[
https://issues.apache.org/jira/browse/JAMES-3620?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17388528#comment-17388528
]
tuister commented on JAMES-3620:
--------------------------------
The created timer will be hold by com.codahale.metrics.MetricRegistry#metrics
for each name.
{code:java}
// com.codahale.metrics.MetricRegistry#metrics
private final ConcurrentMap<String, Metric> metrics;
{code}
Because of wrong command, the name will be random(like SMTP-FADLFAJDSF), so the
size of metrics will be increasing continuously, and causing memory leak
finally.
NOTICE: The line below will not remove metric from metrics,
{code:java}
timeMetric.stopAndPublish();
{code}
If you want to reproduct the memory leak, just send a large number of wrong
commands to James smtp server, and observe the size of
com.codahale.metrics.MetricRegistry#metrics, you can get it.
> Memory leak at org.apache.james.protocols.smtp.core.AbstractHookableCmdHandler
> ------------------------------------------------------------------------------
>
> Key: JAMES-3620
> URL: https://issues.apache.org/jira/browse/JAMES-3620
> Project: James Server
> Issue Type: Improvement
> Components: protocols, SMTPServer
> Reporter: tuister
> Priority: Blocker
> Original Estimate: 168h
> Remaining Estimate: 168h
>
> For each command, the command handler will use a timer record the rt. So, if
> someone input error commands, the command handler will create a new timer to
> record, this will cause memory leak.
> {code:java}
> //AbstractHookableCmdHandler#onCommand
> //see also org.apache.james.protocols.smtp.core.UnknownCmdHandler
> TimeMetric timeMetric = metricFactory.timer("SMTP-" +
> request.getCommand().toLowerCase(Locale.US));{code}
>
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
