HADOOP-2357 Compaction cleanup; less deleting + prevent possible file leaks

git-svn-id: https://svn.apache.org/repos/asf/lucene/hadoop/trunk/src/contrib/hbase@601383 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
Michael Stack 2007-12-05 16:06:25 +00:00
parent afeed7bfc6
commit 0ce1829c73
5 changed files with 155 additions and 153 deletions

View File

@ -91,6 +91,7 @@ Trunk (unreleased changes)
(Edward Yoon via Stack)
HADOOP-2299 Support inclusive scans (Bryan Duxbury via Stack)
HADOOP-2333 Client side retries happen at the wrong level
HADOOP-2357 Compaction cleanup; less deleting + prevent possible file leaks
Release 0.15.1

View File

@ -151,11 +151,12 @@ public class HLog implements HConstants {
try {
for (int i = 0; i < logfiles.length; i++) {
if (LOG.isDebugEnabled()) {
LOG.debug("Splitting " + logfiles[i]);
LOG.debug("Splitting " + i + " of " + logfiles.length + ": " +
logfiles[i]);
}
// Check for empty file.
if (fs.getFileStatus(logfiles[i]).getLen() <= 0) {
LOG.warn("Skipping " + logfiles[i].toString() +
LOG.info("Skipping " + logfiles[i].toString() +
" because zero length");
continue;
}
@ -164,26 +165,29 @@ public class HLog implements HConstants {
try {
HLogKey key = new HLogKey();
HLogEdit val = new HLogEdit();
while (in.next(key, val)) {
int count = 0;
for (; in.next(key, val); count++) {
Text regionName = key.getRegionName();
SequenceFile.Writer w = logWriters.get(regionName);
if (w == null) {
Path logfile = new Path(HRegion.getRegionDir(rootDir,
HRegionInfo.encodeRegionName(regionName)),
HREGION_OLDLOGFILE_NAME);
if (LOG.isDebugEnabled()) {
LOG.debug("getting new log file writer for path " + logfile);
LOG.debug("Creating new log file writer for path " + logfile);
}
w = SequenceFile.createWriter(fs, conf, logfile, HLogKey.class,
HLogEdit.class);
logWriters.put(regionName, w);
}
if (LOG.isDebugEnabled()) {
LOG.debug("Edit " + key.toString() + "=" + val.toString());
if (count % 100 == 0 && count > 0 && LOG.isDebugEnabled()) {
LOG.debug("Applied " + count + " edits");
}
w.append(key, val);
}
if (LOG.isDebugEnabled()) {
LOG.debug("Applied " + count + " total edits");
}
} finally {
in.close();
}

View File

@ -245,9 +245,9 @@ public class HMaster extends Thread implements HConstants, HMasterInterface,
String serverName = Writables.bytesToString(results.get(COL_SERVER));
long startCode = Writables.bytesToLong(results.get(COL_STARTCODE));
if (LOG.isDebugEnabled()) {
LOG.debug(Thread.currentThread().getName() + " scanner: " +
Long.valueOf(scannerId) + " regioninfo: {" + info.toString() +
"}, server: " + serverName + ", startCode: " + startCode);
LOG.debug(Thread.currentThread().getName() + " regioninfo: {" +
info.toString() + "}, server: " + serverName + ", startCode: " +
startCode);
}
// Note Region has been assigned.
@ -447,9 +447,7 @@ public class HMaster extends Thread implements HConstants, HMasterInterface,
storedInfo = serversToServerInfo.get(serverName);
deadServer = deadServers.contains(serverName);
}
if (LOG.isDebugEnabled()) {
LOG.debug("Checking " + info.getRegionName() + " is assigned");
}
/*
* If the server is not dead and either:
* the stored info is not null and the start code does not match

View File

@ -476,7 +476,9 @@ public class HRegion implements HConstants {
return this.conf;
}
/** @return region directory Path */
/** @return region directory Path
* @see HRegion#getRegionDir(Path, String)
*/
public Path getRegionDir() {
return this.regiondir;
}
@ -878,11 +880,6 @@ public class HRegion implements HConstants {
*/
private boolean internalFlushcache(long startTime) throws IOException {
if (startTime == -1) {
if (LOG.isDebugEnabled()) {
LOG.debug("Not flushing cache for region " +
regionInfo.getRegionName() +
": snapshotMemcaches() determined that there was nothing to do");
}
return false;
}
@ -1633,13 +1630,17 @@ public class HRegion implements HConstants {
*
* @param fs the file system object
* @param baseDirectory base directory for HBase
* @param name region file name
* @param name region file name ENCODED!
* @throws IOException
* @return True if deleted.
* @see HRegionInfo#encodeRegionName(Text)
*/
static boolean deleteRegion(FileSystem fs, Path baseDirectory, String name)
throws IOException {
Path p = HRegion.getRegionDir(fs.makeQualified(baseDirectory), name);
if (LOG.isDebugEnabled()) {
LOG.debug("DELETING region " + p.toString());
}
return fs.delete(p);
}
@ -1647,8 +1648,9 @@ public class HRegion implements HConstants {
* Computes the Path of the HRegion
*
* @param dir hbase home directory
* @param name region file name
* @param name region file name ENCODED!
* @return Path of HRegion directory
* @see HRegionInfo#encodeRegionName(Text)
*/
public static Path getRegionDir(final Path dir, final String name) {
return new Path(dir, new Path(HREGIONDIR_PREFIX + name));

