Hi Amit, On Fri, 2017-01-20 at 10:51 +0000, Amit Mishra wrote: > Thank you very much Thomas, after making said changes RSUpdate time > is in control but Scan RS is still high. >
can you summarize your GC parameters again, and also, are these high Scan RS times you observe regular or not? Did you check with -XX:+G1SummarizeRSetStats - XX:G1SummarizeRSetStatsPeriod=<some value> whether there are still coarsenings? Could you also please provide a longer snippet of your log file containing a few gcs before and after this occurrence? Other worthwhile tries could be to fudge with -XX:ParallelGCThreads, either decreasing or increasing them (up to your vcpus) by ten or so to see whether that phase suffers from synchronization issues. The high scan rs times could maybe also a bit mitigated by increasing G1NewSizePercent to maybe two or three to see whether g1 can avoid large young remembered sets by that (Object copy might go up a bit then though). It may be easier to think of causes for the issues you are experiencing by having a little bit of more information. Thanks, Thomas _______________________________________________ hotspot-gc-use mailing list hotspot-gc-use@openjdk.java.net http://mail.openjdk.java.net/mailman/listinfo/hotspot-gc-use
