RFR: 8152100 : Rework and unify the GC phase logging (original) (raw)

Stefan Karlsson stefan.karlsson at oracle.com
Thu Mar 17 14:50:28 UTC 2016

Thanks, Stefan.

StefanK

On 2016-03-17 15:09, Stefan Johansson wrote:

Hi Stefan,

On 2016-03-17 14:29, Stefan Karlsson wrote: Hi all,

Please review and comment on this patch to rework and unify the GC phase logging. http://cr.openjdk.java.net/~stefank/8152100/webrev.00/ Looks good, Stefan https://bugs.openjdk.java.net/browse/JDK-8152100 The patch: 1) Adds the 'phases' tag to all GCTraceTimer calls, except those that correspond to the top-level GC measurement that is logged when -Xlog:gc is used. 2) Change the 'phases' levels so that: - Info - gives the first-level GC phase, such as marking, sweeping, etc. - Debug - also gives better breakdown of what we do in the different phases: reference processing, unloading, string table scrubbing, etc. - Trace - also gives non-descriptive phases that we might want to cleanup our restructure. 3) Moved "gc, compaction, phase" to "gc, compaction", so that -Xlog:gc+phases*=trace is usable. 4) Added and removed some phases so that the different GCs log similar phases. Examples of what this looks like in our different GCs can be found below. The logs where produced with -Xlog:gc,start*,gc+phases=. Serial GC: http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.serial.off.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.serial.info.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.serial.debug.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.serial.trace.txt

Parallel GC: http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.parallel.off.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.parallel.info.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.parallel.debug.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.parallel.trace.txt CMS with a Full GC: http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.cms.off.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.cms.info.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.cms.debug.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.cms.trace.txt CMS with a concurrent GC cycle: http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.cms-conc.off.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.cms-conc.info.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.cms-conc.info.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.cms-conc.trace.txt G1 with a Full GC: http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.g1.off.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.g1.info.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.g1.debug.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.g1.trace.txt G1 with a concurrent GC cycle: http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.g1-conc.off.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.g1-conc.info.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.g1-conc.info.txt http://cr.openjdk.java.net/~stefank/8152100/logs.00/gctracetimer.g1-conc.trace.txt Thanks, StefanK

More information about the hotspot-gc-dev mailing list