Please wait. This can take some minutes ...
Many resources are needed to download a project. Please understand that we have to compensate our server costs. Thank you in advance.
Project price only 1 $
You can buy this project and download/modify it how often you want.
org.apache.activemq.leveldb.LevelDBClient.scala Maven / Gradle / Ivy
/**
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.activemq.leveldb
import java.{lang=>jl}
import java.{util=>ju}
import java.util.concurrent.locks.ReentrantReadWriteLock
import collection.immutable.TreeMap
import collection.mutable.{HashMap, ListBuffer}
import org.iq80.leveldb._
import org.fusesource.hawtdispatch._
import record.{CollectionKey, EntryKey, EntryRecord, CollectionRecord}
import org.apache.activemq.leveldb.util._
import java.util.concurrent._
import org.fusesource.hawtbuf._
import java.io._
import scala.Option._
import org.apache.activemq.command.{MessageAck, Message}
import org.apache.activemq.util.{IOExceptionSupport, ByteSequence}
import java.text.SimpleDateFormat
import java.util.{Date, Collections}
import org.fusesource.leveldbjni.internal.JniDB
import org.apache.activemq.ActiveMQMessageAuditNoSync
import org.apache.activemq.leveldb.util.TimeMetric
import org.fusesource.hawtbuf.ByteArrayInputStream
import org.apache.activemq.leveldb.RecordLog.LogInfo
import scala.Some
import scala.Serializable
import org.fusesource.hawtbuf.ByteArrayOutputStream
import org.apache.activemq.broker.SuppressReplyException
/**
* @author Hiram Chirino
*/
object LevelDBClient extends Log {
class WriteThread(r:Runnable) extends Thread(r) {
setDaemon(true)
}
final val STORE_SCHEMA_PREFIX = "activemq_leveldb_store:"
final val STORE_SCHEMA_VERSION = 1
final val THREAD_POOL_STACK_SIZE = System.getProperty("leveldb.thread.stack.size", "" + 1024 * 512).toLong
final val THREAD_POOL: ThreadPoolExecutor = new ThreadPoolExecutor(0, Integer.MAX_VALUE, 10, TimeUnit.SECONDS, new SynchronousQueue[Runnable], new ThreadFactory {
def newThread(r: Runnable): Thread = {
var rc: Thread = new Thread(null, r, "LevelDB Store Task", THREAD_POOL_STACK_SIZE)
rc.setDaemon(true)
return rc
}
}) {
override def shutdown: Unit = {}
override def shutdownNow = Collections.emptyList[Runnable]
}
val PLIST_WRITE_OPTIONS = new WriteOptions().sync(false)
final val DIRTY_INDEX_KEY = bytes(":dirty")
final val LOG_REF_INDEX_KEY = bytes(":log-refs")
final val LOGS_INDEX_KEY = bytes(":logs")
final val PRODUCER_IDS_INDEX_KEY = bytes(":producer_ids")
final val COLLECTION_META_KEY = bytes(":collection-meta")
final val TRUE = bytes("true")
final val FALSE = bytes("false")
final val ACK_POSITION = new AsciiBuffer("p")
final val COLLECTION_PREFIX = 'c'.toByte
final val COLLECTION_PREFIX_ARRAY = Array(COLLECTION_PREFIX)
final val ENTRY_PREFIX = 'e'.toByte
final val ENTRY_PREFIX_ARRAY = Array(ENTRY_PREFIX)
final val LOG_ADD_COLLECTION = 1.toByte
final val LOG_REMOVE_COLLECTION = 2.toByte
final val LOG_ADD_ENTRY = 3.toByte
final val LOG_REMOVE_ENTRY = 4.toByte
final val LOG_DATA = 5.toByte
final val LOG_TRACE = 6.toByte
final val LOG_UPDATE_ENTRY = 7.toByte
final val LOG_SUFFIX = ".log"
final val INDEX_SUFFIX = ".index"
implicit def toByteArray(buffer:Buffer) = buffer.toByteArray
