参考文章
JVM监控利器：jstat命令全方位详解
JVM内存监控shell脚本
JVM学习-内存监控工具(五)
jvm 内存监控脚本 jvm常用的监控和调试工具
脚本
#!/bin/bash
# 功能：监测JVM内存配置和GC状态，记录到滚动日志文件
# 使用方式：./jvm_gc_monitor.sh <Java进程PID> [监测间隔(秒，默认10)] [日志目录，默认/var/log/jvm_monitor]
# 数据大小单位：MB,时间单位ms# ===================== 配置项 =====================
MAX_LOG_SIZE="100M" # 单个日志文件最大大小（达到则滚动）
LOG_RETENTION=7 # 保留历史日志天数
# ==================================================# 检查参数
if [ $# -lt 1 ]; thenecho "用法错误！正确用法：$0 <Java_PID> [监测间隔(秒)] [日志目录]"echo "示例：$0 12345 5 /data/logs/jvm"exit 1
fi# 定义核心变量
JAVA_PID=$1
INTERVAL=${2:-10}
LOG_DIR=${3:-/var/log/jvm_monitor}
LOG_FILE="${LOG_DIR}/jvm_gc_monitor.log"
SCRIPT_NAME=$(basename $0)# 检查依赖命令
check_dependency() {local cmd=$1if ! command -v $cmd &> /dev/null; thenecho "错误：未找到 $cmd 命令，请确保JDK已安装并配置环境变量！"exit 1fi
}
check_dependency jstat
check_dependency awk
check_dependency bc
check_dependency logrotate# 检查PID是否为数字且对应Java进程
if ! [[ "$JAVA_PID" =~ ^[0-9]+$ ]]; thenecho "错误：PID必须是数字！"exit 1
fi
if ! ps -p $JAVA_PID -o comm= | grep -i java &> /dev/null; thenecho "错误：PID $JAVA_PID 不是Java进程！"exit 1
fi# 创建日志目录（若不存在）
mkdir -p $LOG_DIR || { echo "错误：无法创建日志目录 $LOG_DIR，请检查权限！"; exit 1; }# 配置logrotate（实现日志滚动）
LOGROTATE_CONF="/etc/logrotate.d/jvm_gc_monitor"
cat > $LOGROTATE_CONF << EOF
$LOG_FILE {size $MAX_LOG_SIZErotate $LOG_RETENTIONcompressmissingoknotifemptycreate 0644 root root
}
EOF
echo "日志滚动配置已写入 $LOGROTATE_CONF"# 写入日志头部（仅首次运行时）
if [ ! -s "$LOG_FILE" ]; thenecho -e "=== JVM GC 监测日志 ===\n监测时间\tPID\tJVM版本" > "$LOG_FILE"# 获取JVM版本信息JVM_VERSION=$(jcmd $JAVA_PID VM.version 2>/dev/null | grep -E "JDK|JRE" | head -1 | awk '{$1=$2=""; print $0}' | sed 's/^ *//g')if [ -z "$JVM_VERSION" ]; thenJVM_VERSION=$(java -version 2>&1 | grep -i version | awk -F '"' '{print $2}')fiecho -e "$(date "+%Y-%m-%d %H:%M:%S")\t$JAVA_PID\t$JVM_VERSION\n" >> "$LOG_FILE"# 写入内存指标表头echo -e "ts,survivor_capacity,survivor_usage,eden_capacity,eden_usage,old_capacity,old_usage,meta_capacity,meta_usage,younggc_times,younggc_elapse,fullgc_times,fullgc_elapse" >> "$LOG_FILE"
fi# 核心监测循环
echo "开始监测Java进程(PID: $JAVA_PID)，日志文件: $LOG_FILE，监测间隔: $INTERVAL 秒（按Ctrl+C停止）"
while true; doCURRENT_TIME=$(date "+%Y-%m-%d %H:%M:%S")# 检查Java进程是否存活if ! ps -p $JAVA_PID > /dev/null 2>&1; thenecho -e "$CURRENT_TIME\t-\t-\t-\t-\t-\t-\t-\t-\t-\t-\t-\t进程已退出" >> "$LOG_FILE"echo "[$CURRENT_TIME] 警告：Java进程(PID: $JAVA_PID)已退出，脚本将在3次重试后退出"RETRY_COUNT=$((RETRY_COUNT + 1))if [ $RETRY_COUNT -ge 3 ]; thenecho "[$CURRENT_TIME] 错误：Java进程已退出，脚本退出"exit 1fisleep $INTERVALcontinuefiRETRY_COUNT=0# ==================== 解析jstat -gc（GC实时状态） ====================# jstat -gc 