
 * consumer with prefix: OK, address and queue are created, as ANYCAST

Almost same situation here, but the handling of the third test case now does not fail. that might be accidental, but it still counts.

> Broker doesn't respect prefixes from AMQP client
> ------------------------------------------------
>
>                 Key: ARTEMIS-3153
>                 URL: https://issues.apache.org/jira/browse/ARTEMIS-3153
>             Project: ActiveMQ Artemis
>          Issue Type: Bug
>            Reporter: Justin Bertram
>            Priority: Major
>
> The broker doesn't appear to respect an AMQP client using a prefix (e.g. {{anycast://}} or {{multicast://}}) when sending or consuming messages.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)