Github user roshannaik commented on a diff in the pull request:
https://github.com/apache/storm/pull/936#discussion_r49248342
--- Diff:
external/storm-hdfs/src/main/java/org/apache/storm/hdfs/spout/FileLock.java ---
@@ -0,0 +1,328 @@
+/**
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ * <p/>
+ * http://www.apache.org/licenses/LICENSE-2.0
+ * <p/>
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.storm.hdfs.spout;
+
+
+import org.apache.hadoop.fs.FSDataInputStream;
+import org.apache.hadoop.fs.FSDataOutputStream;
+import org.apache.hadoop.fs.FileSystem;
+import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.hdfs.DistributedFileSystem;
+import org.apache.hadoop.hdfs.protocol.AlreadyBeingCreatedException;
+import org.apache.hadoop.ipc.RemoteException;
+import org.apache.storm.hdfs.common.HdfsUtils;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.io.BufferedReader;
+import java.io.IOException;
+import java.io.InputStreamReader;
+import java.util.Collection;
+
+/**
+ * Facility to synchronize access to HDFS files. Thread gains exclusive
access to a file by acquiring
+ * a FileLock object. The lock itself is represented as file on HDFS.
Relies on atomic file creation.
+ * Owning thread must heartbeat periodically on the lock to prevent the
lock from being deemed as
+ * stale (i.e. lock whose owning thread have died).
+ */
+public class FileLock {
+
+ private final FileSystem fs;
+ private final String componentID;
+ private final Path lockFile;
+ private final FSDataOutputStream lockFileStream;
+ private LogEntry lastEntry;
+
+ private static final Logger LOG =
LoggerFactory.getLogger(FileLock.class);
+
+ private FileLock(FileSystem fs, Path lockFile, FSDataOutputStream
lockFileStream, String spoutId)
+ throws IOException {
+ this.fs = fs;
+ this.lockFile = lockFile;
+ this.lockFileStream = lockFileStream;
+ this.componentID = spoutId;
+ logProgress("0", false);
+ }
+
+ private FileLock(FileSystem fs, Path lockFile, String spoutId, LogEntry
entry)
+ throws IOException {
+ this.fs = fs;
+ this.lockFile = lockFile;
+ this.lockFileStream = fs.append(lockFile);
+ this.componentID = spoutId;
+ LOG.info("Acquired abandoned lockFile {}, Spout {}", lockFile,
spoutId);
+ logProgress(entry.fileOffset, true);
+ }
+
+ public void heartbeat(String fileOffset) throws IOException {
+ logProgress(fileOffset, true);
+ }
+
+ // new line is at beginning of each line (instead of end) for better
recovery from
+ // partial writes of prior lines
+ private void logProgress(String fileOffset, boolean prefixNewLine)
+ throws IOException {
+ long now = System.currentTimeMillis();
+ LogEntry entry = new LogEntry(now, componentID, fileOffset);
+ String line = entry.toString();
+ if(prefixNewLine) {
+ lockFileStream.writeBytes(System.lineSeparator() + line);
+ }
+ else {
+ lockFileStream.writeBytes(line);
+ }
+ lockFileStream.hflush();
+
+ lastEntry = entry; // update this only after writing to hdfs
+ }
+
+ /** Release lock by deleting file
+ * @throws IOException if lock file could not be deleted
+ */
+ public void release() throws IOException {
+ lockFileStream.close();
+ if(!fs.delete(lockFile, false)) {
+ LOG.warn("Unable to delete lock file, Spout = {}", componentID);
+ throw new IOException("Unable to delete lock file");
+ }
+ LOG.debug("Released lock file {}. Spout {}", lockFile, componentID);
+ }
+
+ // For testing only.. invoked via reflection
+ private void forceCloseLockFile() throws IOException {
+ lockFileStream.close();
+ }
+
+ /** returns lock on file or null if file is already locked. throws if
unexpected problem */
+ public static FileLock tryLock(FileSystem fs, Path fileToLock, Path
lockDirPath, String spoutId)
+ throws IOException {
+ String lockFileName = lockDirPath.toString() + Path.SEPARATOR_CHAR +
fileToLock.getName();
+ Path lockFile = new Path(lockFileName);
+
+ try {
+ FSDataOutputStream ostream = HdfsUtils.tryCreateFile(fs, lockFile);
+ if (ostream != null) {
+ LOG.debug("Acquired lock on file {}. LockFile= {}, Spout = {}",
fileToLock, lockFile, spoutId);
+ return new FileLock(fs, lockFile, ostream, spoutId);
+ } else {
+ LOG.debug("Cannot lock file {} as its already locked. Spout = {}",
fileToLock, spoutId);
+ return null;
+ }
+ } catch (IOException e) {
+ LOG.error("Error when acquiring lock on file " + fileToLock + "
Spout = " + spoutId, e);
+ throw e;
+ }
+ }
+
+ /**
+ * checks if lockFile is older than 'olderThan' UTC time by examining
the modification time
+ * on file and (if necessary) the timestamp in last log entry in the
file. If its stale, then
+ * returns the last log entry, else returns null.
+ * @param fs
+ * @param lockFile
+ * @param olderThan time (millis) in UTC.
+ * @return the last entry in the file if its too old. null if last entry
is not too old
+ * @throws IOException
+ */
+ public static LogEntry getLastEntryIfStale(FileSystem fs, Path lockFile,
long olderThan)
+ throws IOException {
+ long modifiedTime = fs.getFileStatus(lockFile).getModificationTime();
+ if( modifiedTime <= olderThan ) { // look
+ //Impt: HDFS timestamp may not reflect recent appends, so we double
check the
+ // timestamp in last line of file to see when the last update was
made
+ LogEntry lastEntry = getLastEntry(fs, lockFile);
+ if(lastEntry==null) {
+ throw new RuntimeException(lockFile.getName() + " is empty. this
file is invalid.");
--- End diff --
i see. good point. On discovering an empty lock file, we can
- simply skip over it after printing a log
- or delete it after logging .. as not much can be done with empty files
anyway.
i suggest the latter. what do you think ?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---