implicit def toBuffer(buffer:Array[Byte]) = new Buffer(buffer)
def encodeCollectionRecord(v: CollectionRecord.Buffer) = v.toUnframedByteArray
def decodeCollectionRecord(data: Buffer):CollectionRecord.Buffer = CollectionRecord.FACTORY.parseUnframed(data)
def encodeCollectionKeyRecord(v: CollectionKey.Buffer) = v.toUnframedByteArray
def decodeCollectionKeyRecord(data: Buffer):CollectionKey.Buffer = CollectionKey.FACTORY.parseUnframed(data)
def encodeEntryRecord(v: EntryRecord.Buffer) = v.toUnframedBuffer
def decodeEntryRecord(data: Buffer):EntryRecord.Buffer = EntryRecord.FACTORY.parseUnframed(data)
def encodeEntryKeyRecord(v: EntryKey.Buffer) = v.toUnframedByteArray
def decodeEntryKeyRecord(data: Buffer):EntryKey.Buffer = EntryKey.FACTORY.parseUnframed(data)
def encodeLocator(pos:Long, len:Int):Array[Byte] = {
val out = new DataByteArrayOutputStream(
AbstractVarIntSupport.computeVarLongSize(pos)+
AbstractVarIntSupport.computeVarIntSize(len)
)
out.writeVarLong(pos)
out.writeVarInt(len)
out.getData
}
def decodeLocator(bytes:Buffer):(Long, Int) = {
val in = new DataByteArrayInputStream(bytes)
(in.readVarLong(), in.readVarInt())
}
def decodeLocator(bytes:Array[Byte]):(Long, Int) = {
val in = new DataByteArrayInputStream(bytes)
(in.readVarLong(), in.readVarInt())
}
def encodeLongLong(a1:Long, a2:Long) = {
val out = new DataByteArrayOutputStream(8)
out.writeLong(a1)
out.writeLong(a2)
out.toBuffer
}
def decodeLongLong(bytes:Array[Byte]):(Long, Long) = {
val in = new DataByteArrayInputStream(bytes)
(in.readLong(), in.readLong())
}
def encodeLong(a1:Long) = {
val out = new DataByteArrayOutputStream(8)
out.writeLong(a1)
out.toBuffer
}
def encodeVLong(a1:Long):Array[Byte] = {
val out = new DataByteArrayOutputStream(
AbstractVarIntSupport.computeVarLongSize(a1)
)
out.writeVarLong(a1)
out.getData
}
def decodeVLong(bytes:Array[Byte]):Long = {
val in = new DataByteArrayInputStream(bytes)
in.readVarLong()
}
def encodeLongKey(a1:Byte, a2:Long):Array[Byte] = {
val out = new DataByteArrayOutputStream(9)
out.writeByte(a1.toInt)
out.writeLong(a2)
out.getData
}
def decodeLongKey(bytes:Array[Byte]):(Byte, Long) = {
val in = new DataByteArrayInputStream(bytes)
(in.readByte(), in.readLong())
}
def decodeLong(bytes:Buffer):Long = {
val in = new DataByteArrayInputStream(bytes)
in.readLong()
}
def decodeLong(bytes:Array[Byte]):Long = {
val in = new DataByteArrayInputStream(bytes)
in.readLong()
}
def encodeEntryKey(a1:Byte, a2:Long, a3:Long):Array[Byte] = {
val out = new DataByteArrayOutputStream(17)
out.writeByte(a1.toInt)
out.writeLong(a2)
out.writeLong(a3)
out.getData
}
def encodeEntryKey(a1:Byte, a2:Long, a3:Buffer):Array[Byte] = {
val out = new DataByteArrayOutputStream(9+a3.length)
out.writeByte(a1.toInt)
out.writeLong(a2)
out.write(a3)
out.getData
}
def decodeEntryKey(bytes:Array[Byte]):(Byte, Long, Buffer) = {
val in = new DataByteArrayInputStream(bytes)
(in.readByte(), in.readLong(), in.readBuffer(in.available()))
}
final class RichDB(val db: DB) {
val isPureJavaVersion = db.getClass.getName == "org.iq80.leveldb.impl.DbImpl"
def getProperty(name:String) = db.getProperty(name)
def getApproximateSizes(ranges:Range*) = db.getApproximateSizes(ranges:_*)
