PrintGCDate/TimeStamps

[Bernd Eckenfels]

Hello,

I wondering about some strange behaviour. Sorry to bother with this minor
observation, but I do hope the datestamp ioption gets more popular by
mention it again .)

Often +PrintGCTimeStamps is recommended to be able to see the intervall
between GC events. However I found some places talking about
+PrintGCDateStamps which is much more convenient for some problems (for
example correlating SLA violations with the STW pause times).

Some discusions suggest that you can combine both in a way that it does not
print both timestamps:

-XX:+PrintGCDateStamps -XX:-PrintGCTimeStamps

However on the win 64bit JDKs 1.6.0_33 and 1.7.0_03 I have tried this, it
does not work (i.e. the logfiles always contans both, actually the
timestamp typically 2 times):

2012-08-21T22:08:29.989+0200: 1.292: [GC 1.292: [ParNew:
5033216K->3826K(5662336K), 0.0015120 secs] 5033216K->3826K(6710912K),
0.0015790 secs] [Times: user=0.00 sys=0.00, real=0.00 secs]
Total time for which application threads were stopped: 0.0018006 seconds
2012-08-21T22:08:30.605+0200: 1.907: [GC 1.907: [ParNew:
5037042K->5230K(5662336K), 0.0023563 secs] 5037042K->5230K(6710912K),
0.0024133 secs] [Times: user=0.00 sys=0.00, real=0.00 secs]
Total time for which application threads were stopped: 0.0026457 seconds

Just some additional information: I was able to tuen off the timestamps in
the HotSpotDiagnostics MBean:

2012-08-21T22:19:54.858+0200: [GC [ParNew: 5039624K->5836K(5662336K),
0.0030131 secs] 5349974K->316537K(6710912K), 0.0030753 secs] [Times:
user=0.00 sys=0.00, real=0.00 secs]

Greetings
Bernd
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/hotspot-gc-use/attachments/20120821/3faaa333/attachment.html