OpenTSDB源码详解之TSDMain类
1.TSDMain类
Main class of the TSD, the Time Series Daemon.
TSD的主类，时间序列的守护进程。
2.源代码
package net.opentsdb.tools;
import java.io.IOException;
import java.lang.reflect.Constructor;
import java.net.InetAddress;
import java.net.InetSocketAddress;
import java.util.concurrent.Executor;
import java.util.concurrent.Executors;
import java.util.HashMap;
import java.util.Map;
import org.jboss.netty.bootstrap.ServerBootstrap;
import org.jboss.netty.channel.socket.ServerSocketChannelFactory;
import org.jboss.netty.channel.socket.nio.NioServerBossPool;
import org.jboss.netty.channel.socket.nio.NioServerSocketChannelFactory;
import org.jboss.netty.channel.socket.nio.NioWorkerPool;
import org.jboss.netty.channel.socket.oio.OioServerSocketChannelFactory;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import net.opentsdb.tools.BuildData;
import net.opentsdb.core.TSDB;
import net.opentsdb.core.Const;
import net.opentsdb.tsd.PipelineFactory;
import net.opentsdb.tsd.RpcManager;
import net.opentsdb.utils.Config;
import net.opentsdb.utils.FileSystem;
import net.opentsdb.utils.Pair;
import net.opentsdb.utils.PluginLoader;
import net.opentsdb.utils.Threads;
/**
* Main class of the TSD, the Time Series Daemon.
*/
final class TSDMain {
/** Prints usage and exits with the given retval. */
static void usage(final ArgP argp, final String errmsg, final int retval) {
System.err.println(errmsg);
System.err.println("Usage: tsd --port=PORT"
+ " --staticroot=PATH --cachedir=PATH\n"
+ "Starts the TSD, the Time Series Daemon");
if (argp != null) {
System.err.print(argp.usage());
}
System.exit(retval);
}
/** A map of configured filters for use in querying */
private static Map<String, Pair<Class<?>, Constructor<? extends StartupPlugin>>>
startupPlugin_filter_map = new HashMap<String,
Pair<Class<?>, Constructor<? extends StartupPlugin>>>();
private static final short DEFAULT_FLUSH_INTERVAL = 1000;
private static TSDB tsdb = null;
public static void main(String[] args) throws IOException {
Logger log = LoggerFactory.getLogger(TSDMain.class);//将TSDMain.class这个类作为log标记
log.info("Starting.");//打印 starting
//BuildData :Build data for net.opentsdb.tools
//revisionString() 和 buildString()都只是获取一个string消息的过程
log.info(BuildData.revisionString());
log.info(BuildData.buildString());
try {
System.in.close(); // Release a FD we don't need.
} catch (Exception e) {
log.warn("Failed to close stdin", e);
}
//Argp :A dead simple command-line argument parser.
final ArgP argp = new ArgP();
CliOptions.addCommon(argp);
argp.addOption("--port", "NUM", "TCP port to listen on.");
argp.addOption("--bind", "ADDR", "Address to bind to (default: 0.0.0.0).");
argp.addOption("--staticroot", "PATH",
"Web root from which to serve static files (/s URLs).");
argp.addOption("--cachedir", "PATH",
"Directory under which to cache result of requests.");
argp.addOption("--worker-threads", "NUM",
"Number for async io workers (default: cpu * 2).");
argp.addOption("--async-io", "true|false",
"Use async NIO (default true) or traditional blocking io");
argp.addOption("--read-only", "true|false",
"Set tsd.mode to ro (default false)");
argp.addOption("--disable-ui", "true|false",
"Set tsd.core.enable_ui to false (default true)");
argp.addOption("--disable-api", "true|false",
"Set tsd.core.enable_api to false (default true)");
argp.addOption("--backlog", "NUM",
"Size of connection attempt queue (default: 3072 or kernel"
+ " somaxconn.");
argp.addOption("--max-connections", "NUM",
"Maximum number of connections to accept");
argp.addOption("--flush-interval", "MSEC",
"Maximum time for which a new data point can be buffered"
+ " (default: " + DEFAULT_FLUSH_INTERVAL + ").");
argp.addOption("--statswport", "Force all stats to include the port");
CliOptions.addAutoMetricFlag(argp);
args = CliOptions.parse(argp, args);
args = null; // free(). why args need to be null?
