JDK-6291341: jstat: FGC counter confusing when CMS GC is used

Yasumasa Suenaga yasuenag at gmail.com
Wed Mar 9 12:41:55 UTC 2016


Hi,

> The jdk changes look good to me.

Thanks! Staffan


Jon,

>> Meaning, without the jdk changes, will jstat continue to work
>> as it currently does?

I uploaded new webrev. I updated webrev for hotspot only:
    http://cr.openjdk.java.net/~ysuenaga/JDK-6291341/poc-2/hotspot/

This patch works the same as the current by default.
If you set -XX:+EnableConcGCPerfCounter, CGC counter will work fine.
(I want to set +EnableConcGCPerfCounter by default :-) )


>> Would the changes in the JVM be useful to you without the jstat changes?

We can override jstat_options if we want to watch CGC counters.
Also we can check it through PerfCounter.print jcmd.

I'm developing JVMTI agent [1] which uses sun.gc.collector PerfCounters.
Our agent watches major collection through PerfCounter and JVMTI events.

Thus I'm sure that this change is very useful for us.


Thanks,

Yasumasa


[1] http://icedtea.classpath.org/wiki/HeapStats


On 2016/03/09 16:13, Staffan Larsen wrote:
> The jdk changes look good to me. Jon, if you integrate the JVM changes you can do the jdk changes at the same time in the hs-rt repo.
> 
> Thanks,
> /Staffan
> 
> 
>> On 8 mars 2016, at 23:08, Jon Masamitsu <jon.masamitsu at oracle.com> wrote:
>>
>> Yasumasa,
>>
>> We're discussing this.
>>
>> What would the result be if the GC team made the
>> JVM changes that you want to contribute.  Meaning,
>> without the jdk changes, will jstat continue to work
>> as it currently does?  Would the changes in the JVM
>> be useful to you without the jstat changes?  By the
>> JVM changes, I mean the changes for the counters.
>>
>> The reason I ask is that the GC team does not
>> own jstat.   If I can integrate the JVM changes
>> without the jstat changes, it makes it a little
>> simpler.
>>
>> Jon
>>
>>
>> On 3/8/2016 6:38 AM, Yasumasa Suenaga wrote:
>>> Hi all,
>>>
>>> I discussed about FullGC counter for Concurrent GC:
>>>      http://mail.openjdk.java.net/pipermail/hotspot-gc-dev/2016-March/016774.html
>>>
>>> I found same issue on JBS:
>>>      JDK-6291341: jstat: FGC counter confusing when CMS GC is used
>>>
>>> JDK-6291341 was closed with Cannot Reproduce.
>>> However, I showed cause of this issue in above thread.
>>>
>>> I uploaded patches for this issue:
>>>
>>>      hotspot:http://cr.openjdk.java.net/~ysuenaga/JDK-6291341/poc/hotspot/
>>>          jdk:http://cr.openjdk.java.net/~ysuenaga/JDK-6291341/poc/jdk/
>>>
>>> JDK webrev includes testcase update.
>>> All tests for jstat work fine.
>>>
>>>
>>> Should I work for it as new issue? Or should I reopen JDK-6291341?
>>>
>>> I cannot access JPRT.
>>> So I need a sponsor.
>>>
>>>
>>> Thanks,
>>>
>>> Yasumasa
>>>
>>>
>>
> 
> 


More information about the hotspot-gc-dev mailing list