Tomcat（JVM）性能监控方法

Tomcat（JVM）监控方法

1、Tomcat自带的监控页面

配置详见Tomcat安装配置监控一文，如图所示为监控页面：

double atof (const char *string);
extern char* strtok(char *token, const char *delimiter);

CollectMetrics()
{
int countP, countS, countF, countR, countK;
int numValues;
static int loggedVersionInfo = FALSE;

lr_save_string("127.0.0.1:8080", "ServerName");
web_set_max_html_param_len("65536");

web_set_user("admin",
"admin",
"{ServerName}");

lr_start_transaction("monitor tomcat");

web_reg_save_param("JVMFreeMemory",
"LB=Free memory: ",
"RB= MB",
"Ord=1",
LAST);

web_reg_save_param("JVMTotalMemory",
"LB=Total memory: ",
"RB= MB",
"Ord=1",
LAST);

web_reg_save_param("JVMMaxMemory",
"LB=Max memory: ",
"RB= MB",
"Ord=1",
LAST);

web_reg_save_param("HTTPMaxThreads",
"LB=Max threads: ",
"RB= ",
"Ord=1",
LAST);

web_reg_save_param("HTTPMinSpareThreads",
"LB=Min spare threads: ",
"RB= ",
"Ord=1",
LAST);

web_reg_save_param("HTTPMaxSpareThreads",
"LB=Max spare threads: ",
"RB= ",
"Ord=1",
LAST);

web_reg_save_param("HTTPCurrentSpareThreads",
"LB=Current thread count: ",
"RB= ",
"Ord=1",
LAST);

web_reg_save_param("HTTPCurrentThreadBusy",
"LB=Current thread busy: ",
"RB= ",
"Ord=1",
LAST);

web_reg_save_param("HTTPMaxProcessingTime",
"LB=Max processing time: ",
"RB= ",
"Ord=1",
LAST);

web_reg_save_param("HTTPRequestCount",
"LB=Request count: ",
"RB= ",
"Ord=1",
LAST);

web_reg_save_param("HTTPErrorCount",
"LB=Error count: ",
"RB= ",
"Ord=1",
LAST);

web_reg_save_param("HTTPBytesReceived",
"LB=Bytes received: ",
"RB= ",
"Ord=1",
LAST);

web_reg_save_param("HTTPBytesSent",
"LB=Bytes sent: ",
"RB= ",
"Ord=1",
LAST);

web_reg_save_param("JKMaxThreads",
"LB=Max threads: ",
"RB= ",
"Ord=2",
LAST);

web_reg_save_param("JKMinSpareThreads",
"LB=Min spare threads: ",
"RB= ",
"Ord=2",
LAST);

web_reg_save_param("JKMaxSpareThreads",
"LB=Max spare threads: ",
"RB= ",
"Ord=2",
LAST);

web_reg_save_param("JKCurrentSpareThreads",
"LB=Current thread count: ",
"RB= ",
"Ord=2",
LAST);

web_reg_save_param("JKCurrentThreadBusy",
"LB=Current thread busy: ",
"RB= ",
"Ord=2",
LAST);

web_reg_save_param("JKMaxProcessingTime",
"LB=Max processing time: ",
"RB= ",
"Ord=2",
LAST);

web_reg_save_param("JKRequestCount",
"LB=Request count: ",
"RB= ",
"Ord=2",
LAST);

web_reg_save_param("JKErrorCount",
"LB=Error count: ",
"RB= ",
"Ord=2",
LAST);

web_reg_save_param("JKBytesReceived",
"LB=Bytes received: ",
"RB= ",
"Ord=2",
LAST);

web_reg_save_param("JKBytesSent",
"LB=Bytes sent: ",
"RB= ",
"Ord=2",
LAST);

web_reg_find("Text=/manager",
LAST);

web_url("status",
"URL=http://{ServerName}/manager/status",
"Resource=0",
"RecContentType=text/html",
"Referer=",
"Snapshot=t1.inf",
"Mode=HTTP",
LAST);

lr_end_transaction("monitor tomcat", LR_AUTO);

