org.apache.jackrabbit.oak.plugins.blob

Class MarkSweepGarbageCollector

java.lang.Object

org.apache.jackrabbit.oak.plugins.blob.MarkSweepGarbageCollector

All Implemented Interfaces:

BlobGarbageCollector

public class MarkSweepGarbageCollector
extends Object
implements BlobGarbageCollector

Mark and sweep garbage collector. Uses the file system to store internal state while in process to account for huge data. This class is not thread safe.

Field Summary

Fields

Modifier and Type	Field and Description
static int	DEFAULT_BATCH_COUNT
static String	DELIM
static org.slf4j.Logger	LOG
static String	TEMP_DIR

Constructor Summary

Constructors

Constructor and Description
MarkSweepGarbageCollector(BlobReferenceRetriever marker, GarbageCollectableBlobStore blobStore, Executor executor, long maxLastModifiedInterval, String repositoryId, Whiteboard whiteboard, StatisticsProvider statisticsProvider) Instantiates a new blob garbage collector.
MarkSweepGarbageCollector(BlobReferenceRetriever marker, GarbageCollectableBlobStore blobStore, Executor executor, String root, int batchCount, long maxLastModifiedInterval, String repositoryId)
MarkSweepGarbageCollector(BlobReferenceRetriever marker, GarbageCollectableBlobStore blobStore, Executor executor, String root, int batchCount, long maxLastModifiedInterval, String repositoryId, Whiteboard whiteboard, StatisticsProvider statisticsProvider) Creates an instance of MarkSweepGarbageCollector

Method Summary

Modifier and Type	Method and Description
long	checkConsistency() Checks for the DataStore consistency and reports the number of missing blobs still referenced.
void	collectGarbage(boolean markOnly) Marks garbage blobs from the passed node store instance.
void	collectGarbage(boolean markOnly, boolean forceBlobRetrieve) Marks garbage blobs from the passed node store instance.
OperationsStatsMBean	getOperationStats() Returns operation statistics
List<GarbageCollectionRepoStats>	getStats() Returns the stats related to GC for all repos
protected void	iterateNodeTree(GarbageCollectorFileState fs, boolean logPath) Iterates the complete node tree and collect all blob references
protected void	mark(GarbageCollectorFileState fs) Mark phase of the GC.
protected void	markAndSweep(boolean markOnly, boolean forceBlobRetrieve) Mark and sweep.
protected long	sweep(GarbageCollectorFileState fs, long markStart, boolean forceBlobRetrieve) Sweep phase of gc candidate deletion.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

LOG

public static final org.slf4j.Logger LOG

TEMP_DIR

public static final String TEMP_DIR

DEFAULT_BATCH_COUNT

public static final int DEFAULT_BATCH_COUNT

See Also:

Constant Field Values

DELIM

public static final String DELIM

See Also:

Constant Field Values

Constructor Detail

MarkSweepGarbageCollector

public MarkSweepGarbageCollector(BlobReferenceRetriever marker,
                                 GarbageCollectableBlobStore blobStore,
                                 Executor executor,
                                 String root,
                                 int batchCount,
                                 long maxLastModifiedInterval,
                                 @Nullable
                                 String repositoryId,
                                 @Nullable
                                 Whiteboard whiteboard,
                                 @Nullable
                                 StatisticsProvider statisticsProvider)
                          throws IOException

Creates an instance of MarkSweepGarbageCollector

Parameters:

marker - BlobReferenceRetriever instanced used to fetch refereed blob entries

blobStore - the blob store instance

executor - executor

root - the root absolute path of directory under which temporary files would be created

batchCount - batch sized used for saving intermediate state

maxLastModifiedInterval - lastModifiedTime in millis. Only files with time less than this time would be considered for GC

repositoryId - unique repository id for this node

whiteboard - whiteboard instance

statisticsProvider - statistics provider instance

Throws:

IOException

MarkSweepGarbageCollector

public MarkSweepGarbageCollector(BlobReferenceRetriever marker,
                                 GarbageCollectableBlobStore blobStore,
                                 Executor executor,
                                 String root,
                                 int batchCount,
                                 long maxLastModifiedInterval,
                                 @Nullable
                                 String repositoryId)
                          throws IOException

Throws:

IOException

MarkSweepGarbageCollector

public MarkSweepGarbageCollector(BlobReferenceRetriever marker,
                                 GarbageCollectableBlobStore blobStore,
                                 Executor executor,
                                 long maxLastModifiedInterval,
                                 @Nullable
                                 String repositoryId,
                                 @Nullable
                                 Whiteboard whiteboard,
                                 @Nullable
                                 StatisticsProvider statisticsProvider)
                          throws IOException

Instantiates a new blob garbage collector.