// get a config object
//读取配置文件的入口 => 下面这个就是设置properties文件的程序入口
Config config = CliOptions.getConfig(argp);
// check for the required parameters
try {
if (config.getString("tsd.http.staticroot").isEmpty())
usage(argp, "Missing static root directory", 1);
} catch(NullPointerException npe) {
usage(argp, "Missing static root directory", 1);
}
try {
if (config.getString("tsd.http.cachedir").isEmpty())
usage(argp, "Missing cache directory", 1);
} catch(NullPointerException npe) {
usage(argp, "Missing cache directory", 1);
}
try {
if (!config.hasProperty("tsd.network.port"))
usage(argp, "Missing network port", 1);
config.getInt("tsd.network.port");
} catch (NumberFormatException nfe) {
usage(argp, "Invalid network port setting", 1);
}
// validate the cache and staticroot directories
try {
FileSystem.checkDirectory(config.getString("tsd.http.cachedir"),
Const.MUST_BE_WRITEABLE, Const.CREATE_IF_NEEDED);
FileSystem.checkDirectory(config.getString("tsd.http.staticroot"),
!Const.MUST_BE_WRITEABLE, Const.DONT_CREATE);
} catch (IllegalArgumentException e) {
usage(argp, e.getMessage(), 3);
}
//--------------------------think over----------------------------
//A ChannelFactory which creates a ServerSocketChannel.
// 创建一个ServerSocketChannel 的ChannelFactory
final ServerSocketChannelFactory factory;
int connections_limit = 0;
try {
//默认的tsd.core.connections.limit的值是0 => default_map.put("tsd.core.connections.limit", "0");在Config类中
connections_limit = config.getInt("tsd.core.connections.limit");
} catch (NumberFormatException nfe) {
usage(argp, "Invalid connections limit", 1);
}
//Config类中设置的值是：default_map.put("tsd.network.async_io", "true");
if (config.getBoolean("tsd.network.async_io")) {
int workers = Runtime.getRuntime().availableProcessors() * 2;
//Config 类中的default_map.put("tsd.network.worker_threads", "");
//根据tsd.network.worker_threads这个配置得到线程数
if (config.hasProperty("tsd.network.worker_threads")) {
try {
workers = config.getInt("tsd.network.worker_threads");
} catch (NumberFormatException nfe) {
usage(argp, "Invalid worker thread count", 1);
}
}
//都是使用xxpool去提高性能
final Executor executor = Executors.newCachedThreadPool();
final NioServerBossPool boss_pool =
new NioServerBossPool(executor, 1, new Threads.BossThreadNamer());
final NioWorkerPool worker_pool = new NioWorkerPool(executor,
workers, new Threads.WorkerThreadNamer());
factory = new NioServerSocketChannelFactory(boss_pool, worker_pool);
} else {
factory = new OioServerSocketChannelFactory(
Executors.newCachedThreadPool(), Executors.newCachedThreadPool(),
new Threads.PrependThreadNamer());
}
StartupPlugin startup = null;
try {
startup = loadStartupPlugins(config);//load startUpPlugins
} catch (IllegalArgumentException e) {
usage(argp, e.getMessage(), 3);
} catch (Exception e) {
throw new RuntimeException("Initialization failed", e);
}
try {
tsdb = new TSDB(config);
if (startup != null) {
tsdb.setStartupPlugin(startup);
}
tsdb.initializePlugins(true);
if (config.getBoolean("tsd.storage.hbase.prefetch_meta")) {
tsdb.preFetchHBaseMeta();
}
// Make sure we don't even start if we can't find our tables.
tsdb.checkNecessaryTablesExist().joinUninterruptibly();
registerShutdownHook();
//ServerBootstrap: A helper class which creates a new server-side Channel and accepts incoming connections.
