JMC-5596: Rule to detect if there has been a Full GC with G1 or CMS

Marcus Hirt marcus.hirt at oracle.com
Thu Dec 13 18:00:53 UTC 2018


Hi Andrew,

Looks good! Thanks for the contribution!

/M

On 2018-12-13, 18:32, "jmc-dev on behalf of Andrew Azores" <jmc-dev-bounces at openjdk.java.net on behalf of aazores at redhat.com> wrote:

    On 2018-12-12 3:30 p.m., Andrew Azores wrote:
    > Hi Marcus,
    > 
    > On 2018-12-12 7:43 a.m., Marcus Hirt wrote:
    >> Hi Andrew,
    >>
    >> There are, sadly, no official guidelines for adding recordings.
    >>
    >> That said, try to:
    >>
    >> 1. Keep the recording small, e.g. make the recording as short as 
    >> possible,
    >>     use a template where some unrelated events are disabled (take care,
    >>     some events are more or less expected, e.g. Flight Recorder meta 
    >> events).
    >>
    >> 2. Look through the data to ensure it doesn't contain something you 
    >> don't want
    >>     to share. For example, it's all too easy to get a password or 
    >> password hash
    >>     in the environment variable or system property events.
    >>
    >> Kind regards,
    >> Marcus
    >>
    > Thanks for the tips. I have attached another updated patch. This one now 
    > includes two flight recordings, exercising the "Full GCs occurred" paths 
    > for the rule for both G1 and CMS scenarios, as well as an updated test 
    > baseline for the expected result reports for these recordings. I 
    > disabled nearly all event types and ran the recording for as little time 
    > as I was easily able to get the reproducer applet I had prepared to 
    > produce full collections. If the recordings are still too large then I'm 
    > sure I can make my applet hungrier for memory and get the desired events 
    > to occur even sooner, which should cut down on the total number of 
    > events in the recording.
    > 
    
    This got stuck in the moderation queue due to the large size of the 
    patch, so with Mario's help I have uploaded the patch instead as a 
    webrev. Available here:
    
    http://cr.openjdk.java.net/~neugens/JMC-5596/webrev.01/
    
    -- 
    Andrew Azores
    Software Engineer, OpenJDK Team
    Red Hat
    
    




More information about the jmc-dev mailing list