Skip to Main Content
Forums

Java APIs

Announcement

For appeals, questions and feedback about Oracle Forums, please email oracle-forums-moderators_us@oracle.com. Technical questions should be asked in the appropriate category. Thank you!

Full GC performed when not needed

alevineFeb 5 2016 — edited Feb 5 2016

I've been monitoring a stress test being done and need to understand why a Full GC occurs?

The jstat -gccause shows:

  S0     S1     E      O      P     YGC     YGCT    FGC    FGCT     GCT    LGCC                 GCC                

  0.00   0.00  78.12  15.57  99.83    234   16.378   154  163.136  179.514 System.gc()          No GC              

  0.00   0.00  80.65  15.57  99.83    234   16.378   154  163.136  179.514 System.gc()          No GC              

  0.00  33.97   0.00  15.57  99.83    235   16.430   155  163.136  179.566 No GC                System.gc()

As far as I understand, system.gc is a manual request for Full GC? Can this be called by the jvm itself for some reason?

The jstat -gccapacity shows:

NGCMN    NGCMX     NGC     S0C   S1C       EC      OGCMN      OGCMX       OGC         OC      PGCMN    PGCMX     PGC       PC     YGC    FGC

1048576.0 1048576.0 1048576.0 83968.0 80896.0 883712.0  2097152.0  2097152.0  2097152.0  2097152.0  21504.0 524288.0 243200.0 243200.0    266   181

1048576.0 1048576.0 1048576.0 77824.0 80896.0 883712.0  2097152.0  2097152.0  2097152.0  2097152.0  21504.0 524288.0 243200.0 243200.0    267   182

1048576.0 1048576.0 1048576.0 77824.0 80896.0 883712.0  2097152.0  2097152.0  2097152.0  2097152.0  21504.0 524288.0 243200.0 243200.0    267   182

The jstat -gcutil shows:

S0     S1     E      O      P     YGC     YGCT    FGC    FGCT     GCT  

  0.00   0.00  46.76  15.26  99.89    270   18.568   185  195.039  213.607

  0.00  36.94   0.00  15.26  99.89    271   18.620   186  195.039  213.659

  0.00   0.00   7.90  15.14  99.89    271   18.620   186  196.042  214.663

  0.00   0.00  46.57  15.14  99.89    271   18.620   186  196.042  214.663

  0.00   0.00  65.36  15.14  99.89    271   18.620   186  196.042  214.663

  0.00   0.00  70.43  15.14  99.89    271   18.620   186  196.042  214.663

  0.00   0.00  73.01  15.14  99.89    271   18.620   186  196.042  214.663

40.18   0.00   0.00  15.14  99.89    272   18.676   187  196.042  214.718

  0.00   0.00   3.20  15.17  99.89    272   18.676   187  196.998  215.674

And verbosegc shows:

[GC [PSYoungGen: 723691K->39704K(962048K)] 1041343K->357355K(3059200K), 0.0484510 secs] [Times: user=0.09 sys=0.00, real=0.05 secs]

[Full GC [PSYoungGen: 39704K->0K(962048K)] [ParOldGen: 317651K->327322K(2097152K)] 357355K->327322K(3059200K) [PSPermGen: 243053K->243053K(243200K)], 1.0875260 secs] [Times: user=1.58 sys=0.01, real=1.09 secs]

[GC [PSYoungGen: 623020K->38985K(957952K)] 950343K->366308K(3055104K), 0.0537650 secs] [Times: user=0.08 sys=0.00, real=0.05 secs]

[Full GC [PSYoungGen: 38985K->0K(957952K)] [ParOldGen: 327322K->326737K(2097152K)] 366308K->326737K(3055104K) [PSPermGen: 243053K->243053K(243200K)], 0.9313790 secs] [Times: user=1.57 sys=0.00, real=0.93 secs]

[GC [PSYoungGen: 684235K->29918K(966656K)] 1010972K->356655K(3063808K), 0.0459150 secs] [Times: user=0.08 sys=0.01, real=0.05 secs]

[Full GC [PSYoungGen: 29918K->0K(966656K)] [ParOldGen: 326737K->317554K(2097152K)] 356655K->317554K(3063808K) [PSPermGen: 243054K->243054K(243200K)], 0.9135880 secs] [Times: user=1.55 sys=0.00, real=0.91 secs]

Why would a Full GC be performed to clean up PSYoungGen, using System.gc() calls?

Any help would be greatly appreciated.

Comments

Frank Kulash

Hi, User_97XVQ
If you want the sub-string of str from the start of the string up to (but not including) the last substring of hyphen-digits-hyphen, then you can use:

REGEXP_SUBSTR ( str
	      , '(.*)-\d+-'
	      , 1
	      , 1
	      , NULL
	      , 1
	      )

If you'd care to post CREATE TABLE and INSERT statements for the sample data, then I could test it.
If str does not contain a sub-0string of all digits surrounded by hyphens, then the expression above returns NULL.

Frank Kulash

Hi, User_97XVQ
I see that you changed your original message after I replied. Please don't do that: it makes the thread hard to read and your changes easy to miss. Make any corrections and additions in a new reply, at the end of the thread.
If tst is a string containing one or more digits, immediately preceded and followed by a hyphen, then

REGEXP_SUBSTR ( tst
	      , '(.*?)-\d+-'
	      , 1
	      , 1
	      , NULL
	      , 1
	      )

returns the sub-string before the first occurrence of the hyphen-digits-hyphen pattern. The only change for your new requirements is the '?' in the second argument, making '.*' non-greedy (that is, matching as little as possible when there is a choice).
As before, the expression returns NULL if tst does not contain a hyphen-digits-hyphen pattern.

1 - 2
Locked Post
New comments cannot be posted to this locked post.

Post Details

Locked on Mar 4 2016
Added on Feb 5 2016
#jdk-s-monitoring-troubleshooting-tools
0 comments
694 views