[
https://issues.apache.org/jira/browse/AMQ-5338?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14234645#comment-14234645
]
Tim Bain commented on AMQ-5338:
-------------------------------
What GC strategy are you using? And which JVM implementation? (Hotspot,
OpenJDK, IBM, etc.) And are you sure that your "full GCs" were actually *full*
GCs rather than just GCs that collected from Old Gen?
Using G1GC in Hotspot 7u21, I've seen behavior like what you described (with an
upward-sloping sawtooth pattern for Old Gen usage), but once a real full GC
happened, Old Gen dropped to nearly nothing and then started the pattern again,
which makes me think that what I saw was a poor implementation of the G1GC
algorithm rather than anything wrong with ActiveMQ. Did you actually let
things run until you hit an OutOfMemoryError? If not, you're just speculating
that a memory leak actually existed, because the GC strategy is under no
obligation to collect everything that can be GCed at any point in time.
> Run into Memory Leak problem after upgrade from 5.5.1 to 5.10
> -------------------------------------------------------------
>
> Key: AMQ-5338
> URL: https://issues.apache.org/jira/browse/AMQ-5338
> Project: ActiveMQ
> Issue Type: Bug
> Affects Versions: 5.9.0, 5.10.0
> Environment: Windows 2012 R2, JDK 1.7.40, CPU I5, Memory 8G
> Reporter: Edwin Wang
>
> Our production environment has used ActiveMQ 5.5.1 for a long time and it
> works well. Recently we decided to verify ActiveMQ 5.10 for prepared for
> upgrade. However, after we upgrade the test environment to 5.10, the MQ is
> easy ran into out of memory every 7-10 days.
> In order to compare the two, I use the same environment and same
> configuration for both 5.5.1 and 5.10.0. Finally, 5.10.0 ran into out of
> memory while 5.5.1 works well.
> It looks like a bug for 5.10.
> Below is the main part of the configration
> <destinationPolicy>
> <policyMap>
> <policyEntries>
> <policyEntry topic=">" advisoryForConsumed="false"
> advisoryForDelivery="false" gcInactiveDestinations="true"
> inactiveTimoutBeforeGC="600000" producerFlowControl="true" memoryLimit="3mb">
> <pendingSubscriberPolicy>
> <vmCursor />
> </pendingSubscriberPolicy>
> </policyEntry>
> <policyEntry queue=">" advisoryForConsumed="false"
> advisoryForDelivery="false" gcInactiveDestinations="true"
> inactiveTimoutBeforeGC="600000" producerFlowControl="true" memoryLimit="3mb">
> </policyEntry>
> </policyEntries>
> </policyMap>
> </destinationPolicy>
> <managementContext>
> <managementContext createConnector="true"/>
> </managementContext>
> <plugins>
> <simpleAuthenticationPlugin>
> <users>
> <authenticationUser
> username="msgmanager" password="Br0c@d3s0ft" groups="admins"/>
> </users>
> </simpleAuthenticationPlugin>
> <authorizationPlugin>
> <map>
> <authorizationMap>
> <authorizationEntries>
> <authorizationEntry queue=">"
> read="admins" write="admins" admin="admins" />
> <authorizationEntry topic=">"
> read="admins" write="admins" admin="admins" />
> </authorizationEntries>
> </authorizationMap>
> </map>
> </authorizationPlugin>
> </plugins>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
