[
https://issues.apache.org/jira/browse/ARTEMIS-1900?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16499763#comment-16499763
]
Lionel Cons commented on ARTEMIS-1900:
--------------------------------------
There are two scenarios (in that simple failing test).
For queues, all clients (producers and consumers) start at the same time. The
producers stop when they have sent enough messages. The consumers stop when
haven't received new message for a given time.
For topics, the only difference is the starting phase: all consumers start
first (with receipts to make sure the subscription is active), then all
producers start.
In both cases, each consumer uses a different subscription id.
In all cases, the "standard" STOMP setup (with anycastPrefix=/queue/ and
multicastPrefix=/topic/) is used.
> Unexpected STOMP subscription errors (AMQ339016+AMQ119017)
> ----------------------------------------------------------
>
> Key: ARTEMIS-1900
> URL: https://issues.apache.org/jira/browse/ARTEMIS-1900
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Reporter: Lionel Cons
> Priority: Major
>
> When stress testing Artemis (latest snapshot) using STOMP, I sometimes see
> subscription creation errors (AMQ339016). This is not reproducible so this is
> probably a concurrency issue.
> The STOMP client receives an {{ERROR}} frame that contains in its {{message}}
> header:
> {code}
> AMQ339016 Error creating subscription xyz
> {code}
> and in its body:
> {code}
> AMQ119017: Queue abc does not exist.
> {code}
> Also, this error is only sent to the client and not logged by the broker.
> IMHO, every time the broker reports a fatal client error (i.e. STOMP
> {{ERROR}} frame) it should also log this as a warning on its side. Let me
> know if this is specific to this case or if I should log a separate Jira
> issue to track this.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
