我正在开发一个简单的分布式分类帐。我希望能够在不同的端口上启动可以相互通信的节点。然后,每个程序都有一个文件,它将在其中写入新发现的节点。
首先,仅将最可靠的节点硬编码到该文件中。 程序上会发生以下情况:
1)我启动一个新节点,该节点启动一个HTTP服务器(我使用com.sun.HttpServer)。服务器具有GetAddress处理程序,该处理程序侦听转到指定URI的请求。然后,它获取IP和PORT(在URI查询参数中指定),获取 known_nodes.txt 文件的信号量,并将新接收的对等地址写入该文件(如果尚未存在) ,然后将新更新文件的内容作为json列表发送回请求者。
2)在我的Node类中(如前所述,该类在单独的线程上启动HTTPServer),我创建了ScheduledExecutorService并赋予它每隔几秒钟运行一次的运行时间,其工作是连接到URL出现在 known_nodes.txt 文件中,并询问他们的known_nodes。如果收到的已知节点文件中以前不存在节点,则将覆盖文件。
现在! 如果我启动一个节点,然后尝试从浏览器请求它,那么一切都会按计划进行-我们收到一个请求,将其写入我们的文件,然后我们的runnable将尝试连接到请求中指定的地址。如果我们捕获到一个 SocketTimeoutException ,则将从我们的known_nodes.txt文件中删除该地址。
问题出现了,当我启动两个节点时,假设在端口8001和8002上运行。请注意,每个节点都有其自己的known_nodes文件。 发生的情况是,其中一个节点将停止运行DiscoverAddresses任务,而另一个将不会。如此有效,一个节点停止了接收请求。
NB!将要停止其计划任务的节点将至少发送一个发现请求,然后死亡/阻止(?)。
以下是可运行任务的代码:
@Override
public void run() {
log.info("still running ");
PeerAddressesHolder inactiveNodes = new PeerAddressesHolder();
ApplicationConfiguration appConf = ApplicationConfiguration.getInstance();
for (PeerAddress peerAddress : knownNodes.getAddresses()) {
if (isSameNode(peerAddress)) {
continue;
}
String urlString = String.format("http://%s:%s%s?myport=%d", peerAddress.getIP(), peerAddress.getPort(), Constants.GET_ADDRESS, myPort);
try {
StringBuilder result = new StringBuilder();
URL url = new URL(urlString);
HttpURLConnection conn = (HttpURLConnection) url.openConnection();
conn.setConnectTimeout(5000);
conn.setRequestMethod("GET");
try (InputStream connInputStream = conn.getInputStream();
InputStreamReader ir = new InputStreamReader(connInputStream);
BufferedReader br = new BufferedReader(ir)){
String line;
while ((line = br.readLine()) != null) {
result.append(line).append('\n');
}
} catch (Exception e) {
log.warn("Couldn't read from connection input stream",e);
}
PeerAddressesHolder peerAddressesHolder = gson.fromJson(result.toString(), PeerAddressesHolder.class);
boolean fetchedNew = false;
for (PeerAddress fetchedAddress : peerAddressesHolder.getAddresses()) {
if (!isValidAddress(peerAddress)) {
log.warn("Peer has sent us a null-address. It will be ignored.");
return;
}
if (!knownNodes.contains(fetchedAddress)) {
knownNodes.addAddress(fetchedAddress);
fetchedNew = true;
}
}
if (fetchedNew) {
FileUtils.writeToFile(appConf.getKnownNodesFilePath(), gson.toJson(knownNodes), false);
}
} catch (SocketTimeoutException e) {
if (appConf.getMostReliableNodes().contains(peerAddress)) {
log.warn("Most reliable node not available: " + peerAddress);
} else {
inactiveNodes.addAddress(peerAddress);
log.warn("Connection timeout from " + peerAddress + ". It will be removed.");
}
} catch (Exception e) {
log.warn("Couldn't discover new addresses." + e);
}
}
try {
knownNodes.removeAll(inactiveNodes.getAddresses());
FileUtils.writeToFile(appConf.getKnownNodesFilePath(), gson.toJson(knownNodes), false);
} catch (IOException ioe) {
log.warn("Couldn't write to file after deleting dead node", ioe);
}
}
这是我在创建节点时启动它的方式。
public NetworkNode(int port) {
this.appConf = ApplicationConfiguration.getInstance();
this.port = port;
log.info("Starting a new node on port " + port);
try {
this.knownNodes = FileUtils.createPeerAddressesList(appConf.getKnownNodesFilePath());
} catch (Exception e) {
log.error("Error while trying to construct a list of peer addresses from file content on path: " + appConf.getKnownNodesFilePath());
}
scheduledExecutorService = Executors.newScheduledThreadPool(4);
scheduledExecutorService.scheduleAtFixedRate(new DiscoverAddressesTask(knownNodes, this.port), 3, 4, TimeUnit.SECONDS);
处理文件读/写的方法都是使用try-with-resources构造完成的,因此我最初的想法(即由于某些未关闭的流而导致可运行停止)可能无效。