When a receiving transaction is committed in a paging situation,
if a page happens to be completed and it will be deleted in a
transaction operation (PageCursorTx). The other tx operation
RefsOperation needs to access the page (in PageCache) to finish
its job. There is a chance that the PageCursorTx removes the
page before RefsOperation and it will cause the RefsOperation
failed to find a message in a page.
Optionally specify a queue routing type when creating a broker. Example:
"create --queues myqueue,mytopic:multicast". Defaults to anycast if
unspecified.
Update to latest proton-j release and refactor the dispostion code to use
the new type enums to better deal with the dispistions. Updates to Qpid JMS
0.37.0 which still uses the current netty 4.1.28.Final dependency.
When a node tries to reconnects to another node in a scale down cluster,
the reconnect request gets denied by the other node and keeps retrying,
which causes tasks in the ordered executor accumulate and eventually OOM.
The fix is to change the ActiveMQPacketHandler#handleCheckForFailover
to allow reconnect if the scale down node is the node itself.
Added new cases:
- export from anycast queue, then import to multicast topic where number of queues > 1
- export from anycast queue, then import to multicast fqqn where number of queues > 1
- export from multicast topic, then import to anycast queue (both address)
- export from anycast queue, then import to multicast topic (both address)
Refactor ServerJMSMessage so it correctly transposes all JMSX headers.
Push common JMSX mappings for JMS to Message Interface mappings into MessageUtil to avoid duplication in ActiveMQMessage and ServerJMSMessage
With the current Maven config, the maven-source-plugin is executed twice, this results in the sources-jars being attached twice which leads the deploy goal in trying to upload the same sources jars twice to the repository. This slows down the build process and also fails on repositories that disallow updating of existing artifacts.
Additionally the "jar" goal is now replaced with "jar-no-fork" which prevents the maven life cycle being executed twice too.
Not closing the InputStream makes this test flaky on Windows. The test
breaks because FileMoveManager::delete(java.io.File) Line 221 fails
to delete the file if it's still "owned" by the JVM process on Windows.
Jiri Danek
Justin Bertram
Matt Slonetsky
Josh Seidel
Clebert Suconic
Howard Gao
Timothy Bish
Francesco Nigro
Michael Andre Pearce
King Ramos
Emmanuel Hugonnet
Ingo Weiss
feuillemorte
Sebastian Thomschke
Andreas Frohwerk