Why Now?

Are you ever surprised by the occurrence of a full collection (aka major collection)? Generally a full collection only happens when an allocation is attempted and there is not enough space available in any of the generations. But there are exceptions to that rule and here are some of them. Actually, these are all of them that I could think of. I thought I would have a longer list so to fill out this blog I've included some bugs that relate to full collections that might interest you. And we're thinking about a policy change for the low pause collector that would affect full collections.

The simplest reason for an unexpected full collection is that you asked for it. System.gc() can be called from most anywhere in the program. Try using -XX:+DisableExplicitGC to have the garbage collector ignore such requests.

The permanent generation is collected only by a full collection. It's very seldom the cause for a full collection but don't overlook the possibility. If you turn on PrintGCDetails, you can see information about the permanent generation collection.

0.167: [Full GC [PSYoungGen: 504K->0K(10368K)] [ParOldGen: 27780K->28136K(41408K)] 28284K->28136K(51776K) [PSPermGen: 1615K->1615K(16384K)], 0.4558222 secs]

This output is from the throughput collector. The "PSPermGen" denotes the permanent generation. The size of the permanent generation currently is 16384K. It's occupancy before the collection is only 1615K and is probably the same after the collection. If the permanent generation needed collecting, the occupancy before the collection would have been closer to the 16384K.

The garbage collector tracks the average of the rate of promotion of live objects from the young generation into the tenured generation. If the average exceeds the free space in the tenured generation, the next time the young generation fills up, a full collection is done instead of a minor collection. There is a pathological situation where, after the full collection, the free space in the tenured generation is still not large enough to take all the live objects expected to be promoted from the next minor collection. This will result in another full collection the next time the young generation fills up. The bad part about this is that the average value for the amount promoted only changes when a minor collection is done and if no minor collections are done, then the average does not change. Increasing the size of the heap usually avoids this problem.

Sometimes it's not really a full collection. There was a bug affecting JDK5 (6432427) which print "Full" when the collection was not a full collection. When JNI critical sections are in use, GC can be locked out. When the JNI critical sections are exited, if a GC had been requested during the lock out, a GC is done. That GC most likely would be a minor collection but was mistakenly labeled a full collection. A second place where "Full" was also printed erroneously was when -XX:+CMSScavengeBeforeRemark is used. The same bug report explains and fixes that error.

An attempt to allocate an object A larger than the young generation can cause a minor collection. That minor collection will fail to free enough space in the young generation to satisfy the allocation of A. A full collection is then attempted and A is then allocated out of the tenured generation. This behavior was a bug (6369448). It has been fixed in JDK's 1.4.2_13, 5.0_10, and 6. The correct behavior is for the GC to recognize when an object is too large to be allocated out of the young generation and to attempt the allocation out of the tenured generation before doing a collection.

When an allocation cannot be satisfied by the tenured generation a full collection is done. That is the correct behavior but that behavior has an adverse affect on CMS (the low pause collector). Early in a run before the CMS (tenured) generation has expanded to a working size, concurrent mode failures (resulting in full collections) happen. These concurrent mode failures perhaps can be avoided. We are thinking about a change to the policy such that CMS expands the generation to satisfy the allocation and then starts a concurrent collection. We're cautious about this approach because similar policies in the past has led to inappropriate expansion of the CMS generation to its maximum size. To avoid these full collections try specifying a larger starting size for the CMS generation.

So just because these are all the unexpected "Full" collections that I could think of, that doesn't mean that these are all there are. If you have a mysterious "Full" collection happening, submit a description to

http://java.sun.com/docs/forms/gc-sendusmail.html

I'll post any additional cases of unexpected "Full" collections that I get.

Late addition (I'm allowed to edit my blog even after they've been posted). If you ask about unexpected "Full" collections and send in a gc log, please add the command line flags

-XX:+PrintGCDetails -XX:+PrintHeapAtGC -XX:+PrintGCTimeStamps

The additional output (of which there will be plenty) may help me see what's happening.

Comments:

Hi,

Thanks: interesting as always...

Does that set of occasions that you list include the "WARNING: bailing out to foreground collection" that I sometimes see when my Web server is starting up and working VERY hard to get ready for the first hits ASAP?

Rgds

Damon

Posted by Damon Hart-Davis on October 26, 2006 at 03:31 AM PDT #

The message about "bailing out" (from the low pause collector /CMS collector) is a concurrent mode failure. A concurrent collection has started but not finished before the tenured generation becomes full. This situration might better be described in my blog on "concurrent mode failures". Two other possibilities are that the permanent generation is too small at start up or that you just need to start with a larger initial heap size. If it's the permanent generation, try a larger size with -XX:PermSize=<nnn>.

Posted by Jon Umasamsit on October 27, 2006 at 12:54 AM PDT #

From http://java.sun.com/docs/hotspot/gc5.0/gc_tuning_5.html#1.1.Other%20Considerations%7Coutline "RMI forces periodic full collection... default rate of once per minute."

Posted by Doug on October 31, 2006 at 05:49 AM PST #

Post a Comment:
Comments are closed for this entry.
About

jonthecollector

Search

Categories
Archives
« April 2014
SunMonTueWedThuFriSat
  
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
   
       
Today