输出字段说明（单位：KB/次/毫秒）：# S0C S1C S0U S1U EC EU OC OU MC MU CCSC CCSU YGC YGCT FGC FGCT GCTGC_STAT=$(jstat -gc $JAVA_PID 2>/dev/null)# 提取使用量和GC指标S0U_CAPACITY=$(echo "$GC_STAT" | tail -1 | awk '{print $1}') # S0U区capacity(KB)S0U_CAPACITY_MB=$(echo "scale=2; $S0U_CAPACITY / 1024" | bc)S1U_CAPACITY=$(echo "$GC_STAT" | tail -1 | awk '{print $2}') # S1U区capacity(KB)S1U_CAPACITY_MB=$(echo "scale=2; $S1U_CAPACITY / 1024" | bc)S0U_USAGE=$(echo "$GC_STAT" | tail -1 | awk '{print $3}') # S0U区usage(KB)S0U_USAGE_MB=$(echo "scale=2; $S0U_USAGE / 1024" | bc)S1U_USAGE=$(echo "$GC_STAT" | tail -1 | awk '{print $4}') # S1U区capacity(KB)S1U_USAGE_MB=$(echo "scale=2; $S1U_USAGE / 1024" | bc)SURVIOR_CAPACITY_MB=$(echo "scale=2; $S0U_CAPACITY_MB + $S1U_CAPACITY_MB" | bc) # SURVIOR区capacity(KB)SURVIOR_USAGE_MB=$(echo "scale=2; $S0U_USAGE_MB + $S1U_USAGE_MB" | bc) # SURVIOR区usage(KB)EC_CAPACTIY=$(echo "$GC_STAT" | tail -1 | awk '{print $5}') # Eden区capacity(KB)EC_CAPACTIY_MB=$(echo "scale=2; $EC_CAPACTIY / 1024" | bc)EU_USAGE=$(echo "$GC_STAT" | tail -1 | awk '{print $6}') # Eden区usage(KB)EU_USAGE_MB=$(echo "scale=2; $EU_USAGE / 1024" | bc)OC_CAPACTIY=$(echo "$GC_STAT" | tail -1 | awk '{print $7}') # 老年代capacity(KB)OC_CAPACTIY_MB=$(echo "scale=2; $OC_CAPACTIY / 1024" | bc)OU_USAGE=$(echo "$GC_STAT" | tail -1 | awk '{print $8}') # 老年代usage(KB)OU_USAGE_MB=$(echo "scale=2; $OU_USAGE / 1024" | bc)MC_CAPACTIY=$(echo "$GC_STAT" | tail -1 | awk '{print $9}') # 元空间capacity(KB)MC_CAPACTIY_MB=$(echo "scale=2; $MC_CAPACTIY / 1024" | bc)MU_USAGE=$(echo "$GC_STAT" | tail -1 | awk '{print $10}') # 元空间使用(KB)MU_USAGE_MB=$(echo "scale=2; $MU_USAGE / 1024" | bc)YGC_COUNT=$(echo "$GC_STAT" | tail -1 | awk '{print $13}') # YoungGC次数YGCT_TIME=$(echo "$GC_STAT" | tail -1 | awk '{print $14}') # YoungGC耗时(秒)YGCT_MS=$(echo "scale=0; $YGCT_TIME * 1000" | bc) # 转换为毫秒FGC_COUNT=$(echo "$GC_STAT" | tail -1 | awk '{print $15}') # FullGC次数FGCT_TIME=$(echo "$GC_STAT" | tail -1 | awk '{print $16}') # FullGC耗时(秒)FGCT_MS=$(echo "scale=0; $FGCT_TIME * 1000" | bc) # 转换为毫秒# ==================== 3. 写入日志 ====================echo -e "$CURRENT_TIME,$SURVIOR_CAPACITY_MB,$SURVIOR_USAGE_MB,$EC_CAPACTIY_MB,$EU_USAGE_MB,$OC_CAPACTIY_MB,$OU_USAGE_MB,$MC_CAPACTIY_MB,$MU_USAGE_MB,$YGC_COUNT,$YGCT_MS,$FGC_COUNT,$FGCT_MS" >> "$LOG_FILE"# 实时打印监测结果（可选）echo "[$CURRENT_TIME] PID: $JAVA_PID | Eden: ${EU_USAGE_MB}MB/${EC_CAPACTIY_MB}MB | old: ${OU_USAGE_MB}MB/${OC_CAPACTIY_MB}MB | YoungGC: $YGC_COUNT 次($YGCT_MS ms) | FullGC: $FGC_COUNT 次($FGCT_MS ms)"# 触发logrotate检查（可选，确保日志及时滚动）logrotate -f $LOGROTATE_CONF > /dev/null 2>&1sleep $INTERVAL
done