如何讓Java應用成爲殺不死的小強?(下篇)

【這是一猿小講的第 49 篇原創分享】java


各位坐穩扶好,咱們要開車了。不過在開車以前,咱們仍是例行回顧一下上期分享的要點。windows


通過前兩期的鋪墊及燒腦的分享,咱們大概對「如何實現 Java 應用進程的狀態監控,若是被監控的進程 down 掉,是否有機制能啓動起來?」問題自己有了一個新的認識,那這期咱們不妨拿出攻城獅的絕招 Ctrl + C、Ctrl + V,從 Resin 源碼中摘取一二,稍微簡單實踐一下。bash



按照圖示,我們先演示一下實踐效果吧,首先找到並運行程序入口 MonitorApp,日誌輸出以下。服務器


此時咱們不妨在控制檯輸入 jps 命令,看一看效果。app


18830 MonitorAppjvm

18831 Resinsocket


發現成功啓動了 MonitorApp、Resin 兩個進程,和 Resin 應用服務器是如出一轍的,若是咱們把進程號爲 18831 的 kill 掉,會是什麼效果?發現控制檯日誌輸出又多了一些,貌似丫環 Resin 又被從新給啓動了。ide


在控制檯輸入 jps 命令再確認一下是否真的變了。函數


18830 MonitorAppui

18935 Resin


那咱們到底該如何實現?那不妨照葫蘆畫瓢,模仿一下 Resin 的實現一下(這就是絕招:仿一仿)。


首先定義咱們的監控應用入口 MonitorApp,很簡單就是把建立子進程的任務給啓動起來。

package com.caucho.server.resin;
public class MonitorApp {
    public static void main(String[] args) {
        new WatchdogChildTask().start();
    }
}複製代碼

接下來再編寫 WatchdogChildTask 子進程任務的代碼,大部分來源於 Resin 的源碼,只是剔除了不少不少不少,簡化了不少不少不少。仔細看發現也很簡單,就有一個循環一直調用 WatchdogChildProcess 的 run 方法,目的也就是一直讓丫環進程跑起來。

package com.caucho.server.resin;

import java.util.concurrent.Executors;
import java.util.logging.Level;
import java.util.logging.Logger;

class WatchdogChildTask implements Runnable {

    private static final Logger log = Logger.getLogger(WatchdogChildTask.class.getName());

    private WatchdogChildProcess _process;

    /**
     * Starts management of the watchdog process
     */
    public void start() {
        //TODO 手動建立線程池會更好 【阿里開發規約】
        Executors.newFixedThreadPool(1).execute(this);
    }

    /**
     * Main thread watching over the health of the Resin instances.
     */
    public void run() {
        try {
            int i = 0;
            long retry = Long.MAX_VALUE;
            while (i++ < retry) {
                WatchdogChildProcess process = new WatchdogChildProcess();
                _process = process;
                try {
                    log.log(Level.INFO, "我是大總管,準備讓乳名爲Resin的丫環跑起來");
                    _process.run();
                } catch (Exception e) {
                    log.log(Level.WARNING, e.toString(), e);
                } finally {
                    _process = null;
                    if (process != null) {
                        log.log(Level.INFO, "我是大總管,發現乳名爲Resin的丫環出情況了,須要讓她釋放資源,從新跑起來");
                        process.kill();
                    }
                }
            }
        } catch (Exception e) {
            log.log(Level.WARNING, e.toString(), e);
        } finally {
            if (_process != null) {
                _process.kill();
                _process = null;
            }
        }
    }
}複製代碼

具體是怎麼把丫環進程跑起來的,這個事情專門交給 WatchdogChildProcess 去作了,先啓動了一個 socket 通信端口;而後採用 ProcessBuilder 啓動 Resin 進程;而後等待丫環進程創建 socket 鏈接通信。大部分也是來源於 Resin 的源碼,只不過作了大量刪減。另外重點提一嘴:拿下去只需修改 com.caucho.server.resin.Resin 爲你要監控應用的主函數便可。

package com.caucho.server.resin;

import java.io.*;
import java.net.*;
import java.util.*;
import java.util.concurrent.atomic.AtomicReference;
import java.util.logging.Level;
import java.util.logging.Logger;

