JVM--调优--04--案例08--元空间不够导致FULL GC

JVM–调优–04–案例08–元空间变大导致FULL GC

---

1、现象

应用经常宕机。

1.1、日志

{Heap before GC invocations=1 (full 1):
def new generation total 1258368K, used 581671K [0x00000005c0000000, 0x0000000615560000, 0x000000066aaa0000)
eden space 1118592K, 52% used [0x00000005c0000000, 0x00000005e3809c70, 0x0000000604460000)
from space 139776K, 0% used [0x0000000604460000, 0x0000000604460000, 0x000000060cce0000)
to space 139776K, 0% used [0x000000060cce0000, 0x000000060cce0000, 0x0000000615560000)
tenured generation total 2796224K, used 15985K [0x000000066aaa0000, 0x0000000715550000, 0x00000007c0000000)
the space 2796224K, 0% used [0x000000066aaa0000, 0x000000066ba3c6e0, 0x000000066ba3c800, 0x0000000715550000)
Metaspace used 34435K, capacity 35373K, committed 35416K, reserved 1079296K
class space used 4370K, capacity 4653K, committed 4696K, reserved 1048576K
9.469: [Full GC (Metadata GC Threshold) TLAB: gc thread: 0x000055fa30914000 [id: 31] desired_size: 22371KB slow allocs: 0 refill waste: 357944B alloc: 1.00000 581671KB refills: 1 waste 99.9% gc: 22895624B slow: 0B fast: 0B
TLAB: gc thread: 0x000055fa30911800 [id: 30] desired_size: 22371KB slow allocs: 0 refill waste: 357944B alloc: 1.00000 581671KB refills: 1 waste 100.0% gc: 22908760B slow: 0B fast: 0B
TLAB: gc thread: 0x000055fa308eb800 [id: 27] desired_size: 22371KB slow allocs: 0 refill waste: 357944B alloc: 1.00000 581671KB refills: 1 waste 99.9% gc: 22890808B slow: 0B fast: 0B
TLAB: gc thread: 0x000055fa308e9800 [id: 26] desired_size: 22371KB slow allocs: 0 refill waste: 357944B alloc: 1.00000 581671KB refills: 1 waste 95.9% gc: 21973168B slow: 0B fast: 0B
TLAB: gc thread: 0x000055fa308bf800 [id: 25] desired_size: 22371KB slow allocs: 0 refill waste: 357944B alloc: 1.00000 581671KB refills: 1 waste 95.9% gc: 21963984B slow: 0B fast: 0B
TLAB: gc thread: 0x000055fa3082f000 [id: 22] desired_size: 22371KB slow allocs: 0 refill waste: 357944B alloc: 1.00000 581671KB refills: 1 waste 94.2% gc: 21572864B slow: 0B fast: 0B
TLAB: gc thread: 0x000055fa30810000 [id: 21] desired_size: 22371KB slow allocs: 0 refill waste: 357944B alloc: 1.00000 581671KB refills: 1 waste 99.6% gc: 22819488B slow: 0B fast: 0B
TLAB: gc thread: 0x000055fa30800800 [id: 20] desired_size: 22371KB slow allocs: 0 refill waste: 357944B alloc: 1.00000 581671KB refills: 1 waste 99.3% gc: 22737328B slow: 0B fast: 0B
TLAB: gc thread: 0x000055fa307ab000 [id: 18] desired_size: 22371KB slow allocs: 0 refill waste: 357944B alloc: 1.00000 581671KB refills: 1 waste 99.9% gc: 22890280B slow: 0B fast: 0B
TLAB: gc thread: 0x000055fa2f127000 [id: 16] desired_size: 22371KB slow allocs: 0 refill waste: 357944B alloc: 1.00000 581671KB refills: 1 waste 100.0% gc: 22908672B slow: 0B fast: 0B
TLAB: gc thread: 0x000055fa2f07c800 [id: 13] desired_size: 22371KB slow allocs: 0 refill waste: 357944B alloc: 1.00000 581671KB refills: 15 waste 3.3% gc: 11358880B slow: 1720B fast: 0B
TLAB totals: thrds: 11 refills: 25 max: 15 slow allocs: 0 max 0 waste: 41.4% gc: 236919856B max: 22908760B slow: 1720B max: 1720B fast: 0B max: 0B
9.470: [Tenured: 15985K->31407K(2796224K), 0.1904343 secs] 597656K->31407K(4054592K), [Metaspace: 34435K->34435K(1079296K)], 0.1906582 secs] [Times: user=0.24 sys=0.01, real=0.19 secs]
Heap after GC invocations=2 (full 2):
def new generation total 1258368K, used 0K [0x00000005c0000000, 0x0000000615560000, 0x000000066aaa0000)
eden space 1118592K, 0% used [0x00000005c0000000, 0x00000005c0000000, 0x0000000604460000)
from space 139776K, 0% used [0x0000000604460000, 0x0000000604460000, 0x000000060cce0000)
to space 139776K, 0% used [0x000000060cce0000, 0x000000060cce0000, 0x0000000615560000)
tenured generation total 2796224K, used 31407K [0x000000066aaa0000, 0x0000000715550000, 0x00000007c0000000)
the space 2796224K, 1% used [0x000000066aaa0000, 0x000000066c94bef8, 0x000000066c94c000, 0x0000000715550000)
Metaspace used 34435K, capacity 35373K, committed 35416K, reserved 1079296K
class space used 4370K, capacity 4653K, committed 4696K, reserved 1048576K
}
9.660: Total time for which application threads were stopped: 0.1913185 seconds, Stopping threads took: 0.0001483 seconds
10.142: Application time: 0.4812853 seconds
10.143: Total time for which application threads were stopped: 0.0012404 seconds, Stopping threads took: 0.0000766 seconds
10.144: Application time: 0.0007587 seconds
10.144: Total time for which application threads were stopped: 0.0007876 seconds, Stopping threads took: 0.0000530 seconds
10.176: Application time: 0.0320494 seconds
10.178: Total time for which application threads were stopped: 0.0015167 seconds, Stopping threads took: 0.0001305 seconds
10.366: Application time: 0.1881530 seconds

