JVM内存参数的学习之三

JVM内存参数的学习之三

---

背景

研究启动性能时, 顺便看到了jmap -heap 1 的部分信息
看到:
MinHeapFreeRatio、MaxHeapFreeRatio
自己突然以为是 Percentage的参数, 恍惚了好久.
才发现自己对内存的学习不够, 所以想多学习一下.

---

参数解释

MinHeapFreeRatio:
空闲堆空间的最小百分比，计算公式为：HeapFreeRatio =(CurrentFreeHeapSize/CurrentTotalHeapSize) * 100，
值的区间为0到100，默认值为 40。如果HeapFreeRatio < MinHeapFreeRatio，则需要进行堆扩容，扩容的时机应该在每次垃圾回收之后。

MaxHeapFreeRatio :
空闲堆空间的最大百分比，计算公式为：HeapFreeRatio =(CurrentFreeHeapSize/CurrentTotalHeapSize) * 100，
值的区间为0到100，默认值为 70。如果HeapFreeRatio > MaxHeapFreeRatio，则需要进行堆缩容，缩容的时机应该在每次垃圾回收之后。

来源:
https://blog.csdn.net/claram/article/details/104635114

---

参数默认值

如果启动参数为:
"-XX:MaxRAMPercentage=80.0","-XX:MinRAMPercentage=50.0"
参数值应该是默认的:
Heap Configuration:
   MinHeapFreeRatio         = 40
   MaxHeapFreeRatio         = 70
   MaxHeapSize              = 25891438592 (24692.0MB)
   NewSize                  = 169148416 (161.3125MB)
   MaxNewSize               = 8630435840 (8230.625MB)
   OldSize                  = 338362368 (322.6875MB)

如果关闭自适应的话:
"-XX:-UseAdaptiveSizePolicy","-XX:MaxRAMPercentage=80.0","-XX:MinRAMPercentage=50.0"
发现这个参数值并不会发生变化
   Heap Configuration:
   MinHeapFreeRatio         = 40
   MaxHeapFreeRatio         = 70
   MaxHeapSize              = 25891438592 (24692.0MB)
   NewSize                  = 169148416 (161.3125MB)
   MaxNewSize               = 8630435840 (8230.625MB)
   OldSize                  = 338362368 (322.6875MB)

增加初始化的设置时
"-XX:InitialRAMPercentage=75.0","-XX:MaxRAMPercentage=75.0","-XX:MinRAMPercentage=75.0"
Heap Configuration:
   MinHeapFreeRatio         = 0
   MaxHeapFreeRatio         = 100
   MaxHeapSize              = 22548578304 (21504.0MB)
   NewSize                  = 7516192768 (7168.0MB)
   MaxNewSize               = 7516192768 (7168.0MB)
   OldSize                  = 15032385536 (14336.0MB)

感觉增加初始化 InitialRAMPercentage 非常管用,会自动分配堆区设置, 并且能够将old区域 固定下来. 

如果不设置 InitialRAMPercentage时
"-XX:MaxRAMPercentage=75.0","-XX:MinRAMPercentage=75.0"
Heap Configuration:
   MinHeapFreeRatio         = 0
   MaxHeapFreeRatio         = 100
   MaxHeapSize              = 22548578304 (21504.0MB)
   NewSize                  = 234881024 (224.0MB)
   MaxNewSize               = 11274289152 (10752.0MB)
   OldSize                  = 234881024 (224.0MB)
但是这个时候内存的分配不是严格的 1:2配比的new和old   old区域一开始很小.  感觉这样并不是很优秀. 

可以增加几个内存参数对象的变量设置.
"-XX:InitialRAMPercentage=75.0","-XX:MaxRAMPercentage=75.0","-XX:MinRAMPercentage=75.0","-XX:SurvivorRatio=3","-XX:NewRatio=1"
   MinHeapFreeRatio         = 0
   MaxHeapFreeRatio         = 100
   MaxHeapSize              = 22548578304 (21504.0MB)
   NewSize                  = 11274289152 (10752.0MB)
   MaxNewSize               = 11274289152 (10752.0MB)
   OldSize                  = 11274289152 (10752.0MB)

---

参数的理解

JVM的参数其实很复杂.
建议还是 最大最小区域保持一致, 这样的话就不会因为 freeMemory的原因
导致堆区被收缩和扩张因为性能衰退.
毕竟操作系统进行内存的收缩和扩展是要耗费资源和时间的.

---

不改容器内启动脚本实现不同参数的方法

apiVersion: apps/v1
kind: Deployment
metadata:
  labels:
    app: myapp
  name: myapp-deployment
  namespace: default
spec:
  replicas: 1
  selector:
    matchLabels:
      app: myapp
  template:
    metadata:
      labels:
        app: myapp
    spec:
      containers:
      - image: myapp:v1.0
        ports:
        - containerPort: 5200
        name: myapp
        command: ["java","-Djava.security.egd=file:/dev/./urandom","-Dloader.path=./server/runtime/3rd,./server/runtime/libs","-Dparallel.startup=true","-Dserver.runtime.path.name=server","-XX:+UseContainerSupport","-XX:InitialRAMPercentage=75.0","-XX:MaxRAMPercentage=75.0","-XX:MinRAMPercentage=75.0","-XX:SurvivorRatio=3","-XX:NewRatio=1","-XX:-DisableExplicitGC","-XX:+PrintFlagsFinal","-XX:+PrintGC","-XX:+PrintGCDateStamps","-XX:+PrintGCDetails","-XX:+PrintGCTimeStamps","-Xloggc:/app/dump/gclog","-XX:+HeapDumpOnOutOfMemoryError","-XX:HeapDumpPath=/app/dump","-jar","./server/runtime/caf-bootstrap.jar","--spring.config.location=./server/runtime/"]
        resources:
          limits:
            cpu: 6
            memory: 28Gi
          requests:
            cpu: 6
            memory: 28Gi
---
apiVersion: v1
kind: Service
metadata:
  labels:
   app: myapp
  name: myapp-service
  namespace: default
spec:
  sessionAffinity: ClientIP
  ports:
  - port: 80
    name: myapp-service
    protocol: TCP
    targetPort: 5200
  selector:
    app: myapp
  type: NodePort