org.apache.tephra.hbase.txprune

Class DataJanitorState

java.lang.Object

org.apache.tephra.hbase.txprune.DataJanitorState

public class DataJanitorState
extends Object

Persist data janitor state into an HBase table. This is used by both TransactionProcessor and by the HBaseTransactionPruningPlugin to persist and read the compaction state.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static interface	DataJanitorState.TableSupplier Supplies table for persisting state

Field Summary

Fields

Modifier and Type	Field and Description
static byte[]	FAMILY
static byte[]	PRUNE_UPPER_BOUND_COL

Constructor Summary

Constructors

Constructor and Description
DataJanitorState(DataJanitorState.TableSupplier stateTableSupplier)

Method Summary

Modifier and Type	Method and Description
void	deleteAllRegionsOnOrBeforeTime(long time) Delete all the regions that were recorded for all times equal or less than the given time.
void	deleteEmptyRegionsOnOrBeforeTime(long time) Delete empty region records saved on or before the given time.
void	deleteInactiveTransactionBoundsOnOrBeforeTime(long time) Delete all inactive transaction bounds recorded for a time less than the given time
void	deletePruneUpperBounds(long deletionPruneUpperBound, SortedSet<byte[]> excludeRegions) Delete prune upper bounds for the regions that are not in the given exclude set, and the prune upper bound is less than the given value.
SortedSet<byte[]>	getEmptyRegionsAfterTime(long time, SortedSet<byte[]> includeRegions) Return regions that were recorded as empty after the given time.
long	getInactiveTransactionBoundForTime(long time) Return inactive transaction bound for the given time.
RegionPruneInfo	getPruneInfoForRegion(byte[] regionId) Get the latest RegionPruneInfo for a given region.
List<RegionPruneInfo>	getPruneInfoForRegions(SortedSet<byte[]> regions) Gets a list of RegionPruneInfo for given regions.
long	getPruneUpperBoundForRegion(byte[] regionId) Get latest prune upper bound for a given region.
Map<byte[],Long>	getPruneUpperBoundForRegions(SortedSet<byte[]> regions) Get latest prune upper bounds for given regions.
TimeRegions	getRegionsOnOrBeforeTime(long time) Return the set of regions saved for the time at or before the given time.
void	saveEmptyRegionForTime(long time, byte[] regionId) Save the given region as empty as of the given time.
void	saveInactiveTransactionBoundForTime(long time, long inactiveTransactionBound) Persist inactive transaction bound for a given time.
void	savePruneUpperBoundForRegion(byte[] regionId, long pruneUpperBound) Persist the latest prune upper bound for a given region.
void	saveRegionsForTime(long time, Set<byte[]> regions) Persist the regions for the given time.

Methods inherited from class java.lang.Object

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

Field Detail

FAMILY

public static final byte[] FAMILY

PRUNE_UPPER_BOUND_COL

public static final byte[] PRUNE_UPPER_BOUND_COL

Constructor Detail

DataJanitorState

public DataJanitorState(DataJanitorState.TableSupplier stateTableSupplier)

Method Detail

savePruneUpperBoundForRegion

public void savePruneUpperBoundForRegion(byte[] regionId,
                                         long pruneUpperBound)
                                  throws IOException

Persist the latest prune upper bound for a given region. This is called by TransactionProcessor after major compaction.

Parameters:

regionId - region id

pruneUpperBound - the latest prune upper bound for the region

Throws:

IOException - when not able to persist the data to HBase

getPruneUpperBoundForRegion

public long getPruneUpperBoundForRegion(byte[] regionId)
                                 throws IOException

Get latest prune upper bound for a given region. This indicates the largest invalid transaction that no longer has writes in this region.

Parameters:

regionId - region id

Returns:

latest prune upper bound for the region

Throws:

IOException - when not able to read the data from HBase

getPruneInfoForRegion

@Nullable
public RegionPruneInfo getPruneInfoForRegion(byte[] regionId)
                                                throws IOException

Get the latest RegionPruneInfo for a given region.