class WatchdogChildProcess {

    private static final Logger log = Logger.getLogger(WatchdogChildProcess.class.getName());

    private Socket _childSocket;
    private OutputStream _stdOs;
    private int _status = -1;
    private AtomicReference _processRef = new AtomicReference();

    public void run() {
        ServerSocket ss = null;
        Socket s = null;
        try {
            ss = new ServerSocket(0, 5, InetAddress.getByName("127.0.0.1"));
            int port = ss.getLocalPort();
            log.log(Level.INFO, "我是大總管,我啓動一個端口爲{0}的socket,讓丫環們實時與我通信",port);

            Process process = createProcess(port);
            if (process != null) {
                _processRef.compareAndSet(null, process);

                InputStream stdIs = process.getInputStream();
                _stdOs = process.getOutputStream();

                //TODO 不要顯式建立線程,請使用線程池【阿里開發規約】
                new Thread(new WatchdogProcessLogThread(stdIs)).start();

                s = connectToChild(ss);
                _status = process.waitFor();
                logStatus(_status);
            }
        } catch (Exception e) {
            log.log(Level.WARNING, e.toString(), e);
            try {
                Thread.sleep(5000);
            } catch (Exception e1) {
            }
        } catch (Throwable e) {
            log.log(Level.WARNING, e.toString(), e);
        } finally {
            if (ss != null) {
                try {
                    ss.close();
                } catch (Throwable e) {
                }
            }
            try {
                if (s != null) {
                    s.close();
                }
            } catch (Throwable e) {
                log.log(Level.FINER, e.toString(), e);
            }
            kill();
            synchronized (this) {
                notifyAll();
            }
        }
    }

    private void logStatus(int status) {
        String code = " (exit code=" + status + ")";
        log.warning("大總管忽然發現丫環進程罷工了!!");
    }

    void kill() {
        Process process = _processRef.getAndSet(null);
        if (process != null) {
            try {
                process.destroy();
            } catch (Exception e) {
                log.log(Level.FINE, e.toString(), e);
            }
        }

        OutputStream stdOs = _stdOs;
        _stdOs = null;
        if (stdOs != null) {
            try {
                stdOs.close();
            } catch (Throwable e) {
                log.log(Level.FINE, e.toString(), e);
            }
        }

        Socket childSocket = _childSocket;
        _childSocket = null;

        if (childSocket != null) {
            try {
                childSocket.close();
            } catch (Throwable e) {
                log.log(Level.FINE, e.toString(), e);
            }
        }

        if (process != null) {
            try {
                process.waitFor();
            } catch (Exception e) {
                log.log(Level.INFO, e.toString(), e);
            }
        }
    }

    /**
     * Waits for a socket connection from the child, returning the socket
     *
     * @param ss TCP ServerSocket from the watchdog for the child to connect to
     */
    private Socket connectToChild(ServerSocket ss)
            throws IOException {
        Socket s = null;
        try {
            ss.setSoTimeout(60000);
            for (int i = 0; i < 120 && s == null; i++) {
                try {
                    s = ss.accept();
                } catch (SocketTimeoutException e) {
                }
            }

            if (s != null) {
                _childSocket = s;
            }
        } catch (Exception e) {
            log.log(Level.WARNING, e.toString(), e);
        } finally {
            ss.close();
        }
        return s;
    }

    /**
     * Creates a new Process for the Resin JVM, initializing the environment
     * and passing value to the new process.
     *
     * @param socketPort the watchdog socket port
     * @param out        the debug log jvm-default.log
     */
    private Process createProcess(int socketPort)
            throws IOException {
        HashMap env = buildEnv();
        ArrayList jvmArgs = buildJvmArgs();

        jvmArgs.add("com.caucho.server.resin.Resin");
        jvmArgs.add("-socketwait");
        jvmArgs.add(String.valueOf(socketPort));

        ProcessBuilder builder = new ProcessBuilder();
        builder.environment().putAll(env);
        builder = builder.command(jvmArgs);
        builder.redirectErrorStream(true);
        return builder.start();
    }

    private HashMap buildEnv()
            throws IOException {
        HashMap env = new HashMap();
        env.putAll(System.getenv());

