生产环境中，有大规模的面向服务的架构和分布式应用，怎么样监控到这些服务接口的调用频次以及接口的性能表现呢？
per4j 日志组件是个不错的选择，具体使用可以参考这篇文章：使用Perf4J进行性能分析和监控
per4j 自身带的日志分析功能结合 Google Chart API 是个不错的可视化方案；可是有 GFW，服务器部署是个问题，难道自己搭建代理吗？境外服务器？
对per4j的日志格式简单的分析之后，我决定自己在per4j的基础上做一个监视系统。大致思路是：
- 自定义一个
log4j的appender，将per4j日志内容按记录点的名字、开始时间、消耗时长写入到一张日志表； - 然后使用
nodejs搭建一个简单的提供统计接口的服务，并使用chartjs展示统计结果；
具体的步骤如下：
创建日志表
我这里使用提 MySQL数据库，考虑到日志记录的效率，我将日志表的存储引擎设置为 ARCHIVE；因日志量庞大，我将日志按日期写入31张表（一个月最多有31天）；
CREATE TABLE `performance_log_{1-31}` (
`id` bigint(20) NOT NULL AUTO_INCREMENT,
`tag` varchar(32) NOT NULL,
`start` bigint(20) NOT NULL,
`duration` int(11) NOT NULL,
PRIMARY KEY (`id`)
) ENGINE=ARCHIVE;
解析per4j的日志格式
per4j的日志格式是这样的：
start[开始时间戳] time[消耗时长] tag[记录点] message[消息]
使用正则方式匹配出开始时间戳、消耗市场、记录点三个数据
public class Lite {
private String name;
private long start;
private int duration;
private static final Pattern pattern = Pattern.compile("^start\\["+ "(\\d+)" +"\\]\\s+time\\["+ "(\\d+)" +"\\]\\s+tag\\["+ "(\\w+)" +"\\]");
………
………
public static Lite from(String message) {
Matcher matcher = pattern.matcher(message);
if(matcher.find()) {
Long start = Long.parseLong(matcher.group(1));
int duration = Integer.parseInt(matcher.group(2));
String name = matcher.group(3);
return new Lite(name, start, duration);
} else {
return null;
}
}
自定义一个log4j的 database appender
package pfm.log4j;
import org.apache.commons.dbcp2.BasicDataSource;
import org.apache.log4j.AppenderSkeleton;
import org.apache.log4j.Logger;
import org.apache.log4j.spi.LoggingEvent;
import javax.sql.DataSource;
import java.sql.Connection;
import java.sql.PreparedStatement;
import java.sql.SQLException;
import java.util.*;
public class DailyDBAppender extends AppenderSkeleton {
private final Vector<Lite> buffer;
private int bufferSize = 10;
private String url;
private String username;
private String password;
private String driverClass;
private int initialSize = 5;
private int maxTotal = 50;
private int maxIdle = 8;
private int maxWaitMillis = 5000;
private int minIdle = 30000;
private DataSource dataSource;
private static Logger logger = Logger.getLogger(DailyDBAppender.class);
public DailyDBAppender() {
buffer = new Vector<>();
Timer timer = new Timer();
timer.schedule( new FlushTimerTask() , 10000, 3000 );
}
protected void append(LoggingEvent event) {
String message = (String) event.getMessage();
Lite lite = Lite.from(message);
if(lite == null) {
logger.warn("could not parse: + '"+ message +"'.");
return ;
}
buffer.add(lite);
if( buffer.size() >= bufferSize ) {
(new Thread(this::flush)).start();
}
}
private void flush() {
if (dataSource == null) {
dataSource = initDataSource(url,
driverClass,
username,
password,
initialSize,
maxTotal,
maxIdle,
maxWaitMillis,
minIdle);
}
Calendar calendar = Calendar.getInstance();
synchronized (buffer) {
try (Connection connection = dataSource.getConnection()) {
int dayOfMonth = calendar.get(Calendar.DAY_OF_MONTH);
String tableName = "performance_log_" + dayOfMonth;
String sql = "insert into "+ tableName +" (tag, start, duration) values (?, ?, ?)";
try (PreparedStatement statement =
connection.prepareStatement(sql)) {