// 一个帮助类:去创建一个新的服务器端Channel 并且接受即将到来的连接
final ServerBootstrap server = new ServerBootstrap(factory);
// This manager is capable of lazy init, but we force an init here to fail fast.
// RpcManager : Manager for the lifecycle of HttpRpcs,TelnetRpcs,RpcPlugins, and HttpRpcPlugin
final RpcManager manager = RpcManager.instance(tsdb);
server.setPipelineFactory(new PipelineFactory(tsdb, manager, connections_limit));
if (config.hasProperty("tsd.network.backlog")) {
server.setOption("backlog", config.getInt("tsd.network.backlog"));
}
server.setOption("child.tcpNoDelay",
config.getBoolean("tsd.network.tcp_no_delay"));
server.setOption("child.keepAlive",
config.getBoolean("tsd.network.keep_alive"));
server.setOption("reuseAddress",
config.getBoolean("tsd.network.reuse_address"));
// null is interpreted as the wildcard address. null被解释为通配符地址。
//This class represents an Internet Protocol (IP) address. 这个类代表一个IP地址
InetAddress bindAddress = null;
if (config.hasProperty("tsd.network.bind")) {
bindAddress = InetAddress.getByName(config.getString("tsd.network.bind"));
}
// we validated the network port config earlier
final InetSocketAddress addr = new InetSocketAddress(bindAddress,
config.getInt("tsd.network.port"));
server.bind(addr);
if (startup != null) {
startup.setReady(tsdb);
}
log.info("Ready to serve on " + addr);
} catch (Throwable e) {
factory.releaseExternalResources();
try {
if (tsdb != null)
tsdb.shutdown().joinUninterruptibly();
} catch (Exception e2) {
log.error("Failed to shutdown HBase client", e2);
}
throw new RuntimeException("Initialization failed", e);
}
// The server is now running in separate threads, we can exit main.
System.out.println("main end"); // you can't read this output in stdout
}
private static StartupPlugin loadStartupPlugins(Config config) {
Logger log = LoggerFactory.getLogger(TSDMain.class);
// load the startup plugin if enabled
StartupPlugin startup = null;
if (config.getBoolean("tsd.startup.enable")) {
log.debug("Startup Plugin is Enabled");
final String plugin_path = config.getString("tsd.core.plugin_path");
final String plugin_class = config.getString("tsd.startup.plugin");
log.debug("Plugin Path: " + plugin_path);
try {
TSDB.loadPluginPath(plugin_path);
} catch (Exception e) {
log.error("Error loading plugins from plugin path: " + plugin_path, e);
}
log.debug("Attempt to Load: " + plugin_class);
startup = PluginLoader.loadSpecificPlugin(plugin_class, StartupPlugin.class);
if (startup == null) {
throw new IllegalArgumentException("Unable to locate startup plugin: " +
config.getString("tsd.startup.plugin"));
}
try {
startup.initialize(config);
} catch (Exception e) {
throw new RuntimeException("Failed to initialize startup plugin", e);
}
log.info("Successfully initialized startup plugin [" +
startup.getClass().getCanonicalName() + "] version: "
+ startup.version());
} else {
startup = null;
}
return startup;
}
private static void registerShutdownHook() {
final class TSDBShutdown extends Thread {
public TSDBShutdown() {
super("TSDBShutdown");
}
public void run() {
try {
if (RpcManager.isInitialized()) {
// Check that its actually been initialized. We don't want to
// create a new instance only to shutdown!
