public void dump(OutputStream outputStream, IndexWriter indexWriter, Lock commitLock)
throws IOException {
IndexCommit indexCommit = null;
String segmentsFileName = null;
// Lock up to stop external commits while recording dump IndexCommit
commitLock.lock();
try {
// Commit all pending changes to ensure dumped index is integrated
indexWriter.commit();
// Record IndexCommit, then release the lock to other dumpers and
// committers to move on
indexCommit = _lastIndexCommit;
segmentsFileName = indexCommit.getSegmentsFileName();
_dumpingSegmensFiletNames.add(segmentsFileName);
} finally {
commitLock.unlock();
}
try {
IndexCommitSerializationUtil.serializeIndex(indexCommit, outputStream);
} finally {
// Clear the dumping segments file name, so the old index can be
// removed.
_dumpingSegmensFiletNames.remove(segmentsFileName);
}
}
@Override
public void delete() {
Long gen = delegate.getGeneration();
Long reserve = reserves.get(gen);
if (reserve != null && System.currentTimeMillis() < reserve) return;
if (savedCommits.containsKey(gen)) return;
delegate.delete();
}
public void onCommit(List<? extends IndexCommit> list) throws IOException {
_lastIndexCommit = list.get(list.size() - 1);
for (int i = 0; i < list.size() - 1; i++) {
IndexCommit indexCommit = list.get(i);
if (!_dumpingSegmensFiletNames.contains(indexCommit.getSegmentsFileName())) {
indexCommit.delete();
}
}
}
LuceneSnapshotIterator(
File indexDirectory, IndexCommit snapshotPoint, SnapshotDeletionPolicy deletionPolicy)
throws IOException {
this.indexDirectory = indexDirectory;
this.deletionPolicy = deletionPolicy;
this.fileNames = snapshotPoint.getFileNames().iterator();
}
public static long getCommitTimestamp(IndexCommit commit) throws IOException {
final Map<String, String> commitData = commit.getUserData();
String commitTime = commitData.get(SolrIndexWriter.COMMIT_TIME_MSEC_KEY);
if (commitTime != null) {
return Long.parseLong(commitTime);
} else {
return 0;
}
}
private String getId(IndexCommit commit) {
StringBuilder sb = new StringBuilder();
Directory dir = commit.getDirectory();
// For anything persistent, make something that will
// be the same, regardless of the Directory instance.
if (dir instanceof FSDirectory) {
FSDirectory fsd = (FSDirectory) dir;
File fdir = fsd.getDirectory();
sb.append(fdir.getPath());
} else {
sb.append(dir);
}
sb.append('/');
sb.append(commit.getGeneration());
sb.append('_');
sb.append(commit.getVersion());
return sb.toString();
}
private void updateCommits(List<IndexCommit> commits) {
// to be safe, we should only call delete on a commit point passed to us
// in this specific call (may be across diff IndexWriter instances).
// this will happen rarely, so just synchronize everything
// for safety and to avoid race conditions
synchronized (this) {
long maxCommitAgeTimeStamp = -1L;
IndexCommit newest = commits.get(commits.size() - 1);
log.info("newest commit = " + newest.getVersion());
int singleSegKept = (newest.getSegmentCount() == 1) ? 1 : 0;
int totalKept = 1;
// work our way from newest to oldest, skipping the first since we always want to keep it.
