让我概括一下我的问题:
大约有4000台服务器,每台服务器有数百万个URL。我的代码需要命中每个url,并将响应代码与url一起写入hdfs文件系统。
这里也添加了一些内容:检查发送到网页的请求数
我使用的是400线程的生产者-消费者模型。这段代码最近对一些web服务器造成了dos攻击,我很难找出问题所在:
主要类别:
public void readURLS(final Path inputPath, final Path outputPath) {
LOG.info("Looking for files to download, queue size: {}, DOWNLOAD_THREADS: {}", queueSize, producerThreads);
final List<Path> files = HdfsUtils.listDirectory(inputPath, hadoopConf);
final BlockingQueue<String> queue = new LinkedBlockingQueue<>(queueSize);
final UrlConsumerWriter consumerWriter =
new UrlConsumerWriter(queue, outputPath, hadoopConf);
LOG.info("Starting download of {} files from: '{}'", files.size(), inputPath);
final ExecutorService writerPool = DownloadUtils.createWriterPool();
CompletableFuture<Void> producer = downloadFilesToQueue(files, queue)
.thenRun(consumerWriter::notifyProducersDone);
CompletableFuture<Void> consumer =
CompletableFuture.runAsync(consumerWriter, writerPool)// Cancel download workers if write worker fails
.whenComplete((result, err) -> {
if (err != null) {
LOG.error("Consumer Write worker failed!", err);
producer.cancel(true);
}
});
writerPool.shutdown();
producer.join();
consumer.join();
LOG.info("Url Validation Job Complete!!!");
}
private CompletableFuture<Void> downloadFilesToQueue(
final List<Path> files,
final BlockingQueue<String> downloadQueue
) {
final ExecutorService pool = DownloadUtils.createDownloadPool(producerThreads);
final List<CompletableFuture<Void>> workers = files.stream()
.map(file -> new UrlDownloadWorker(clock, file, hadoopConf, downloadQueue,
utils, (validatorImpl.emptyTable())))
.map(worker -> CompletableFuture.runAsync(worker, pool))
.collect(Collectors.toList());
pool.shutdown();
final CompletableFuture<Void> allDownloads = CompletableFuture.allOf(workers.toArray(new CompletableFuture[0]));
// When one worker fails, cancel all the other immediately
for (final CompletableFuture<Void> worker : workers) {
worker.whenComplete((v, err) -> {
if (err != null) {
LOG.error("Download worker failed!", err);
allDownloads.cancel(true);
}
});
}
return allDownloads;
}
制片人级别:
@Override
public void run() {
LOG.info("Starting download worker for file: '{}'", file);
long numLines = 0;
try (BufferedReader reader = new BufferedReader(new InputStreamReader(
file.getFileSystem(hadoopConf).open(file), CHARSET))) {
String line;
while ((line = reader.readLine()) != null) {
// LOG.info("Thread {} Reading file: '{}'",Thread.currentThread().getName(), file);
if (Thread.interrupted()) {
throw new InterruptedException();
}
StringBuilder builder = new StringBuilder();
//write into database
final StatusCode statusCode = utils.validateURL(line);
if (statusCode != null) {
queue.put(builder.append(line)
.append(",")
.append(statusCode.name()).toString());
builder.setLength(0);
} else {
throw new UrlValidationException(
"Failed to validate url :'" + line + "'");
}
numLines++;
}
} catch (IOException e) {
throw new DownloadException(file, e);
} catch (InterruptedException e) {
Thread.currentThread().interrupt();
throw new DownloadException("Interrupted while downloading", file, e);
}
LOG.info("Download of {} lines complete for file: '{}'", numLines, file);
}
urlvalidationutils类:
public final class UrlValidationUtils {
private static final String WEBSITENOTCHECK = "uncheck.org";
private final Map<String, StatusCode> blockedHosts = new ConcurrentHashMap<>();
private static final int MAX_REDIRECT = 4;
public StatusCode validateURL(String url) throws IOException {
return validate(url, MAX_REDIRECT);
}
private StatusCode validate(String url, int maxRedirect) throws IOException {
URL urlValue = new URL(url);
HttpURLConnection con;
if (url.contains(WEBSITENOTCHECK)) {
blockedHosts.put(urlValue.getHost(), StatusCode.SUCCESS);
}
//first check if the host is already marked as invalid
// if (blockedHosts.containsKey(urlValue.getHost())) {
// return blockedHosts.get(urlValue.getHost());
// }
StatusCode statusCode;
con = (HttpURLConnection) urlValue.openConnection();
try {
int resCode;
con.setInstanceFollowRedirects(false);
con.setConnectTimeout(3000); //set timeout to 3 seconds
con.connect();
resCode = con.getResponseCode();
LOG.info("thread name {} connection id {} url {} ", Thread.currentThread().getName(), con.toString(), url);
if (resCode == HttpURLConnection.HTTP_OK) {
statusCode = StatusCode.SUCCESS;
} else if (resCode == HttpURLConnection.HTTP_SEE_OTHER || resCode == HttpURLConnection.HTTP_MOVED_PERM
|| resCode == HttpURLConnection.HTTP_MOVED_TEMP) {
String location = con.getHeaderField("Location");
if (location.startsWith("/")) {
location = urlValue.getProtocol() + "://" + urlValue.getHost() + location;
}
statusCode = validateRedirect(location, maxRedirect - 1, con);
} else {
blockedHosts.put(urlValue.getHost(), StatusCode.FAIL);
statusCode = StatusCode.FAIL;
}
} catch (UnknownHostException e) {
blockedHosts.put(urlValue.getHost(), StatusCode.UNKOWNHOST);
statusCode = StatusCode.UNKOWNHOST;
} catch (ConnectException e) {
blockedHosts.put(urlValue.getHost(), StatusCode.CONNECTION_ISSUE);
statusCode = StatusCode.CONNECTION_ISSUE;
} catch (IOException e) {
//if an IOException is caught possible reason is SOCKETTIMEOUT
blockedHosts.put(urlValue.getHost(), StatusCode.SOCKETTIMEOUT);
statusCode = StatusCode.SOCKETTIMEOUT;
}
con.disconnect();
LOG.info("thread name {} connection id {} url {} ", Thread.currentThread().getName(), con.toString(), url);
return statusCode;
}
private StatusCode validateRedirect(String location, int redirectCount, HttpURLConnection connection)
throws IOException {
if (redirectCount >= 0) {
connection.disconnect();
return validate(location, redirectCount);
}
return StatusCode.FAIL;
}
}
1条答案
按热度按时间3yhwsihp1#
为了避免服务器过载,我建议在访问一批url之前等待几毫秒。例如,在点击n个URL之后,您可以等待20毫秒,然后点击下一个n。。等等。批处理的大小(n)取决于您知道服务器在一秒钟内可以处理多少个请求。在性能方面,您是否为他们签订了服务级别协议?