View File

@ -542,7 +542,8 @@ class HStore implements HConstants {
HBaseConfiguration conf) throws IOException {
this.dir = dir;
this.compactionDir = new Path(dir, "compaction.dir");
this.compactionDir = new Path(HRegion.getRegionDir(dir, encodedName),
"compaction.dir");
this.regionName = regionName;
this.encodedRegionName = encodedName;
this.family = family;
@ -603,16 +604,7 @@ class HStore implements HConstants {
// means it was built prior to the previous run of HStore, and so it cannot
// contain any updates also contained in the log.
long maxSeqID = -1;
for (HStoreFile hsf: hstoreFiles) {
long seqid = hsf.loadInfo(fs);
if(seqid > 0) {
if(seqid > maxSeqID) {
maxSeqID = seqid;
}
}
}
this.maxSeqId = maxSeqID;
this.maxSeqId = getMaxSequenceId(hstoreFiles);
if (LOG.isDebugEnabled()) {
LOG.debug("maximum sequence id for hstore " + storeName + " is " +
this.maxSeqId);
@ -641,6 +633,25 @@ class HStore implements HConstants {
}
}
/*
* @param hstoreFiles
* @return Maximum sequence number found or -1.
* @throws IOException
*/
private long getMaxSequenceId(final List<HStoreFile> hstoreFiles)
throws IOException {
long maxSeqID = -1;
for (HStoreFile hsf : hstoreFiles) {
long seqid = hsf.loadInfo(fs);
if (seqid > 0) {
if (seqid > maxSeqID) {
maxSeqID = seqid;
}
}
}
return maxSeqID;
}
long getMaxSequenceId() {
return this.maxSeqId;
}
@ -670,16 +681,17 @@ class HStore implements HConstants {
try {
HLogKey key = new HLogKey();
HLogEdit val = new HLogEdit();
long skippedEdits = 0;
while (login.next(key, val)) {
maxSeqIdInLog = Math.max(maxSeqIdInLog, key.getLogSeqNum());
if (key.getLogSeqNum() <= maxSeqID) {
if (LOG.isDebugEnabled()) {
LOG.debug("Skipping edit <" + key.toString() + "=" +
val.toString() + "> key sequence: " + key.getLogSeqNum() +
" max sequence: " + maxSeqID);
}
skippedEdits++;
continue;
}
if (skippedEdits > 0 && LOG.isDebugEnabled()) {
LOG.debug("Skipped " + skippedEdits +
" edits because sequence id <= " + maxSeqID);
}
// Check this edit is for me. Also, guard against writing
// METACOLUMN info such as HBASE::CACHEFLUSH entries
Text column = val.getColumn();
@ -977,119 +989,88 @@ class HStore implements HConstants {
* @return true if compaction completed successfully
*/
boolean compact() throws IOException {
long maxId = -1;
synchronized (compactLock) {
Path curCompactStore = HStoreFile.getHStoreDir(this.compactionDir,
encodedRegionName, familyName);
Path curCompactStore = getCompactionDir();
if (LOG.isDebugEnabled()) {
LOG.debug("started compaction of " + storefiles.size() + " files in " +
curCompactStore.toString());
LOG.debug("started compaction of " + storefiles.size() +
" files using " + curCompactStore.toString());
}
if (this.fs.exists(curCompactStore)) {
LOG.warn("Cleaning up a previous incomplete compaction at " +
curCompactStore.toString());
if (!this.fs.delete(curCompactStore)) {
LOG.warn("Deleted returned false on " + curCompactStore.toString());
// Clean out its content in prep. for this new compaction. Has either
// aborted previous compaction or it has content of a previous
// compaction.
Path [] toRemove = this.fs.listPaths(new Path [] {curCompactStore});
for (int i = 0; i < toRemove.length; i++) {
this.fs.delete(toRemove[i]);
}
}
// Storefiles are keyed by sequence id. The oldest file comes first.
// We need to return out of here a List that has the newest file first.
List<HStoreFile> filesToCompact =
new ArrayList<HStoreFile>(this.storefiles.values());
Collections.reverse(filesToCompact);
if (filesToCompact.size() < 1 ||
(filesToCompact.size() == 1 && !filesToCompact.get(0).isReference())) {
if (LOG.isDebugEnabled()) {
LOG.debug("nothing to compact for " + this.storeName);
}
return false;
}
if (!fs.exists(curCompactStore) && !fs.mkdirs(curCompactStore)) {
LOG.warn("Mkdir on " + curCompactStore.toString() + " failed");
return false;
}