for (int i = commits.size() - 2; i >= 0; i--) {
IndexCommit commit = commits.get(i);
// delete anything too old, regardless of other policies
try {
if (maxCommitAge != null) {
if (maxCommitAgeTimeStamp == -1) {
DateMathParser dmp = new DateMathParser(DateField.UTC, Locale.US);
maxCommitAgeTimeStamp = dmp.parseMath(maxCommitAge).getTime();
}
if (commit.getTimestamp() < maxCommitAgeTimeStamp) {
commit.delete();
continue;
}
}
} catch (Exception e) {
log.warn("Exception while checking commit point's age for deletion", e);
}
if (singleSegKept < maxOptimizedCommitsToKeep && commit.getSegmentCount() == 1) {
totalKept++;
singleSegKept++;
continue;
}
if (totalKept < maxCommitsToKeep) {
totalKept++;
continue;
}
commit.delete();
}
} // end synchronized
}
static String str(IndexCommit commit) {
StringBuilder sb = new StringBuilder();
try {
sb.append("commit{");
Directory dir = commit.getDirectory();
if (dir instanceof FSDirectory) {
FSDirectory fsd = (FSDirectory) dir;
sb.append("dir=").append(fsd.getDirectory());
} else {
sb.append("dir=").append(dir);
}
sb.append(",segFN=").append(commit.getSegmentsFileName());
sb.append(",version=").append(commit.getVersion());
sb.append(",generation=").append(commit.getGeneration());
sb.append(",filenames=").append(commit.getFileNames());
} catch (Exception e) {
sb.append(e);
}
return sb.toString();
}
@Override
public String getSegmentsFileName() {
return delegate.getSegmentsFileName();
}
@Override
public Map getUserData() throws IOException {
return delegate.getUserData();
}
@Override
public boolean isDeleted() {
return delegate.isDeleted();
}
@Override
public long getGeneration() {
return delegate.getGeneration();
}
@Override
public int hashCode() {
return delegate.hashCode();
}
@Override
public boolean equals(Object o) {
return delegate.equals(o);
}
@Override
public int getSegmentCount() {
return delegate.getSegmentCount();
}
@Override
public Directory getDirectory() {
return delegate.getDirectory();
}
@Override
public Collection getFileNames() throws IOException {
return delegate.getFileNames();
}
/**
* Perform phase1 of the recovery operations. Once this {@link IndexCommit} snapshot has been
* performed no commit operations (files being fsync'd) are effectively allowed on this index
* until all recovery phases are done
*
* <p>Phase1 examines the segment files on the target node and copies over the segments that are
* missing. Only segments that have the same size and checksum can be reused
*/
public void phase1(final IndexCommit snapshot, final Translog.View translogView) {
cancellableThreads.checkForCancel();
// Total size of segment files that are recovered
long totalSize = 0;
// Total size of segment files that were able to be re-used
long existingTotalSize = 0;
final Store store = shard.store();
store.incRef();
try {
StopWatch stopWatch = new StopWatch().start();
final Store.MetadataSnapshot recoverySourceMetadata;
try {
recoverySourceMetadata = store.getMetadata(snapshot);
} catch (CorruptIndexException | IndexFormatTooOldException | IndexFormatTooNewException ex) {
shard.failShard("recovery", ex);
throw ex;
}
for (String name : snapshot.getFileNames()) {
final StoreFileMetaData md = recoverySourceMetadata.get(name);
if (md == null) {
logger.info(
"Snapshot differs from actual index for file: {} meta: {}",
name,
recoverySourceMetadata.asMap());
throw new CorruptIndexException(
"Snapshot differs from actual index - maybe index was removed metadata has "
+ recoverySourceMetadata.asMap().size()
+ " files",
name);
}
}
// Generate a "diff" of all the identical, different, and missing
// segment files on the target node, using the existing files on
// the source node
String recoverySourceSyncId = recoverySourceMetadata.getSyncId();
String recoveryTargetSyncId = request.metadataSnapshot().getSyncId();
final boolean recoverWithSyncId =
recoverySourceSyncId != null && recoverySourceSyncId.equals(recoveryTargetSyncId);
if (recoverWithSyncId) {
final long numDocsTarget = request.metadataSnapshot().getNumDocs();
final long numDocsSource = recoverySourceMetadata.getNumDocs();
if (numDocsTarget != numDocsSource) {
throw new IllegalStateException(
"try to recover "
+ request.shardId()
+ " from primary shard with sync id but number "
+ "of docs differ: "
+ numDocsTarget
+ " ("
+ request.sourceNode().getName()
+ ", primary) vs "
+ numDocsSource
+ "("
+ request.targetNode().getName()
+ ")");
}
// we shortcut recovery here because we have nothing to copy. but we must still start the
// engine on the target.