def get(key:Array[Byte], ro:ReadOptions=new ReadOptions):Option[Array[Byte]] = {
Option(db.get(key, ro))
}
def close:Unit = db.close()
def delete(key:Array[Byte], wo:WriteOptions=new WriteOptions):Unit = {
db.delete(key, wo)
}
def put(key:Array[Byte], value:Array[Byte], wo:WriteOptions=new WriteOptions):Unit = {
db.put(key, value, wo)
}
def write[T](wo:WriteOptions=new WriteOptions, max_write_latency:TimeMetric = TimeMetric())(func: WriteBatch=>T):T = {
val updates = db.createWriteBatch()
try {
val rc=Some(func(updates))
max_write_latency {
db.write(updates, wo)
}
return rc.get
} finally {
updates.close();
}
}
def store[T](write:WriteBatch, wo:WriteOptions=new WriteOptions) = {
db.write(write, wo)
}
def snapshot[T](func: Snapshot=>T):T = {
val snapshot = db.getSnapshot
try {
func(snapshot)
} finally {
snapshot.close()
}
}
def cursorKeys(ro:ReadOptions=new ReadOptions)(func: Array[Byte] => Boolean): Unit = {
val iterator = db.iterator(ro)
iterator.seekToFirst();
try {
while( iterator.hasNext && func(iterator.peekNext.getKey) ) {
iterator.next()
}
} finally {
iterator.close();
}
}
def cursorKeysPrefixed(prefix:Array[Byte], ro:ReadOptions=new ReadOptions)(func: Array[Byte] => Boolean): Unit = {
val iterator = db.iterator(ro)
might_trigger_compaction(iterator.seek(prefix));
try {
def check(key:Buffer) = {
key.startsWith(prefix) && func(key)
}
while( iterator.hasNext && check(iterator.peekNext.getKey) ) {
iterator.next()
}
} finally {
iterator.close();
}
}
def cursorPrefixed(prefix:Array[Byte], ro:ReadOptions=new ReadOptions)(func: (Array[Byte],Array[Byte]) => Boolean): Unit = {
val iterator = db.iterator(ro)
might_trigger_compaction(iterator.seek(prefix));
try {
def check(key:Buffer) = {
key.startsWith(prefix) && func(key, iterator.peekNext.getValue)
}
while( iterator.hasNext && check(iterator.peekNext.getKey) ) {
iterator.next()
}
} finally {
iterator.close();
}
}
def compare(a1:Array[Byte], a2:Array[Byte]):Int = {
new Buffer(a1).compareTo(new Buffer(a2))
}
def cursorRangeKeys(startIncluded:Array[Byte], endExcluded:Array[Byte], ro:ReadOptions=new ReadOptions)(func: Array[Byte] => Boolean): Unit = {
val iterator = db.iterator(ro)
might_trigger_compaction(iterator.seek(startIncluded));
try {
def check(key:Array[Byte]) = {
if ( compare(key,endExcluded) < 0) {
func(key)
} else {
false
}
}
while( iterator.hasNext && check(iterator.peekNext.getKey) ) {
iterator.next()
}
} finally {
iterator.close();
}
}
def cursorRange(startIncluded:Array[Byte], endExcluded:Array[Byte], ro:ReadOptions=new ReadOptions)(func: (Array[Byte],Array[Byte]) => Boolean): Unit = {
val iterator = db.iterator(ro)
might_trigger_compaction(iterator.seek(startIncluded));
try {
def check(key:Array[Byte]) = {
(compare(key,endExcluded) < 0) && func(key, iterator.peekNext.getValue)
}
while( iterator.hasNext && check(iterator.peekNext.getKey) ) {
iterator.next()
}
} finally {
iterator.close();
}
}
def lastKey(prefix:Array[Byte], ro:ReadOptions=new ReadOptions): Option[Array[Byte]] = {
val last = new Buffer(prefix).deepCopy().data
if ( last.length > 0 ) {
val pos = last.length-1
last(pos) = (last(pos)+1).toByte
}
if(isPureJavaVersion) {
// The pure java version of LevelDB does not support backward iteration.