// Step through them, writing to the brand-new TreeMap
HStoreFile compactedOutputFile = new HStoreFile(conf, this.compactionDir,
encodedRegionName, familyName, -1);
MapFile.Writer compactedOut = compactedOutputFile.getWriter(this.fs,
this.compression, this.bloomFilter);
try {
// Storefiles are keyed by sequence id. The oldest file comes first.
// We need to return out of here a List that has the newest file as
// first.
List<HStoreFile> filesToCompact =
new ArrayList<HStoreFile>(this.storefiles.values());
Collections.reverse(filesToCompact);
HStoreFile compactedOutputFile = new HStoreFile(conf,
this.compactionDir, encodedRegionName, familyName, -1);
if (filesToCompact.size() < 1 ||
(filesToCompact.size() == 1 &&
!filesToCompact.get(0).isReference())) {
if (LOG.isDebugEnabled()) {
LOG.debug("nothing to compact for " + this.storeName);
}
return false;
}
if (!fs.mkdirs(curCompactStore)) {
LOG.warn("Mkdir on " + curCompactStore.toString() + " failed");
}
// Compute the max-sequenceID seen in any of the to-be-compacted
// TreeMaps if it hasn't been passed in to us.
if (maxId == -1) {
for (HStoreFile hsf: filesToCompact) {
long seqid = hsf.loadInfo(fs);
if (seqid > 0) {
if (seqid > maxId) {
maxId = seqid;
}
}
}
}
// Step through them, writing to the brand-new TreeMap
MapFile.Writer compactedOut = compactedOutputFile.getWriter(this.fs,
this.compression, this.bloomFilter);
try {
compactHStoreFiles(compactedOut, filesToCompact);
} finally {
compactedOut.close();
}
// Now, write out an HSTORE_LOGINFOFILE for the brand-new TreeMap.
if (maxId >= 0) {
compactedOutputFile.writeInfo(fs, maxId);
} else {
compactedOutputFile.writeInfo(fs, -1);
}
// Write out a list of data files that we're replacing
Path filesToReplace = new Path(curCompactStore, COMPACTION_TO_REPLACE);
DataOutputStream out = new DataOutputStream(fs.create(filesToReplace));
try {
out.writeInt(filesToCompact.size());
for (HStoreFile hsf: filesToCompact) {
hsf.write(out);
}
} finally {
out.close();
}
// Indicate that we're done.
Path doneFile = new Path(curCompactStore, COMPACTION_DONE);
(new DataOutputStream(fs.create(doneFile))).close();
// Move the compaction into place.
completeCompaction();
return true;
compactHStoreFiles(compactedOut, filesToCompact);
} finally {
// Clean up the parent -- the region dir in the compactions directory.
if (this.fs.exists(curCompactStore.getParent())) {
if (!this.fs.delete(curCompactStore.getParent())) {
LOG.warn("Delete returned false deleting " +
curCompactStore.getParent().toString());
}
}
compactedOut.close();
}
// Now, write out an HSTORE_LOGINFOFILE for the brand-new TreeMap.
// Compute max-sequenceID seen in any of the to-be-compacted TreeMaps.
long maxId = getMaxSequenceId(filesToCompact);
compactedOutputFile.writeInfo(fs, maxId);
// Write out a list of data files that we're replacing
Path filesToReplace = new Path(curCompactStore, COMPACTION_TO_REPLACE);
DataOutputStream out = new DataOutputStream(fs.create(filesToReplace));
try {
out.writeInt(filesToCompact.size());
for (HStoreFile hsf : filesToCompact) {
hsf.write(out);
}
} finally {
out.close();
}
// Indicate that we're done.
Path doneFile = new Path(curCompactStore, COMPACTION_DONE);
(new DataOutputStream(fs.create(doneFile))).close();
// Move the compaction into place.
completeCompaction(curCompactStore);
return true;
}
}
/*
* Compact passed <code>toCompactFiles</code> into <code>compactedOut</code>.
* We create a new set of MapFile.Reader objects so we don't screw up
* the caching associated with the currently-loaded ones. Our
* iteration-based access pattern is practically designed to ruin
* the cache.
*
* We work by opening a single MapFile.Reader for each file, and
* iterating through them in parallel. We always increment the
* lowest-ranked one. Updates to a single row/column will appear