// so we don't return here
logger.trace(
"[{}][{}] skipping [phase1] to {} - identical sync id [{}] found on both source and target",
indexName,
shardId,
request.targetNode(),
recoverySourceSyncId);
} else {
final Store.RecoveryDiff diff =
recoverySourceMetadata.recoveryDiff(request.metadataSnapshot());
for (StoreFileMetaData md : diff.identical) {
response.phase1ExistingFileNames.add(md.name());
response.phase1ExistingFileSizes.add(md.length());
existingTotalSize += md.length();
if (logger.isTraceEnabled()) {
logger.trace(
"[{}][{}] recovery [phase1] to {}: not recovering [{}], exists in local store and has checksum [{}],"
+ " size [{}]",
indexName,
shardId,
request.targetNode(),
md.name(),
md.checksum(),
md.length());
}
totalSize += md.length();
}
List<StoreFileMetaData> phase1Files =
new ArrayList<>(diff.different.size() + diff.missing.size());
phase1Files.addAll(diff.different);
phase1Files.addAll(diff.missing);
for (StoreFileMetaData md : phase1Files) {
if (request.metadataSnapshot().asMap().containsKey(md.name())) {
logger.trace(
"[{}][{}] recovery [phase1] to {}: recovering [{}], exists in local store, but is different: remote "
+ "[{}], local [{}]",
indexName,
shardId,
request.targetNode(),
md.name(),
request.metadataSnapshot().asMap().get(md.name()),
md);
} else {
logger.trace(
"[{}][{}] recovery [phase1] to {}: recovering [{}], does not exists in remote",
indexName,
shardId,
request.targetNode(),
md.name());
}
response.phase1FileNames.add(md.name());
response.phase1FileSizes.add(md.length());
totalSize += md.length();
}
response.phase1TotalSize = totalSize;
response.phase1ExistingTotalSize = existingTotalSize;
logger.trace(
"[{}][{}] recovery [phase1] to {}: recovering_files [{}] with total_size [{}], reusing_files [{}] with "
+ "total_size [{}]",
indexName,
shardId,
request.targetNode(),
response.phase1FileNames.size(),
new ByteSizeValue(totalSize),
response.phase1ExistingFileNames.size(),
new ByteSizeValue(existingTotalSize));
cancellableThreads.execute(
() ->
recoveryTarget.receiveFileInfo(
response.phase1FileNames,
response.phase1FileSizes,
response.phase1ExistingFileNames,
response.phase1ExistingFileSizes,
translogView.totalOperations()));
// How many bytes we've copied since we last called RateLimiter.pause
final Function<StoreFileMetaData, OutputStream> outputStreamFactories =
md ->
new BufferedOutputStream(
new RecoveryOutputStream(md, translogView), chunkSizeInBytes);
sendFiles(
store,
phase1Files.toArray(new StoreFileMetaData[phase1Files.size()]),
outputStreamFactories);
// Send the CLEAN_FILES request, which takes all of the files that
// were transferred and renames them from their temporary file
// names to the actual file names. It also writes checksums for
// the files after they have been renamed.
//
// Once the files have been renamed, any other files that are not
// related to this recovery (out of date segments, for example)
// are deleted
try {
cancellableThreads.executeIO(
() ->
recoveryTarget.cleanFiles(
translogView.totalOperations(), recoverySourceMetadata));
} catch (RemoteTransportException | IOException targetException) {
final IOException corruptIndexException;
// we realized that after the index was copied and we wanted to finalize the recovery
// the index was corrupted:
// - maybe due to a broken segments file on an empty index (transferred with no
// checksum)
// - maybe due to old segments without checksums or length only checks
if ((corruptIndexException = ExceptionsHelper.unwrapCorruption(targetException))
!= null) {
try {
final Store.MetadataSnapshot recoverySourceMetadata1 = store.getMetadata(snapshot);
StoreFileMetaData[] metadata =
StreamSupport.stream(recoverySourceMetadata1.spliterator(), false)
.toArray(size -> new StoreFileMetaData[size]);
ArrayUtil.timSort(
metadata,
(o1, o2) -> {
return Long.compare(o1.length(), o2.length()); // check small files first
});
for (StoreFileMetaData md : metadata) {
cancellableThreads.checkForCancel();
logger.debug(
"{} checking integrity for file {} after remove corruption exception",
shard.shardId(),
md);
if (store.checkIntegrityNoException(md)
== false) { // we are corrupted on the primary -- fail!
shard.failShard("recovery", corruptIndexException);
logger.warn(
"{} Corrupted file detected {} checksum mismatch", shard.shardId(), md);
throw corruptIndexException;
}
}
} catch (IOException ex) {
targetException.addSuppressed(ex);
throw targetException;
}
// corruption has happened on the way to replica
RemoteTransportException exception =
new RemoteTransportException(
"File corruption occurred on recovery but " + "checksums are ok", null);
exception.addSuppressed(targetException);
logger.warn(
(org.apache.logging.log4j.util.Supplier<?>)
() ->
new ParameterizedMessage(
"{} Remote file corruption during finalization of recovery on node {}. local checksum OK",
shard.shardId(),
request.targetNode()),
corruptIndexException);
throw exception;
} else {
throw targetException;
}
}
}
prepareTargetForTranslog(translogView.totalOperations());
logger.trace(
"[{}][{}] recovery [phase1] to {}: took [{}]",
indexName,
shardId,
request.targetNode(),
stopWatch.totalTime());
response.phase1Time = stopWatch.totalTime().millis();
} catch (Exception e) {
throw new RecoverFilesRecoveryException(
request.shardId(), response.phase1FileNames.size(), new ByteSizeValue(totalSize), e);
} finally {
store.decRef();
}
}
@Override
public boolean isOptimized() {
return delegate.isOptimized();
}
@Override
public long getVersion() {
return delegate.getVersion();
}
@Override
public long getTimestamp() throws IOException {
return delegate.getTimestamp();
}