RpcManager.instance(tsdb).shutdown().join();
}
if (tsdb != null) {
tsdb.shutdown().join();
}
} catch (Exception e) {
LoggerFactory.getLogger(TSDBShutdown.class)
.error("Uncaught exception during shutdown", e);
}
}
}
Runtime.getRuntime().addShutdownHook(new TSDBShutdown());
}
}
3.分析
先看一下TSDMain类的uml类图，如下：
字段有startupPlugin_filter_map，DEFAULT_FLUSH_INTERVAL，tsdb。
方法有usage(ArgP,String,int);main(String []);loadStartupPlugins(Config);registerShutdownHook()
现在依次介绍：
字段：
- startupPlugin_filter_map：用于在查询时的配置的过滤器映射
- DEFAULT_FLUSH_INTERVAL ：默认的刷新间隔，为 1000【但是这个单位是什么？ms？】
- TSDB 的一个对象tsdb
方法：
- main，主方法。用于启动守护进程
- loadStartupPlugins：用于加载插件【但是这些插件的作用是什么？我暂时不了解】
- registerShutdownHook()：【暂不清楚】
4.运行结果
尝试运行TSDMain类中的main方法，可能会一直卡在某个地方，我就不大理解了，我的环境都已经配置通了，为什么什么都没有显示出来呢？我在浏览器输入localhost:4399时，报无法连接的错误。
后来再次研究main方法中的代码，存在如下的代码：
// Make sure we don't even start if we can't find our tables.
tsdb.checkNecessaryTablesExist().joinUninterruptibly();
这个意思就是确定OpenTSDB所需要的表存在才会继续运行，否则会一直卡住。再查看checkNecessaryTablesExist()方法，如下：
/**
* Verifies that the data and UID tables exist in HBase and optionally the
* tree and meta data tables if the user has enabled meta tracking or tree
* building
* 验证数据以及UID表存在于HBase中，并且可选的是，如果用户开启了meta跟踪或树构建，则也需要tree和meta数据表。
*
* @return An ArrayList of objects to wait for
* @throws TableNotFoundException
* @since 2.0
*/
public Deferred<ArrayList<Object>> checkNecessaryTablesExist() {
final ArrayList<Deferred<Object>> checks =
new ArrayList<Deferred<Object>>(2);
checks.add(client.ensureTableExists(
config.getString("tsd.storage.hbase.data_table")));//default_map.put("tsd.storage.hbase.data_table", "tsdb");
checks.add(client.ensureTableExists(
config.getString("tsd.storage.hbase.uid_table")));//default_map.put("tsd.storage.hbase.uid_table", "tsdb-uid");
//判断是否开启了tree building功能，如果开启了，则需要使用tree_table
if (config.enable_tree_processing()) {
checks.add(client.ensureTableExists(
config.getString("tsd.storage.hbase.tree_table")));
}
//如果开启了元数据统计，则需要表tesdb-meta
// default_map.put("tsd.storage.hbase.meta_table", "tsdb-meta");
if (config.enable_realtime_ts() || config.enable_realtime_uid() ||
config.enable_tsuid_incrementing()) {
checks.add(client.ensureTableExists(
config.getString("tsd.storage.hbase.meta_table")));
}
return Deferred.group(checks);
}
恍然大悟，是进程没有找到这个tsdb、tsdb-uid等表。然后接着查看集群进程，发现HMaster已经死亡。接着重启HBase进程，再次运行即可。然后得到最后的输出界面如下：
main end表示整个main方法都已经运行结束了。打开网页localhost:4399，可以看到如下的页面【只不过是图像莫名丢失啦：（】
下面就来解决这个图片消失的问题。
在浏览器中触发这个数据查询时，会在后台触发java代码，并且在intellij的标准输出中输出。可以看到此时的输出结果为：
