1

我在 java 8 下运行我的 java 应用程序，我注意到我的 gc 日志文件中有一个奇怪的行为：

    Java HotSpot(TM) 64-Bit Server VM (25.131-b11) for linux-amd64 JRE (1.8.0_131-b11), built on Mar 15 2017 01:23:40 by "java_re" with gcc 4.3.0 20080428 (Red Hat 4.3.0-8)
Memory: 4k page, physical 131855880k(737780k free), swap 8388604k(8192864k free)
CommandLine flags: -XX:InitialHeapSize=1073741824 -XX:MaxHeapSize=21474836480 -XX:MaxMetaspaceSize=1073741824 -XX:MaxNewSize=2147483648 -XX:+PrintGC -XX:+PrintGCApplicationConcurrentTime -XX:+PrintGCApplicationStoppedTime -XX:+PrintGCCause -XX:+PrintGCDateStamps -XX:+PrintGCDetails -XX:+PrintGCTimeStamps -XX:+PrintPromotionFailure -XX:+PrintReferenceGC -XX:+PrintTLAB -XX:+PrintTenuringDistribution -XX:+UseCompressedClassPointers -XX:+UseCompressedOops -XX:+UseParallelGC 
...
2019-04-15T02:15:29.588+0200: 12.457: [Full GC (Metadata GC Threshold) 2019-04-15T02:15:29.599+0200: 12.467: [SoftReference, 0 refs, 0.0000372 secs]2019-04-15T02:15:29.599+0200: 12.467: [WeakReference, 264 refs, 0.0000308 secs]2019-04-15T02:15:29.599+0200: 12.467: [FinalReference, 556 refs, 0.0000449 secs]2019-04-15T02:15:29.599+0200: 12.468: [PhantomReference, 0 refs, 1 refs, 0.0000053 secs]2019-04-15T02:15:29.599+0200: 12.468: [JNI Weak Reference, 0.0000087 secs][PSYoungGen: 98556K->0K(1908224K)] [ParOldGen: 189998K->108966K(699392K)] 288554K->108966K(2607616K), [Metaspace: 20670K->20670K(1069056K)], 0.0491849 secs] [Times: user=0.25 sys=0.04, real=0.05 secs] 
...
2019-04-18T01:45:00.974+0200: 257383.843: [Full GC (Metadata GC Threshold) 2019-04-18T01:45:01.468+0200: 257384.337: [SoftReference, 253 refs, 0.0000833 secs]2019-04-18T01:45:01.468+0200: 257384.337: [WeakReference, 757380 refs, 0.1353332 secs]2019-04-18T01:45:01.604+0200: 257384.472: [FinalReference, 1164 refs, 0.0004652 secs]2019-04-18T01:45:01.604+0200: 257384.472: [PhantomReference, 1 refs, 48 refs, 0.0000144 secs]2019-04-18T01:45:01.604+0200: 257384.472: [JNI Weak Reference, 0.0000551 secs][PSYoungGen: 4573K->0K(333824K)] [ParOldGen: 5596982K->5594855K(8593920K)] 5601555K->5594855K(8927744K), [Metaspace: 799214K->799212K(1769472K)], 1.3237147 secs] [Times: user=6.77 sys=0.01, real=1.33 secs] 
...
2019-04-18T02:10:00.972+0200: 258883.841: [Full GC (Metadata GC Threshold) 2019-04-18T02:10:01.466+0200: 258884.334: [SoftReference, 227 refs, 0.0000816 secs]2019-04-18T02:10:01.466+0200: 258884.334: [WeakReference, 747600 refs, 0.1386251 secs]2019-04-18T02:10:01.604+0200: 258884.473: [FinalReference, 829 refs, 0.0004398 secs]2019-04-18T02:10:01.605+0200: 258884.473: [PhantomReference, 1 refs, 50 refs, 0.0000148 secs]2019-04-18T02:10:01.605+0200: 258884.473: [JNI Weak Reference, 0.0000486 secs][PSYoungGen: 6331K->0K(334848K)] [ParOldGen: 5594855K->5597194K(9195520K)] 5601186K->5597194K(9530368K), [Metaspace: 799223K->799223K(1769472K)], 1.0449664 secs] [Times: user=6.36 sys=0.01, real=1.05 secs]

首先，由于元数据 GC 阈值，我看到很多 Full GC 触发了一个小的延迟，但是日志没有解释为什么会触发这个 full GC，因为元数据大小仍然很小。另外，在我附加的最后一次完整 GC 中，我看到Metaspace: 799223K->799223K(1769472K)，但是1769472K是从哪里来的？它大于设置的 MaxMetaspaceSize 吗？

4

0 回答 0