* ranked by timestamp. This allows us to throw out deleted values or
* obsolete versions.
* @param compactedOut
* @param toCompactFiles
* @throws IOException
* Compact passed <code>toCompactFiles</code> into <code>compactedOut</code>.
* We create a new set of MapFile.Reader objects so we don't screw up the
* caching associated with the currently-loaded ones. Our iteration-based
* access pattern is practically designed to ruin the cache.
*
* We work by opening a single MapFile.Reader for each file, and iterating
* through them in parallel. We always increment the lowest-ranked one.
* Updates to a single row/column will appear ranked by timestamp. This allows
* us to throw out deleted values or obsolete versions. @param compactedOut
* @param toCompactFiles @throws IOException
*/
private void compactHStoreFiles(final MapFile.Writer compactedOut,
final List<HStoreFile> toCompactFiles) throws IOException {
@ -1107,6 +1088,7 @@ class HStore implements HConstants {
// culprit.
LOG.warn("Failed with " + e.toString() + ": " + hsf.toString() +
(hsf.isReference()? " " + hsf.getReference().toString(): ""));
closeCompactionReaders(rdrs);
throw e;
}
}
@ -1195,13 +1177,17 @@ class HStore implements HConstants {
}
}
} finally {
for (int i = 0; i < rdrs.length; i++) {
if (rdrs[i] != null) {
try {
rdrs[i].close();
} catch (IOException e) {
LOG.warn("Exception closing reader", e);
}
closeCompactionReaders(rdrs);
}
}
private void closeCompactionReaders(final CompactionReader [] rdrs) {
for (int i = 0; i < rdrs.length; i++) {
if (rdrs[i] != null) {
try {
rdrs[i].close();
} catch (IOException e) {
LOG.warn("Exception closing reader", e);
}
}
}
@ -1326,11 +1312,11 @@ class HStore implements HConstants {
* 8) Releasing the write-lock
* 9) Allow new scanners to proceed.
* </pre>
*
* @param curCompactStore Compaction to complete.
*/
private void completeCompaction() throws IOException {
Path curCompactStore = HStoreFile.getHStoreDir(this.compactionDir,
encodedRegionName, familyName);
private void completeCompaction(final Path curCompactStore)
throws IOException {
// 1. Wait for active scanners to exit
newScannerLock.writeLock().lock(); // prevent new scanners
try {
@ -1346,6 +1332,7 @@ class HStore implements HConstants {
// 2. Acquiring the HStore write-lock
this.lock.writeLock().lock();
}
try {
Path doneFile = new Path(curCompactStore, COMPACTION_DONE);
if (!fs.exists(doneFile)) {
@ -1366,7 +1353,6 @@ class HStore implements HConstants {
hsf.readFields(in);
toCompactFiles.add(hsf);
}
} finally {
in.close();
}
@ -1412,13 +1398,13 @@ class HStore implements HConstants {
// 7. Loading the new TreeMap.
Long orderVal = Long.valueOf(finalCompactedFile.loadInfo(fs));
this.readers.put(orderVal,
finalCompactedFile.getReader(this.fs, this.bloomFilter));
finalCompactedFile.getReader(this.fs, this.bloomFilter));
this.storefiles.put(orderVal, finalCompactedFile);
} catch (IOException e) {
LOG.error("Failed replacing compacted files. Compacted file is " +
finalCompactedFile.toString() + ". Files replaced are " +
toCompactFiles.toString() +
" some of which may have been already removed", e);
finalCompactedFile.toString() + ". Files replaced are " +
toCompactFiles.toString() +
" some of which may have been already removed", e);
}
} finally {
// 8. Releasing the write-lock
@ -1479,6 +1465,17 @@ class HStore implements HConstants {
}
}
/*
* @return Path to the compaction directory for this column family.
* Compaction dir is a subdirectory of the region. Needs to have the
* same regiondir/storefamily path prefix; HStoreFile constructor presumes
* it (TODO: Fix).
*/
private Path getCompactionDir() {
return HStoreFile.getHStoreDir(this.compactionDir,
this.encodedRegionName, this.familyName);
}
private MapFile.Reader [] getReaders() {
return this.readers.values().
toArray(new MapFile.Reader[this.readers.size()]);