20:44:28.852 [OpenTSDB I/O Boss #1] INFO net.opentsdb.tsd.ConnectionManager - [id: 0x99b05f33, /192.168.211.2:6273 => /192.168.211.2:4399] OPEN
20:44:28.854 [OpenTSDB I/O Worker #1] INFO net.opentsdb.tsd.ConnectionManager - [id: 0x99b05f33, /192.168.211.2:6273 => /192.168.211.2:4399] BOUND: /192.168.211.2:4399
20:44:28.854 [OpenTSDB I/O Worker #1] INFO net.opentsdb.tsd.ConnectionManager - [id: 0x99b05f33, /192.168.211.2:6273 => /192.168.211.2:4399] CONNECTED: /192.168.211.2:6273
20:44:28.858 [OpenTSDB I/O Worker #1] DEBUG o.j.n.h.c.http.HttpContentCompressor - -Dio.netty.noJdkZlibEncoder: false
20:44:28.872 [OpenTSDB I/O Worker #1] INFO net.opentsdb.tsd.HttpQuery - [id: 0x99b05f33, /192.168.211.2:6273 => /192.168.211.2:4399] HTTP / done in 8ms
20:44:28.933 [OpenTSDB I/O Worker #1] WARN net.opentsdb.tsd.HttpQuery - [id: 0x99b05f33, /192.168.211.2:6273 => /192.168.211.2:4399] File not found: \root\openTSDB_temp\queryui.nocache.js (系统找不到指定的文件。)
20:44:28.934 [OpenTSDB I/O Worker #1] INFO net.opentsdb.tsd.HttpQuery - [id: 0x99b05f33, /192.168.211.2:6273 => /192.168.211.2:4399] HTTP /s/queryui.nocache.js done in 1ms
20:44:28.936 [OpenTSDB I/O Worker #1] WARN net.opentsdb.tsd.HttpQuery - [id: 0x99b05f33, /192.168.211.2:6273 => /192.168.211.2:4399] File not found: \root\openTSDB_temp\opentsdb_header.jpg (系统找不到指定的文件。)
20:44:28.937 [OpenTSDB I/O Worker #1] INFO net.opentsdb.tsd.HttpQuery - [id: 0x99b05f33, /192.168.211.2:6273 => /192.168.211.2:4399] HTTP /s/opentsdb_header.jpg done in 0ms
20:44:28.943 [OpenTSDB I/O Boss #1] INFO net.opentsdb.tsd.ConnectionManager - [id: 0xadcd8ee4, /192.168.211.2:6274 => /192.168.211.2:4399] OPEN
20:44:28.944 [OpenTSDB I/O Worker #2] INFO net.opentsdb.tsd.ConnectionManager - [id: 0xadcd8ee4, /192.168.211.2:6274 => /192.168.211.2:4399] BOUND: /192.168.211.2:4399
20:44:28.944 [OpenTSDB I/O Worker #2] INFO net.opentsdb.tsd.ConnectionManager - [id: 0xadcd8ee4, /192.168.211.2:6274 => /192.168.211.2:4399] CONNECTED: /192.168.211.2:6274
20:44:30.019 [OpenTSDB I/O Worker #1] INFO net.opentsdb.tsd.HttpQuery - [id: 0x99b05f33, /192.168.211.2:6273 => /192.168.211.2:4399] HTTP / done in 0ms
20:44:30.073 [OpenTSDB I/O Worker #1] WARN net.opentsdb.tsd.HttpQuery - [id: 0x99b05f33, /192.168.211.2:6273 => /192.168.211.2:4399] File not found: \root\openTSDB_temp\queryui.nocache.js (系统找不到指定的文件。)
20:44:30.074 [OpenTSDB I/O Worker #1] INFO net.opentsdb.tsd.HttpQuery - [id: 0x99b05f33, /192.168.211.2:6273 => /192.168.211.2:4399] HTTP /s/queryui.nocache.js done in 1ms