1.1、字段说明

9.469: [Full GC (Metadata GC Threshold) TLAB: gc thread: 0x000055fa30914000 [id: 31] 
desired_size: 22371KB 
slow allocs: 0  
refill waste: 357944B 
alloc: 1.00000   581671KB 
refills: 1 waste 99.9% 
gc: 22895624B slow: 0B fast: 0B

1. 9.469: 这是GC事件发生的时间戳，表示自JVM启动以来的秒数。
2. [Full GC (Metadata GC Threshold)]: 这表示发生了一次完全垃圾收集（Full GC），并且触发这次GC的原因是元数据GC阈值（Metadata GC Threshold）被达到。在JVM中，元数据通常指的是类的元数据，比如类的结构信息。当这些元数据占用的空间超过了某个阈值时，就会触发Full GC来清理这些不再使用的元数据。
3. TLAB: 这是线程本地分配缓冲区（Thread-Local Allocation Buffer）的缩写。TLAB是JVM为每个线程分配的私有内存区域，用于加速对象分配。
4. gc thread: 0x000055fa30914000 [id: 31]: 这是执行此次GC的线程的信息，包括其地址和ID。
5. desired_size: 22371KB: 这可能是指TLAB期望的大小，即JVM希望为每个线程分配的TLAB大小。
6. slow allocs: 0: 这表示在TLAB分配过程中，因为TLAB空间不足而需要进行慢速分配的次数为0。
7. refill waste: 357944B: 这表示在重新填充TLAB时浪费的空间大小。
8. alloc: 1.00000: 这可能是表示TLAB的利用率或某种分配率，但具体含义需要查看JVM的文档或源代码来确定。
9. 581671KB: 这可能是与TLAB或此次GC事件相关的某种内存大小。
10. refills: 1: 这表示TLAB被重新填充的次数。
11. waste 99.9%: 这表示在TLAB重新填充或GC过程中浪费的空间比例。
12. gc: 22895624B: 这表示此次GC回收的内存大小。
13. slow: 0B fast: 0B: 这表示慢速和快速分配的内存大小，但在这个日志中它们都是0，可能表示没有发生慢速或快速分配。

1.2、分析

1. 上面可以看到，元空间不够，导致了一次Full GC。
2. TLAB 空间浪费非常高，大于90%，可以设置TLAB的大小
   1. 使用-XX:TLABWasteTargetPercent设置TLAB占用Eden空间的百分比。这个参数允许你调整TLAB的大小，以适应你的应用程序的内存分配模式。默认情况下，TLAB的空间很小，仅占整个Eden空间的1%

1.2、解析

2、解决

1. 调整TLAB大小
2. 增大元数据空间
3. 升级JVM版本