// 写入Tomcat JVM 度量数据
lr_user_data_point("Tomcat JVM Free memory", atof(lr_eval_string("{JVMFreeMemory}")));
lr_user_data_point("Tomcat JVM Total memory", atof(lr_eval_string("{JVMTotalMemory}")));
lr_user_data_point("Tomcat JVM Max memory", atof(lr_eval_string("{JVMMaxMemory}")));

// 写入Tomcat web server 度量数据
lr_user_data_point("Tomcat HTTP Max threads", atof(lr_eval_string("{HTTPMaxThreads}")));
lr_user_data_point("Tomcat HTTP Min spare threads", atof(lr_eval_string("{HTTPMinSpareThreads}")));
lr_user_data_point("Tomcat HTTP Max spare threads", atof(lr_eval_string("{HTTPMaxSpareThreads}")));
lr_user_data_point("Tomcat HTTP Current spare threads", atof(lr_eval_string("{HTTPCurrentSpareThreads}")));
lr_user_data_point("Tomcat HTTP Current thread busy", atof(lr_eval_string("{HTTPCurrentThreadBusy}")));
lr_user_data_point("Tomcat HTTP Max processing time", atof(lr_eval_string("{HTTPMaxProcessingTime}")));
lr_user_data_point("Tomcat HTTP Request count", atof(lr_eval_string("{HTTPRequestCount}")));
lr_user_data_point("Tomcat HTTP Error count", atof(lr_eval_string("{HTTPErrorCount}")));
lr_user_data_point("Tomcat HTTP Bytes received", atof(lr_eval_string("{HTTPBytesReceived}")));
lr_user_data_point("Tomcat HTTP Bytes sent", atof(lr_eval_string("{HTTPBytesSent}")));

// 写入Tomcat servlet container 度量数据
lr_user_data_point("Tomcat JK Max threads", atof(lr_eval_string("{JKMaxThreads}")));
lr_user_data_point("Tomcat JK Min spare threads", atof(lr_eval_string("{JKMinSpareThreads}")));
lr_user_data_point("Tomcat JK Max spare threads", atof(lr_eval_string("{JKMaxSpareThreads}")));
lr_user_data_point("Tomcat JK Current spare threads", atof(lr_eval_string("{JKCurrentSpareThreads}")));
lr_user_data_point("Tomcat JK Current thread busy", atof(lr_eval_string("{JKCurrentThreadBusy}")));
lr_user_data_point("Tomcat JK Max processing time", atof(lr_eval_string("{JKMaxProcessingTime}")));
lr_user_data_point("Tomcat JK Request count", atof(lr_eval_string("{JKRequestCount}")));
lr_user_data_point("Tomcat JK Error count", atof(lr_eval_string("{JKErrorCount}")));
lr_user_data_point("Tomcat JK Bytes received", atof(lr_eval_string("{JKBytesReceived}")));
lr_user_data_point("Tomcat JK Bytes sent", atof(lr_eval_string("{JKBytesSent}")));

return 0;
}

View Code

2）然后在Run-time中设定数据收集的间隔：



3）最后在Controller中设置脚本运行，这样场景运行完毕后，即可在Analysis中添加“User Defined Data Points”查看收集到的Tomcat性能数据：



这种收集数据的弊端是需要通过网络请求Tomcat的页面，会对Tomcat服务器的性能造成一定的影响，如果Tomcat出现性能瓶颈，则收集数据时也可能碰到连接错误，因此，最好把收集数据点的间隔设大一点，例如5秒~10秒收集一个数据点。或尽量不采用LR监控。

3、[b]JConsole监控[/b]

JMX(Java Management Extensions)是一个为应用程序植入管理功能的框架。JMX是一套标准的代理和服务，实际上，用户可以在任何Java应用程序中使用这些代理和服务实现管理。可以利用JDK的JConsole来访问Tomcat JMX接口实施监控，具体步骤如下：