        StringBuilder classPath = new StringBuilder();
        // TODO 系統不同分割符也不一樣 windows爲分號;
        classPath.append(".:");

        String appPath = System.getProperty("user.dir");
        classPath.append(appPath).append("/resin/target/classes");
        env.put("CLASSPATH", classPath.toString());

        // 。。。 刪除了可多可多的代碼 。。。
        return env;
    }

    private ArrayList buildJvmArgs() {
        ArrayList jvmArgs = new ArrayList();
        jvmArgs.add("java");
        // ... 又刪除了可多代碼 ...
        return jvmArgs;
    }

    /**
     * Watchdog thread responsible for writing jvm-default.log by reading the
     * JVM's stdout and copying it to the log. */ class WatchdogProcessLogThread implements Runnable { private InputStream _is; /** * @param is the stdout stream from the Resin */ WatchdogProcessLogThread(InputStream is) { _is = is; } @Override public void run() { try { int len; byte[] data = new byte[4096]; while ((len = _is.read(data, 0, data.length)) > 0) { System.out.print(new String(data, 0, len)); } } catch (Throwable e) { log.log(Level.WARNING, e.toString(), e); } finally { kill(); } } } }複製代碼

下面這個要重點說下,由於這套模型你拿過去,只需修改下面 Resin 這個類的代碼,這個其實也就是咱們要監控的應用。其實很簡單,就有一個 connect 方法主要用於與大總管進行通信,一旦通信失敗自己就退出。

package com.caucho.server.resin;

import java.io.IOException;
import java.io.InputStream;
import java.net.Socket;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.logging.Level;
import java.util.logging.Logger;

public class Resin {

    private static ExecutorService executorService = Executors.newFixedThreadPool(1);

    private static final Logger log = Logger.getLogger(Resin.class.getName());

    public static void main(String[] args) {
        log.log(Level.INFO, "我是乳名爲Resin的丫環,大總管給的通信端口爲{0} {1}", args);
        //獲取傳入的參數 port
        int port = Integer.parseInt(args[1]);

        connect(port);
    }

    public static void connect(final int port) {
        log.log(Level.INFO, "我是乳名爲Resin的丫環,我要開始與端口爲{0}的大總管進行通信",port);
        executorService.execute(new Runnable() {
            @Override
            public void run() {
                Socket socket = null;
                try {
                    socket = new Socket("127.0.0.1", port);
                    InputStream s = socket.getInputStream();
                    byte[] buf = new byte[1024];
                    int len;
                    while ((len = s.read(buf)) != -1) {
                        log.log(Level.INFO, "通信信息 {0}", new String(buf, 0, len));
                    }
                } catch (IOException e) {
                    log.log(Level.WARNING, "我是乳名爲Resin的丫環,與端口爲{0}的大總管進行通信發生異常",port);
                } finally {
                    try {
                        socket.close();
                    } catch (IOException e) {
                        log.log(Level.WARNING, e.getMessage(), e);
                    }
                    log.log(Level.INFO, "我是乳名爲Resin的丫環,與端口爲{0}的大總管進行通信結束,我要退下啦",port);
                    System.exit(0);
                }
            }
        });
        log.log(Level.INFO, "丫環與大總管通信完成");
    }
}複製代碼

到這,代碼也就碼完了,不妨把代碼拔下去,運行一下,稍微體驗體驗,看看是否是那回事兒!其中爲了演示須要刪除了 N 多代碼,有些地方很不優雅,還需按照阿里開發規約適當調整調整,不過這些不是我們這期分享的重點,我們重點是思想 + 輕實踐。


好了,思想也落地了,接下來就看你怎麼讓它老樹開新花啦。分享就到這兒吧,但願可以解你所惑;但願能在你前進的道路上,幫你披荊斬棘。若是感受有點幫助,歡迎在看、秒贊,瘋狂分享轉發,由於你的每一次分享,我都認真當成了鼓勵與喜歡。

推薦閱讀:

老技術新談,Java應用監控利器JMX(1)

老技術新談,Java應用監控利器JMX(2)

老技術新談,Java應用監控利器JMX(3)

一篇文章講透線上應用監控

相關文章
相關標籤/搜索