var rc:Option[Array[Byte]] = None
cursorRangeKeys(prefix, last) { key=>
rc = Some(key)
true
}
rc
} else {
val iterator = db.iterator(ro)
try {
might_trigger_compaction(iterator.seek(last));
if ( iterator.hasPrev ) {
iterator.prev()
} else {
iterator.seekToLast()
}
if ( iterator.hasNext ) {
val key:Buffer = iterator.peekNext.getKey
if(key.startsWith(prefix)) {
Some(key)
} else {
None
}
} else {
None
}
} finally {
iterator.close();
}
}
}
def compact = {
compact_needed = false
db match {
case db:JniDB =>
db.compactRange(null, null)
// case db:DbImpl =>
// val start = new Slice(Array[Byte]('a'.toByte))
// val end = new Slice(Array[Byte]('z'.toByte))
// db.compactRange(2, start, end)
case _ =>
}
}
private def might_trigger_compaction[T](func: => T): T = {
val start = System.nanoTime()
try {
func
} finally {
val duration = System.nanoTime() - start
// If it takes longer than 100 ms..
if( duration > 1000000*100 ) {
compact_needed = true
}
}
}
@volatile
var compact_needed = false
}
def bytes(value:String) = value.getBytes("UTF-8")
import FileSupport._
def create_sequence_file(directory:File, id:Long, suffix:String) = directory / ("%016x%s".format(id, suffix))
def find_sequence_files(directory:File, suffix:String):TreeMap[Long, File] = {
TreeMap((directory.list_files.flatMap { f=>
if( f.getName.endsWith(suffix) ) {
try {
val base = f.getName.stripSuffix(suffix)
val position = java.lang.Long.parseLong(base, 16);
Some(position -> f)
} catch {
case e:NumberFormatException => None
}
} else {
None
}
}): _* )
}
class CollectionMeta extends Serializable {
var size = 0L
var last_key:Array[Byte] = _
}
def copyIndex(from:File, to:File) = {
for( file <- from.list_files ) {
val name: String = file.getName
if( name.endsWith(".sst") ) {
// SST files don't change once created, safe to hard link.
file.linkTo(to / name)
} else if(name == "LOCK") {
// No need to copy the lock file.
} else {
/// These might not be append only files, so avoid hard linking just to be safe.
file.copyTo(to / name)
}
}
}
}
/**
*
* @author Hiram Chirino
*/
class LevelDBClient(store: LevelDBStore) {
import LevelDBClient._
import FileSupport._
val dispatchQueue = createQueue("leveldb")
/////////////////////////////////////////////////////////////////////
//
// Helpers
//
/////////////////////////////////////////////////////////////////////
def directory = store.directory
def logDirectory = Option(store.logDirectory).getOrElse(store.directory)
/////////////////////////////////////////////////////////////////////
//
// Public interface used by the DBManager
//
/////////////////////////////////////////////////////////////////////
def sync = store.sync;
def verifyChecksums = store.verifyChecksums
var log:RecordLog = _
var index:RichDB = _
var plist:RichDB = _
var indexOptions:Options = _
var lastIndexSnapshotPos:Long = _
val snapshotRwLock = new ReentrantReadWriteLock(true)
var factory:DBFactory = _
val logRefs = HashMap[Long, LongCounter]()
var recoveryLogs:java.util.TreeMap[Long, Void] = _
val collectionMeta = HashMap[Long, CollectionMeta]()
def plistIndexFile = directory / ("plist"+INDEX_SUFFIX)
def dirtyIndexFile = directory / ("dirty"+INDEX_SUFFIX)
def tempIndexFile = directory / ("temp"+INDEX_SUFFIX)
def snapshotIndexFile(id:Long) = create_sequence_file(directory,id, INDEX_SUFFIX)
def size: Long = logRefs.size * store.logSize
def createLog: RecordLog = {
new RecordLog(logDirectory, LOG_SUFFIX)
}
var writeExecutor:ExecutorService = _
def writeExecutorExec(func: =>Unit ) = writeExecutor {
func
}
def storeTrace(ascii:String, force:Boolean=false) = {
assert_write_thread_executing
val time = new SimpleDateFormat("dd/MMM/yyyy:HH:mm::ss Z").format(new Date)
log.appender { appender =>
appender.append(LOG_TRACE, new AsciiBuffer("%s: %s".format(time, ascii)))
if( force ) {
appender.force
}
}
}
def might_fail[T](func : =>T):T = {
def handleFailure(e:IOException) = {
var failure:Throwable = e;
if( store.broker_service !=null ) {
// This should start stopping the broker but it might block,
// so do it on another thread...
