ARTEMIS-2629 ensure queue auto-delete after expiration #2992
Conversation
jbertram
force-pushed
the
ARTEMIS-2629
branch
from
df50a31
to
e6a5efe
Compare
jbertram
force-pushed
the
ARTEMIS-2629
branch
from
e6a5efe
to
bd0ceb1
Compare
| @@ -84,8 +84,7 @@ public static boolean messageCountCheck(Queue queue) { | |||
| } | |||
|
|
|||
| public static boolean delayCheck(Queue queue) { | |||
| long consumerRemovedTimestamp = queue.getConsumerRemovedTimestamp(); | |||
| return consumerRemovedTimestamp != -1 && System.currentTimeMillis() - consumerRemovedTimestamp >= queue.getAutoDeleteDelay(); | |||
| return System.currentTimeMillis() - queue.getConsumerRemovedTimestamp() >= queue.getAutoDeleteDelay(); | |||
There was a problem hiding this comment.
Removing the condition consumerRemovedTimestamp != -1 could change the semantic of the public method delayCheck. With this change delayCheck will return true before removing a consumer.
There was a problem hiding this comment.
It will change the semantic, and that was intended. This particular check is no longer desirable as it will preclude auto-deleting a queue where all the messages have expired but no consumer has ever connected. I implemented a new check in org.apache.activemq.artemis.core.postoffice.impl.PostOfficeImpl.AddressQueueReaper#queueWasUsed which resolves this issue. The delayCheck method is only used internally as it's not part of the user-facing API so I don't have any qualms with changing its semantics.
jbertram
force-pushed
the
ARTEMIS-2629
branch
2 times, most recently
from
73cb5cb
to
7dc3459
Compare
jbertram
force-pushed
the
ARTEMIS-2629
branch
from
7dc3459
to
13c4fbe
Compare
No description provided.