com.sleepycat.je.rep.stream

Class FeederReader

java.lang.Object

com.sleepycat.je.log.FileReader

com.sleepycat.je.rep.stream.FeederReader

public class FeederReader
extends FileReader

The FeederReader is a flavor of VLSNReader which supports replication stream feeding. It assumes that reading will always go forward in the log. Special features are: - The reader can read either from a log buffer or from the file. Sometimes log entries are logged but are not yet available on disk. In general, it's better to read from the log buffers rather then the file. - The reader can block for a given time period, waiting for the next vlsn to appear

Nested Class Summary

Nested classes/interfaces inherited from class com.sleepycat.je.log.FileReader

FileReader.EOFException, FileReader.ReadWindow

Field Summary

Fields inherited from class com.sleepycat.je.log.FileReader

cksumValidator, currentEntryHeader, currentEntryOffset, currentEntryPrevOffset, envImpl, eof, fileManager, forward, logger, nextEntryOffset, startLsn, window

Constructor Summary

Constructors

Constructor and Description
FeederReader(RepImpl envImpl, VLSNIndex vlsnIndex, long startLsn, int readBufferSize)

Method Summary

Modifier and Type	Method and Description
long	initScan(VLSN startVLSN) Set up the FeederReader to start scanning from this VLSN.
protected boolean	isTargetEntry() Return true if this entry is replicated and its VLSN is currentVLSN.
protected FileReader.ReadWindow	makeWindow(int readBufferSize) Use a ReadWindow which can read from LogBuffers as well as the physical file.
protected boolean	processEntry(ByteBuffer entryBuffer) Instantiate a WireRecord to house this log entry.
OutputWireRecord	scanForwards(VLSN vlsn, int waitTime) Forward scanning for feeding the replica: get the log record for this VLSN.

Methods inherited from class com.sleepycat.je.log.FileReader

entryIsReplicated, getAndResetNReads, getLastEntrySize, getLastLsn, getLastVlsn, getLogVersion, getNRepeatIteratorReads, getNumRead, handleGapInBackwardsScan, initStartingPosition, readNextEntry, readNextEntryAllowExceptions, setAlwaysValidateChecksum, setBackwardPosition, setForwardPosition, skipData, skipEntry

Methods inherited from class java.lang.Object

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

Constructor Detail

FeederReader

public FeederReader(RepImpl envImpl,
                    VLSNIndex vlsnIndex,
                    long startLsn,
                    int readBufferSize)

Method Detail

makeWindow

protected FileReader.ReadWindow makeWindow(int readBufferSize)

Use a ReadWindow which can read from LogBuffers as well as the physical file.

Overrides:

makeWindow in class FileReader

Throws:

DatabaseException

initScan

public long initScan(VLSN startVLSN)
              throws IOException

Set up the FeederReader to start scanning from this VLSN. If we find a mapping for this VLSN, we'll start precisely at its LSN, else we'll have to start from an earlier location. This initialization can't be done in the constructor, because the Feeder is set up to require the construction of the source before we know where to start.

Returns:

the startLsn

Throws:

IOException

scanForwards

public OutputWireRecord scanForwards(VLSN vlsn,
                                     int waitTime)
                              throws InterruptedException

Forward scanning for feeding the replica: get the log record for this VLSN. If the log record hasn't been created yet, wait for a period specified by "waitTime". Where possible, the FeederReader fetches the log record from the cache within the VLSNIndex. (See the VLSNIndex for a description of this two level cache). If the requested VLSN is not available from the cache, the reader fetches the item from the JE log -- either from the log buffers or from disk. The FeederReader is like a cursor on the log, and retains a position in the log. When there are log item cache hits, the FeederReader's position can fall behind, because it is being bypassed. It is possible for log cleaning to take place between the point of the FeederReader's stale position and the end of the log. If so, the FeederReader must not attempt to scan from its current position, because it might run afoul of gap created by the cleaned and delete log files. When there have been log item cache hits, the FeederReader must jump its position forward using the vlsnIndex mappings to safely skip over any cleaned gaps in the log.

Throws:

InterruptedException

isTargetEntry

protected boolean isTargetEntry()

Return true if this entry is replicated and its VLSN is currentVLSN.

Overrides:

isTargetEntry in class FileReader

Returns:

true if this reader should process the current logrec further, via the processEntry() method. A logrec must be passed to processEntry if the full logrec (not just the header) must be de-serialized for further processing. Return false if no further processing is needed, in which case the current logrec will be skipped (i.e, not returned to the caller of readNextEntry().

processEntry

protected boolean processEntry(ByteBuffer entryBuffer)

Instantiate a WireRecord to house this log entry.

Specified by:

processEntry in class FileReader

Parameters:

entryBuffer - A ByteBuffer that the logrec data and is positioned at the start of the logrec body (i.e., just after the logrec header).

Returns:

true if this entry should be returned to the caller of readNextEntry().