[jboss-jira] [JBoss JIRA] (WFLY-12167) Memory leak in metrics in standalone-ha configuration

[Brian Stansberry (Jira)]

    [ https://issues.jboss.org/browse/WFLY-12167?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13744298#comment-13744298 ] 

Brian Stansberry commented on WFLY-12167:
-----------------------------------------

[~pferraro] [~jmesnil] During a metrics scan do the same underlying value instances get read for multiple attributes over the course of a single management op? If so can the relevant OSHs cache those in the OperationContext via an Attachment? For example ChannelMetricExecutor could use an attachment of type Map<String, JChannel>.

A weakness is while that read is going on another op could come in and result in the service being stopped, leaving the read op with a ref to a defunct value. That potential exists with the current code too, but caching the value across steps would increase the size of the window wherein that could happen.

Note that this comment is getting a bit OT from the original bug report, which is resolved. This is about further optimization.

> Memory leak in metrics in standalone-ha configuration
> -----------------------------------------------------
>
>                 Key: WFLY-12167
>                 URL: https://issues.jboss.org/browse/WFLY-12167
>             Project: WildFly
>          Issue Type: Bug
>          Components: Clustering, MP Metrics, MSC
>    Affects Versions: 16.0.0.Final
>            Reporter: Bernd Stolle
>            Assignee: Richard Opalka
>            Priority: Blocker
>              Labels: memoryleak
>             Fix For: 17.0.0.Final
>
>         Attachments: Screenshot 2019-06-06 at 11.07.00.png
>
>
> When started in standalone HA configuration every request to the recently added metrics endpoint ({{<management-if>:9990/metrics}}) lead to an increase in memory consumption until the JVM is slowed down significantly by GC to a point where even the requests to {{/health}} fail within a reasonable timeout (2s) and untlimately lead to OOM.
> The same issue does not occur when WildFly is started in the default standalone configuration (non HA).
> I can provide a (compressed) heap dump if required.



--
This message was sent by Atlassian Jira
(v7.12.1#712002)