I have been able to successfully have large number of persistent messages
spool over to the filesystem exactly as intended by using the
cursorMemoryHighWaterMark for messages that have been enqueued to the queue.
The issue I am running into, it happens only a few times per day in our
system, is that a large transaction is created that tries to enqueue a
couple hundred thousand messages and during the transaction none of the
rules in place seem to cause the messages to get spooled over to disk. If
the number of messages is large enough the broker can still run out of
memory.
For this scenario I am not too worried about speed as much as I am about
keeping the broker alive to receive and store the messages.
I am using the file based cursor as I had noticed during some larger
transactions that was causing the queue to hit the cursorMemoryHighWaterMark
messages that had been committed to the queue were getting blocked behind
the larger transaction which does not seem to occur with the file based
cursor.
Any thoughts on how to handle this scenario would be greatly appreciated.
The broker is started with a gig of heap space. -Xmx1G
<destinationPolicy>
<policyMap>
<policyEntries>
<policyEntry topic=">"
producerFlowControl="true">
<pendingMessageLimitStrategy>
<constantPendingMessageLimitStrategy
limit="1000"
/>
</pendingMessageLimitStrategy>
</policyEntry>
<policyEntry queue=">"
producerFlowControl="false" cursorMemoryHighWaterMark="10">
<pendingQueuePolicy><fileQueueCursor/></pendingQueuePolicy>
</policyEntry>
</policyEntries>
</policyMap>
</destinationPolicy>
<systemUsage>
<systemUsage>
<memoryUsage>
<memoryUsage limit="64 mb" />
</memoryUsage>
<storeUsage>
<storeUsage limit="30 gb" />
</storeUsage>
<tempUsage>
<tempUsage limit="2 gb" />
</tempUsage>
</systemUsage>
</systemUsage>
--
Sent from: http://activemq.2283324.n4.nabble.com/ActiveMQ-User-f2341805.html
