[jboss-jira] [JBoss JIRA] (WFLY-5835) Mean response times in soak test exceed 1000 milliseconds

[Paul Ferraro (JIRA)]

     [ https://issues.jboss.org/browse/WFLY-5835?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Paul Ferraro closed WFLY-5835.
------------------------------
    Resolution: Rejected


This is expected behavior for ASYNC cache modes.
In AS7.x, cache reads were lock free, thus would never block, but be prone to dirty reads (i.e. stale data on failover).
In WF10, cache reads require locks, thus they will block until the previous replication finishes, and dirty reads are prevented.

> Mean response times in soak test exceed 1000 milliseconds
> ---------------------------------------------------------
>
>                 Key: WFLY-5835
>                 URL: https://issues.jboss.org/browse/WFLY-5835
>             Project: WildFly
>          Issue Type: Bug
>          Components: Clustering
>    Affects Versions: 10.0.0.CR4
>            Reporter: Michal Vinkler
>            Assignee: Paul Ferraro
>
> The responses in soak test reach really high values (hundreds of ms) compared to EAP 6 (max 10 ms). Also number of unhealthy samples (samples which arrive later than in 3 seconds) is really high in EAP7 run compared to 0 unhealthy samples in EAP6 run. 
> EAP7 run also logs few sampling errors in almost each iteration, EAP6 did not. 
> Compare these results:
> EAP7: [performance.txt|http://jenkins.mw.lab.eng.bos.redhat.com/hudson/job/eap-7x-clustering-soak-http-repl-async-web/lastSuccessfulBuild/artifact/report/performance.txt]
> (example:
> response min: 0 ms, mean: 965 ms, max: 16034 ms, sampling errors: 5, unhealthy samples: 86209)
> EAP6: [performance.txt|http://jenkins.mw.lab.eng.bos.redhat.com/hudson/job/eap-6x-clustering-soak-http-repl-async-web/73/artifact/report/performance.txt]
> (example:
> response min: 0 ms, mean: 6 ms, max: 1053 ms, sampling errors: 0, unhealthy samples: 0)



--
This message was sent by Atlassian JIRA
(v6.4.11#64026)