20:44:30.076 [OpenTSDB I/O Worker #2] WARN net.opentsdb.tsd.HttpQuery - [id: 0xadcd8ee4, /192.168.211.2:6274 => /192.168.211.2:4399] File not found: \root\openTSDB_temp\opentsdb_header.jpg (系统找不到指定的文件。)
20:44:30.077 [OpenTSDB I/O Worker #2] INFO net.opentsdb.tsd.HttpQuery - [id: 0xadcd8ee4, /192.168.211.2:6274 => /192.168.211.2:4399] HTTP /s/opentsdb_header.jpg done in 0ms
20:44:32.184 [OpenTSDB I/O Worker #1] INFO net.opentsdb.tsd.HttpQuery - [id: 0x99b05f33, /192.168.211.2:6273 => /192.168.211.2:4399] HTTP / done in 1ms
20:44:32.239 [OpenTSDB I/O Worker #1] WARN net.opentsdb.tsd.HttpQuery - [id: 0x99b05f33, /192.168.211.2:6273 => /192.168.211.2:4399] File not found: \root\openTSDB_temp\queryui.nocache.js (系统找不到指定的文件。)
20:44:32.240 [OpenTSDB I/O Worker #1] INFO net.opentsdb.tsd.HttpQuery - [id: 0x99b05f33, /192.168.211.2:6273 => /192.168.211.2:4399] HTTP /s/queryui.nocache.js done in 1ms
20:44:32.241 [OpenTSDB I/O Worker #2] WARN net.opentsdb.tsd.HttpQuery - [id: 0xadcd8ee4, /192.168.211.2:6274 => /192.168.211.2:4399] File not found: \root\openTSDB_temp\opentsdb_header.jpg (系统找不到指定的文件。)
20:44:32.242 [OpenTSDB I/O Worker #2] INFO net.opentsdb.tsd.HttpQuery - [id: 0xadcd8ee4, /192.168.211.2:6274 => /192.168.211.2:4399] HTTP /s/opentsdb_header.jpg done in 0ms
20:44:32.270 [OpenTSDB I/O Worker #1] WARN net.opentsdb.tsd.HttpQuery - [id: 0x99b05f33, /192.168.211.2:6273 => /192.168.211.2:4399] File not found: \root\openTSDB_temp\opentsdb_header.jpg (系统找不到指定的文件。)
可以很清楚的看到系统找不到指定的文件错，也就是因为没有拿到这些文件导致出错。主要的文件有：
- opentsdb_header.jpg
- queryui.nocache.js
查看opentsdb.properties文件，其中的两行代码如下：
tsd.http.cachedir = G:\\testdb
tsd.http.staticroot = G:\\testdb
这两行代码就是存储Opentsdb的一个前端页面文件，但是在G:\testdb目录中却是没有文件的。我将之前Centos 7中安装的opentsdb的文件拿了过来，最后就得到了相应的界面，如下所示：
这次出现了opentsdb_header的图像，但是主要的躯干还没有显示出来，原因一样，也是因为文件丢失。从以前的安装包里copy一份即可。再次刷新一下即可：
最后得到的图形界面如上，到这里通过源码编译opentsdb的第一步算是成功了。接下来会逐渐深入。
注
整个源码分析过程中遇到的困难我思考了很久才得以解决，这里将其稍微记录一下：
- 1.在opentsdb.properties文件中，可以看到如下的代码：
# ----------- HTTP -----------
# The location of static files for the HTTP GUI interface.
# *** REQUIRED ***
#i don't know why can't set staticroot to /usr/local/opentsdb-2.3.0/build/staticroot,must to /root/openTSDB_temp
#tsd.http.staticroot = /usr/local/opentsdb-2.3.0/build/staticroot
#tsd.http.staticroot = /root/openTSDB_temp
tsd.http.staticroot = G:\\testdb
tsd.http.staticroot的路径我变换了很多次，从/usr/local/opentsdb-2.3.0/build/staticroot => /root/openTSDB_temp => G:\testdb。才恍然大悟，运行在windows上的程序，你是无法把其文件路径指向其它机器（如虚拟机上）的文件夹。这里必须使用windows本地的路径才可以。
- 2.为什么我们在浏览器中输入了metric，date…之后，就可以触发整个数据的查询呢？原因是，这是由前端语言javascript和ajax控制的，如果前端的页面有变化，则会触发后台的java程序工作。
