Java GC（分配失败）

Question

Why always "GC (Allocation Failure)"?

Java HotSpot(TM) 64-Bit Server VM (25.25-b02) for linux-amd64 JRE ( 1.8.0_25 -b17),

CommandLine flags: 
-XX:CMSInitiatingOccupancyFraction=60 
-XX:GCLogFileSize=10485760 
-XX:+HeapDumpOnOutOfMemoryError 
-XX:InitialHeapSize=32212254720 
-XX:MaxHeapSize=32212254720 
-XX:NewRatio=10 
-XX:OldPLABSize=16 
-XX:ParallelGCThreads=4 
-XX:+PrintGC 
-XX:+PrintGCDetails 
-XX:+PrintGCTimeStamps 
-XX:+PrintStringTableStatistics 
-XX:+PrintTenuringDistribution 
-XX:StringTableSize=1000003 
-XX:SurvivorRatio=4 
-XX:TargetSurvivorRatio=50 
-XX:+UseCompressedClassPointers 
-XX:+UseCompressedOops
-XX:+UseParNewGC 
-XX:+UseConcMarkSweepGC

27.329: [GC (Allocation Failure) 27.329: [ParNew
Desired survivor size 44728320 bytes, new threshold 15 (max 15)
- age   1:   16885304 bytes,   16885304 total
: 349568K->16618K(436928K), 0.2069129 secs] 349568K->16618K(31369920K), 0.2070712 secs] [Times: user=0.78 sys=0.04, real=0.21 secs]


28.210: [GC (Allocation Failure) 28.210: [ParNew
Desired survivor size 44728320 bytes, new threshold 15 (max 15)
- age   1:   28866504 bytes,   28866504 total
- age   2:   12582536 bytes,   41449040 total
: 366186K->47987K(436928K), 0.2144807 secs] 366186K->47987K(31369920K), 0.2146024 secs] [Times: user=0.84 sys=0.01, real=0.22 secs]


29.037: [GC (Allocation Failure) 29.038: [ParNew
Desired survivor size 44728320 bytes, new threshold 2 (max 15)
- age   1:   28443488 bytes,   28443488 total
- age   2:   28386624 bytes,   56830112 total
- age   3:   12579928 bytes,   69410040 total
: 397555K->76018K(436928K), 0.2357352 secs] 397555K->76018K(31369920K), 0.2358535 secs] [Times: user=0.93 sys=0.01, real=0.23 secs]

Answer 1

"Allocation Failure" is a cause of GC cycle to kick in.

"Allocation Failure" means that no more space left in Eden to allocate object. So, it is normal cause of young GC.

Older JVM were not printing GC cause for minor GC cycles.

"Allocation Failure" is almost only possible cause for minor GC. Another reason for minor GC to kick could be CMS remark phase (if +XX:+ScavengeBeforeRemark is enabled).

Answer 2

"Allocation Failure" is cause of GC to kick is not correct. It is an outcome of GC operation.

GC kicks in when there is no space to allocate( depending on region minor or major GC is performed). Once GC is performed if space is freed good enough, but if there is not enough size it fails. Allocation Failure is one such failure. Below document have good explanation https://docs.oracle.com/javase/8/docs/technotes/guides/vm/gctuning/g1_gc.html

Answer 3

在 jdk1.8 中使用 CMS GC 时会出现这个错误，我换了 G1 Gc 解决了这个问题。

 -Xss512k -Xms6g -Xmx6g -XX:+UseG1GC -XX:MaxGCPauseMillis=200 -XX:InitiatingHeapOccupancyPercent=70 -XX:NewRatio=1 -XX:SurvivorRatio=6 -XX:G1ReservePercent=10 -XX:G1HeapRegionSize=32m -XX:ConcGCThreads=6 -Xloggc:gc.log -XX:+HeapDumpOnOutOfMemoryError -XX:+PrintGC -XX:+PrintGCDetails -XX:+PrintGCTimeStamps