Throws:

IOException

Method Detail

collectGarbage

public void collectGarbage(boolean markOnly)
                    throws Exception

Description copied from interface: BlobGarbageCollector

Marks garbage blobs from the passed node store instance. Collects them only if markOnly is false.

Specified by:

collectGarbage in interface BlobGarbageCollector

Parameters:

markOnly - whether to only mark references and not sweep in the mark and sweep operation.

Throws:

Exception - the exception

collectGarbage

public void collectGarbage(boolean markOnly,
                           boolean forceBlobRetrieve)
                    throws Exception

Description copied from interface: BlobGarbageCollector

Marks garbage blobs from the passed node store instance. Collects them only if markOnly is false. Also forces retrieval of blob ids from the blob store rather than using any local tracking.

Specified by:

collectGarbage in interface BlobGarbageCollector

Parameters:

markOnly - whether to only mark references and not sweep in the mark and sweep operation.

forceBlobRetrieve - whether to force retrieve of blob ids from datastore

Throws:

Exception

getStats

public List<GarbageCollectionRepoStats> getStats()
                                          throws Exception

Returns the stats related to GC for all repos

Specified by:

getStats in interface BlobGarbageCollector

Returns:

a list of GarbageCollectionRepoStats objects

Throws:

Exception

getOperationStats

public OperationsStatsMBean getOperationStats()

Description copied from interface: BlobGarbageCollector

Returns operation statistics

Specified by:

getOperationStats in interface BlobGarbageCollector

Returns:

stats object

markAndSweep

protected void markAndSweep(boolean markOnly,
                            boolean forceBlobRetrieve)
                     throws Exception

Mark and sweep. Main entry method for GC.

Parameters:

markOnly - whether to mark only

forceBlobRetrieve - force retrieve blob ids

Throws:

Exception - the exception

mark

protected void mark(GarbageCollectorFileState fs)
             throws IOException,
                    DataStoreException

Mark phase of the GC.

Parameters:

fs - the garbage collector file state

Throws:

IOException

DataStoreException

sweep

protected long sweep(GarbageCollectorFileState fs,
                     long markStart,
                     boolean forceBlobRetrieve)
              throws Exception

Sweep phase of gc candidate deletion.

Performs the following steps depending upon the type of the blob store refer SharedDataStore.Type:

• Shared
• • Merge all marked references (from the mark phase run independently) available in the data store meta store (from all configured independent repositories).
  • Retrieve all blob ids available.
  • Diffs the 2 sets above to retrieve list of blob ids not used.
  • Deletes only blobs created after (earliest time stamp of the marked references - #maxLastModifiedInterval) from the above set.
• Default
• • Mark phase already run.
  • Retrieve all blob ids available.
  • Diffs the 2 sets above to retrieve list of blob ids not used.
  • Deletes only blobs created after (time stamp of the marked references - #maxLastModifiedInterval).

Parameters:

fs - the garbage collector file state

markStart - the start time of mark to take as reference for deletion

forceBlobRetrieve -

Returns:

the number of blobs deleted

Throws:

Exception - the exception

iterateNodeTree

protected void iterateNodeTree(GarbageCollectorFileState fs,
                               boolean logPath)
                        throws IOException

Iterates the complete node tree and collect all blob references

Parameters:

fs - the garbage collector file state

logPath - whether to log path in the file or not

Throws:

IOException

checkConsistency

public long checkConsistency()
                      throws Exception

Checks for the DataStore consistency and reports the number of missing blobs still referenced.

Specified by:

checkConsistency in interface BlobGarbageCollector

Returns:

the missing blobs

Throws:

Exception