[
https://issues.apache.org/jira/browse/CAMEL-18844?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17652260#comment-17652260
]
Rafał Gała commented on CAMEL-18844:
------------------------------------
Shouldn't this be synchronized?
{code:java}
if (exchangeEvents.size() >= capacity) {
exchangeEvents.poll();
} {code}
I maye be wrong but it looks like this poll call was made to prevent the queue
size from exceeding value of initial capacity (default 25). Because it's not
synchronized, the add method is called more often then the poll one.
> Possible memory leak in org.apache.camel.impl.console.EventConsole
> ------------------------------------------------------------------
>
> Key: CAMEL-18844
> URL: https://issues.apache.org/jira/browse/CAMEL-18844
> Project: Camel
> Issue Type: Bug
> Components: camel-core
> Affects Versions: 3.20.0
> Reporter: Rafał Gała
> Priority: Critical
> Fix For: 3.21.0, 3.20.1
>
>
> Today I have upgraded from 3.18.4 to 3.20.0 and did tests under heavy load as
> usual. After processing of approximately 3 millions of exchanges JVM's heap
> usage reached 4GB and it ran out of memory. I analyzed a dump in Eclipse MAT
> and it showed the following:
> {noformat}
> One instance of org.apache.camel.spring.boot.SpringBootCamelContext loaded by
> org.apache.catalina.loader.ParallelWebappClassLoader @ 0x700b5adb0 occupies 3
> 276 629 016 (97,38%) bytes. The memory is accumulated in one instance of
> java.lang.Object[], loaded by <system class loader>, which occupies 3 276 529
> 992 (97,38%)
> bytes.Keywordsorg.apache.camel.spring.boot.SpringBootCamelContextorg.apache.catalina.loader.ParallelWebappClassLoader
> @ 0x700b5adb0java.lang.Object[]{noformat}
> I discovered that the *java.util.ArrayDeque* instance in
> *org.apache.camel.impl.console.EventConsole* instance (most probably the
> exchangeEvents one) contains references to over 12 millions of various
> org.apache.camel.impl.event.* objects, like {*}ExchangeCreatedEvent{*},
> *ExchangeSentEvent* etc.
> I will investigate this further, but it looks like the poll method on
> ArrayDeque does not do something as expected or is used in a wrong way:)
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
