Fix KahaDB index free page recovery on unclean shutdown so that existing
free pages will be tracked and not lost.
(cherry picked from commit 38d85be476)
Remove incorrect dispatch counter increment on message expiration in
TopicSubscription when a pending limit strategy is set
(cherry picked from commit d0c95146c3)
Queue purge now acquires the sendLock to prevent new messages from
coming in while purging. The statistics are no longer zeroed out as
they should properly decrement as messages are removed. These changes
should prevent the statistics from going negative.
(cherry picked from commit 56bb079c82)
On journal start use lastFileLength instead of maxFileLength when
checking for unused space in the last journal file as it is possible the
length of the last journal file is not the same length as maxFileLength
(cherry picked from commit 0ad62f722f)
Stop connections on all connection attempt errors, not just security
exceptions, to prevent dangling open sockets.
(cherry picked from commit d9c74d7317)
Fixing Scheduler so that a rescheduled task will first cancel the
existing task so it does not get orphaned from the task map. Also
fixing Topic start so that it will only start once and not twice.
(cherry picked from commit 2769298cf6)
Add support for the receiver side of a sender link which carries a
desired capabilities array which can request to know if we support
delayed delivery, answer with an offered capability in that case.
(cherry picked from commit 4a1c05b628)
Fixing an issue with syncDurableSubs that cause a bridge failure when
adding multiple bridges between the same brokers
(cherry picked from commit 25703fbd1f)
Fixing the auto+nio+ssl transport so that the protocol detection task
will properly terminate on timeout and not continue to run. Also
lowered the default detection timeout to 15 seconds instead of 30
seconds to match the InactivityMonitor default.
(cherry picked from commit 27238b2dd7)
Fixing a thread safety issue with memoryUsage when using concurrentStoreAndDispatch
that was causing memory usage to get out of sync.
The InnerFutureTask class inside KahaDB was not thread safe which was
the root cause of the problem.
(cherry picked from commit 0f0bdb21ef)
Adding support for Jetty 9.3 by re-adding in the logic to dynamically
load the correct GzipHandler depending on the version
(cherry picked from commit 80f46a8056)
Changing the WebSocket servlet to return a succesful empty response
instead of an error when a GET request is made
(cherry picked from commit f40532a23d)
Better handle prefetch extension and pull consumers over Topics to avoid
the remote not receiving all the messages available based on the credit
it has issued.
(cherry picked from commit e050519ff6)
Fixing the auto transport protocol detection so that the byte buffer
that captures the initial bytes for detection is not shared across
threads. This was causing failed connections under high load and high cpu
usage under NIO
(cherry picked from commit 7e648d512d)
Adding a timeout for websocket sends to prevent the transport thread
from getting stuck and blocking. The default is 30 seconds.
(cherry picked from commit 450cabe4ea)
Include the already received backlog when deciding to grant additional
credit to avoid excessive backlogs of messages during producer flow
control.
(cherry picked from commit 7cf7fba7aa)
Christopher L. Shannon (cshannon)
Timothy Bish
Krzysztof Sobkowiak
Christian Schneider
Daniel Kulp
gtully