int size = buffer.size();
Iterator<Lite> iterator = buffer.iterator();
while(iterator.hasNext()) {
Lite lite = iterator.next();
statement.setString(1, lite.getName());
statement.setLong(2, lite.getStart());
statement.setInt(3, lite.getDuration());
statement.executeUpdate();
iterator.remove();
}
logger.info("flush "+ size +" lites to db.\t\t\t\t\t\t\t\t\t" + buffer.size());
}
} catch (SQLException e) {
e.printStackTrace();
logger.warn("could not write log message to db. " + e.getMessage());
}
}
}
public void close() { }
public boolean requiresLayout() {
return false;
}
//这里省略掉数据库连接的属性和连接池相关的参数的 set 方法
private DataSource initDataSource(String url,
String driverClass,
String username,
String password,
int initialSize,
int maxTotal,
int maxIdle,
int maxWaitMillis,
int minIdle) {
BasicDataSource dataSource = new BasicDataSource();
dataSource.setDriverClassName(driverClass);
dataSource.setUsername(username);
dataSource.setPassword(password);
dataSource.setUrl(url);
dataSource.setInitialSize(initialSize);
dataSource.setMaxTotal(maxTotal);
dataSource.setMaxIdle(maxIdle);
dataSource.setMaxWaitMillis(maxWaitMillis);
dataSource.setMinIdle(minIdle);
return dataSource;
}
private final class FlushTimerTask extends TimerTask {
@Override
public void run() {
logger.debug("start to flush lites to db.");
if(buffer.size() > 0) {
flush();
}
}
}
}
使用了
commons-dbcp2数据库连接池；
增加配置到log4j.xml文件中
<appender name="dailydb" class="pfm.log4j.DailyDBAppender">
<param name="username" value="username" />
<param name="password" value="password" />
<param name="url" value="jdbc:mysql://localhost:3306/pfmlogdb" />
<param name="driverClass" value="com.mysql.jdbc.Driver" />
</appender>
<logger name="org.perf4j.TimingLogger" additivity="false">
<level value="INFO" />
<appender-ref ref="dailydb" />
</logger>
将per4j注解嵌入到记录点
我的要监视的接口程序是使用springmvc编写的，将记录点加到Controller层；
@Controller
public class SignController {
@Profiled(tag = "login")
@RequestMapping(value = "/login", method = RequestMethod.PUT)
@ResponseBody
public Response login(@RequestBody LoginRequest request) {
……
……
}
@Profiled(tag = "register")
@RequestMapping(value = "/register", method = RequestMethod.POST)
@ResponseBody
public Response register(@RequestBody RegisterRequest request) {
……
……
}
另外要将per4j的日志切面记录类以切面自动代理的方式配置到spring容器中；applicationContext.xml文件中增加下面的配置：
<aop:aspectj-autoproxy>
<aop:include name="timingAspect" />
</aop:aspectj-autoproxy>
<bean id="timingAspect" class="org.perf4j.slf4j.aop.TimingAspect" />
将接口程序运行起来后看到如下日志：
[2016-04-01 02:07:54,722] [INFO ] [Timer-0] pfm.log4j.DailyDBAppender - flush 3 lites to db. 0
[2016-04-01 02:08:00,715] [INFO ] [Timer-0] pfm.log4j.DailyDBAppender - flush 1 lites to db.
……
至此，即说明日志已经成功写入到数据库中；
mysql> select * from performance_log_1 limit 3;
+----+-------------------+---------------+----------+
| id | tag | start | duration |
+----+-------------------+---------------+----------+
| 1 | login | 1459440022770 | 31 |
| 2 | get_snapshot | 1459440022871 | 40 |
| 3 | get_snapshot | 1459440022932 | 23 |
+----+-------------------+---------------+----------+
3 rows in set (0.01 sec)
后续(B)篇继续接口访问频次的可视化方案的实现；