1）首先，打开Tomcat5的bin目录中的catalina.bat文件，添加：

JAVA_OPTS="-Xms512m -Xmx512m -Xmn256m  -XX:PermSize=64m -XX:MaxPermSize=64m  -Djava.rmi.server.hostname=192.168.222.132 -Dcom.sun.management.jmxremote.port=1090 -Dcom.sun.management.jmxremote.ssl=false -Dcom.sun.management.jmxremote.authenticate=false"

-Dcom.sun.management.jmxremote：代表开启JMX的管理功能

2）重启tomcat，并查看监控端口（上面配置的1090）是否已启动

3）打开jdk的bin目录（如C:\Program Files\Java\jdk1.7.0_17\bin）下的JConsole，并输入iP和监控端口进行连接



监控结果：



资料：

在Java SE 6中监视和诊断性能问题：http://www.ibm.com/developerworks/cn/java/j-java6perfmon/

4、VisualVM监控

VisualVM（All-in-One Java Trubleshooting Tool）是目前为止随JDK发布（在JDK1.6 update7中首次出现）的功能最强大的运行监视和故障处理程序，相对JProfiler、YourKit等专业收费工具最大的优点:不需要被监视的程序基于特殊Agent运行，它对应用程序实际性能影响很小，可以直接应用在生产环境中。

VisualVM基于NetBean平台开发，支持插件扩展，可以做到：



资料：

使用VisualVM进行性能分析及调优：http://www.ibm.com/developerworks/cn/java/j-lo-visualvm/

5、LambdaProbe监控

LambdaProbe是一款强大的免费开源工具，LambdaProbe拥有几乎所有Tomcat Manager的功能, 可以说是一个增强版本的Tomcat Manager。除此之外，Tomcat Probe 还拥有很多让开发者和系统管理者更方便的功能，从而使得Tomcat对开发者和管理者更加透明，包括：应用程序,数据源,发布,日志,线程,集群,系统信息,状态,连接器状态等功能。

1）下载LambdaProbe

2）解压probe.1.7b.zip文件，把probe.war部署到Tomcat服务器，修改tomcat-users.xml文件，添加账户权限：

<?xml version='1.0' encoding='utf-8'?>
<tomcat-users>
<role rolename="manager"/>
<role rolename="admin"/>
<user username="probe" password="probe" roles="admin,manager"/>
</tomcat-users>

3）设置tomcat启动参数
# vi ../bin/catalina.sh
JAVA_OPTS=-Dcom.sun.management.jmxremote

4）重启动Tomcat服务器

5）访问http://serverip/probe/,输入用户名和密码,进入即可对tomcat及jvm进行监控

6）汉化方法：将messages_zh_CN.zip文件解压缩, 将其中的 messages_zh_CN.properties 放到 $TOMCAT_HOME$\webapps\probe\WEB-INF 下即可

目前配置到第5步时访问遇到如下所示的问题：



问题原因：

Porbe只支持Tomcat服务器的版本: 5.0, 5.5, 6.0，而不支持更高版本，如7.0

解决方案：

从7.0换成6.0版本的tomcat即可：



6、JProfiler监控

JProfiler是一款Java的性能监控工具。可以查看当前应用的对象、对象引用、内存、CPU使用情况、线程、线程运行情况（阻塞、等待等），同时可以查找应用内存使用得热点，即：哪个对象占用的内存比较多；或者CPU热点，即：哪儿方法占用的较大得CPU资源。

1）将jprofiler_linux_7_2_3.tar.gz上传至Linux服务器，解压缩至opt目录下：



2）在客户端windows上安装jprofiler并打开：



3）依次按向导安装即可：



选择远程连接：



根据远程平台类型选择：



选择JVM版本、位数等：





待监控的远程服务器IP地址和Jprofiler监听的端口号（默认8864）：





4）到此步骤时，如下参数需要添加到服务器端的JVM启动参数中：





5）重启服务器端Tomcat服务

6）完成客户端远程连接：





大功告成：



资料：

JProfiler解决Java服务器的性能跟踪：http://www.blogjava.net/anymobile/articles/28248.html

常用Java Profiling工具的分析与比较：http://www.ibm.com/developerworks/cn/java/j-lo-profiling/