[
https://issues.apache.org/jira/browse/IGNITE-3495?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15385568#comment-15385568
]
Alexey Goncharuk commented on IGNITE-3495:
------------------------------------------
In this case you should have tracked the exact moment when the response message
is written to socket and ACKed from remote server. Get futures give us good
enough get estimate, because if there is something besides the Get* futures
that affects get performance, this should be fixed.
> CacheMetrics.getAverage###Time is not calculated properly
> ---------------------------------------------------------
>
> Key: IGNITE-3495
> URL: https://issues.apache.org/jira/browse/IGNITE-3495
> Project: Ignite
> Issue Type: Bug
> Reporter: Denis Magda
> Assignee: Vladislav Pyatkov
> Attachments: ExampleNodeStartupClient.java, IGNITE_3495.patch,
> example-default.xml
>
>
> {{CacheMetrics.getAverageGetTime}}, {{CacheMetrics.getAveragePutTime}} and
> {{CacheMetrics.getAverageRemoveTime}} are not calculated properly in the
> following scenario:
> - start a server node;
> - start a client node that will perform gets and puts;
> - {{CacheMetrics.getAverage###Time}} will always be 0 for the server node's
> cluster group.
> The issue happens because {{CacheMetricsImpl.add###TimeNanos}} method is not
> called on the server side when a metric related event happens.
> In the attache you can find source that showcases the bug.
> - start basic {{ExampleNodeStartup}} using attached configuration
> {{example-default.xml}} conjuration;
> - start {{ExampleNodeStartupClient}}. You will see that average metrics are
> not incremented.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
