continuous cms gc. even though it is not hitting CMSInitiatingOccupancyFraction=68 Is thsi a know issue

[Peter B. Kessler]

Anuj Lal wrote:

> We just moved to 1.5.0.13 from 1.4.12. It sin production.
> I am opening sun support case too.
>  
> -Xms2460m –server
>   -XX:+UseConcMarkSweepGC
> -XX:+UseParNewGC
> -XX:+CMSParallelRemarkEnabled
> -XX:+UseCMSCompactAtFullCollection
> -XX:CMSFullGCsBeforeCompaction=0
> -XX:+HandlePromotionFailure
> -XX:CMSInitiatingOccupancyFraction=68
>  -XX:+UseCMSInitiatingOccupancyOnly
> -Xmx2460m
> -XX:NewSize=462M
> -XX:MaxNewSize=462M
> -XX:MaxTenuringThreshold=5
> -XX:SurvivorRatio=5
> -XX:TargetSurvivorRatio=80
> -XX:PermSize=256M
> -XX:MaxPermSize=256M
> -XX:-DisableExplicitGC
>  -XX:+CMSPermGenSweepingEnabled
> -XX:+CMSClassUnloadingEnabled
>  -XX:CMSMaxAbortablePrecleanTime=1
>  
>  
> Here I sthe snippet of log.  See used is only used 599703K an jvm keep doing cms gc
>  
> Anuj
>  
> ....


How do people on hotspot-gc-dev at openjdk.java.net feel about our
analyzing GC logs on this alias?  I know the Sun people feel it
would be useful for people interested enough to subscribe to
this alias to see some of the logs we get, and our analysis of
them.  But I thought I'd ask.  (E.g., this case is about 1.5.0,
not OpenJDK, though the analysis might still be interesting.)

What I'd like to do is set up a separate mailing list for OpenJDK
GC log analysis (hotspot-gc-feedback at openjdk.java.net?) and have
hotspot-gc-dev at openjdk.java.net on that list.  That way if the
traffic becomes too heavy, we could separate the lists and only
the interested (or paid :-) people would have to see it.  Does
that seem reasonable?

			... peter