Parameters:

regionId - region id

Returns:

RegionPruneInfo for the region

Throws:

IOException - when not able to read the data from HBase

getPruneUpperBoundForRegions

public Map<byte[],Long> getPruneUpperBoundForRegions(SortedSet<byte[]> regions)
                                              throws IOException

Get latest prune upper bounds for given regions. This is a batch operation of method getPruneUpperBoundForRegion(byte[])

Parameters:

regions - a set of regions

Returns:

a map containing region id and its latest prune upper bound value

Throws:

IOException - when not able to read the data from HBase

getPruneInfoForRegions

public List<RegionPruneInfo> getPruneInfoForRegions(@Nullable
                                                    SortedSet<byte[]> regions)
                                             throws IOException

Gets a list of RegionPruneInfo for given regions. Returns all regions if the given regions set is null.

Parameters:

regions - a set of regions

Returns:

list of RegionPruneInfos.

Throws:

IOException - when not able to read the data from HBase

deletePruneUpperBounds

public void deletePruneUpperBounds(long deletionPruneUpperBound,
                                   SortedSet<byte[]> excludeRegions)
                            throws IOException

Delete prune upper bounds for the regions that are not in the given exclude set, and the prune upper bound is less than the given value. After the invalid list is pruned up to deletionPruneUpperBound, we do not need entries for regions that have prune upper bound less than deletionPruneUpperBound. We however limit the deletion to only regions that are no longer in existence (due to deletion, etc.), to avoid update/delete race conditions.

Parameters:

deletionPruneUpperBound - prune upper bound below which regions will be deleted

excludeRegions - set of regions that should not be deleted

Throws:

IOException - when not able to delete data in HBase

saveRegionsForTime

public void saveRegionsForTime(long time,
                               Set<byte[]> regions)
                        throws IOException

Persist the regions for the given time. HBaseTransactionPruningPlugin saves the set of transactional regions existing in the HBase instance periodically.

Parameters:

time - timestamp in milliseconds

regions - set of regions at the time

Throws:

IOException - when not able to persist the data to HBase

getRegionsOnOrBeforeTime

@Nullable
public TimeRegions getRegionsOnOrBeforeTime(long time)
                                               throws IOException

Return the set of regions saved for the time at or before the given time. This method finds the greatest time that is less than or equal to the given time, and then returns all regions with that exact time, but none that are older than that.

Parameters:

time - timestamp in milliseconds

Returns:

set of regions and time at which they were recorded, or null if no regions found

Throws:

IOException - when not able to read the data from HBase

deleteAllRegionsOnOrBeforeTime

public void deleteAllRegionsOnOrBeforeTime(long time)
                                    throws IOException

Delete all the regions that were recorded for all times equal or less than the given time.

Parameters:

time - timestamp in milliseconds

Throws:

IOException - when not able to delete data in HBase

saveInactiveTransactionBoundForTime

public void saveInactiveTransactionBoundForTime(long time,
                                                long inactiveTransactionBound)
                                         throws IOException

Persist inactive transaction bound for a given time. This is the smallest not in-progress transaction that will not have writes in any HBase regions that are created after the given time.

Parameters:

time - time in milliseconds

inactiveTransactionBound - inactive transaction bound for the given time

Throws:

IOException - when not able to persist the data to HBase

getInactiveTransactionBoundForTime

public long getInactiveTransactionBoundForTime(long time)
                                        throws IOException

Return inactive transaction bound for the given time.

Parameters:

time - time in milliseconds

Returns:

inactive transaction bound for the given time

Throws:

IOException - when not able to read the data from HBase

deleteInactiveTransactionBoundsOnOrBeforeTime

public void deleteInactiveTransactionBoundsOnOrBeforeTime(long time)
                                                   throws IOException

Delete all inactive transaction bounds recorded for a time less than the given time

Parameters:

time - time in milliseconds

Throws:

IOException - when not able to delete data in HBase

saveEmptyRegionForTime

public void saveEmptyRegionForTime(long time,
                                   byte[] regionId)
                            throws IOException

Save the given region as empty as of the given time.

Parameters:

time - time in milliseconds

regionId - region id

Throws:

IOException

getEmptyRegionsAfterTime

public SortedSet<byte[]> getEmptyRegionsAfterTime(long time,
                                                  @Nullable
                                                  SortedSet<byte[]> includeRegions)
                                           throws IOException

Return regions that were recorded as empty after the given time.

Parameters:

time - time in milliseconds

includeRegions - If not null, the returned set will be an intersection of the includeRegions set and the empty regions after the given time

Throws:

IOException

deleteEmptyRegionsOnOrBeforeTime

public void deleteEmptyRegionsOnOrBeforeTime(long time)
                                      throws IOException

Delete empty region records saved on or before the given time.

Parameters:

time - time in milliseconds

Throws:

IOException