new Thread("LevelDB IOException handler.") {
override def run() {
try {
store.broker_service.handleIOException(e)
} catch {
case e:RuntimeException =>
failure = e
} finally {
store.stop()
}
}
}.start()
// Lets wait until the broker service has started stopping. Once the
// stopping flag is raised, errors caused by stopping the store should
// not get propagated to the client.
while( !store.broker_service.isStopping ) {
Thread.sleep(100);
}
}
throw new SuppressReplyException(failure);
}
try {
func
} catch {
case e:IOException => handleFailure(e)
case e:Throwable => handleFailure(IOExceptionSupport.create(e))
}
}
def start() = {
init()
replay_init()
might_fail {
log.open()
}
var startPosition = lastIndexSnapshotPos;
// if we cannot locate a log for a snapshot, replay from
// first entry of first available log
if (log.log_info(startPosition).isEmpty) {
if (!log.log_infos.isEmpty) {
startPosition = log.log_infos.firstKey();
}
}
replay_from(startPosition, log.appender_limit)
replay_write_batch = null;
}
def assert_write_thread_executing = assert(Thread.currentThread().getClass == classOf[WriteThread])
def init() ={
// Lets check store compatibility...
directory.mkdirs()
val version_file = directory / "store-version.txt"
if (version_file.exists()) {
val ver = try {
var tmp: String = version_file.readText().trim()
if (tmp.startsWith(STORE_SCHEMA_PREFIX)) {
tmp.stripPrefix(STORE_SCHEMA_PREFIX).toInt
} else {
-1
}
} catch {
case e:Throwable => throw new Exception("Unexpected version file format: " + version_file)
}
ver match {
case STORE_SCHEMA_VERSION => // All is good.
case _ => throw new Exception("Cannot open the store. It's schema version is not supported.")
}
}
version_file.writeText(STORE_SCHEMA_PREFIX + STORE_SCHEMA_VERSION)
writeExecutor = Executors.newFixedThreadPool(1, new ThreadFactory() {
def newThread(r: Runnable) = new WriteThread(r)
})
val factoryNames = store.indexFactory
factory = factoryNames.split("""(,|\s)+""").map(_.trim()).flatMap { name=>
try {
Some(this.getClass.getClassLoader.loadClass(name).newInstance().asInstanceOf[DBFactory])
} catch {
case e:Throwable =>
debug("Could not load factory: "+name+" due to: "+e)
None
}
}.headOption.getOrElse(throw new Exception("Could not load any of the index factory classes: "+factoryNames))
if( factory.getClass.getName == "org.iq80.leveldb.impl.Iq80DBFactory") {
info("Using the pure java LevelDB implementation.")
}
if( factory.getClass.getName == "org.fusesource.leveldbjni.JniDBFactory") {
info("Using the JNI LevelDB implementation.")
}
indexOptions = new Options();
indexOptions.createIfMissing(true);
indexOptions.maxOpenFiles(store.indexMaxOpenFiles)
indexOptions.blockRestartInterval(store.indexBlockRestartInterval)
indexOptions.paranoidChecks(store.paranoidChecks)
indexOptions.writeBufferSize(store.indexWriteBufferSize)
indexOptions.blockSize(store.indexBlockSize)
indexOptions.compressionType( store.indexCompression.toLowerCase match {
case "snappy" => CompressionType.SNAPPY
case "none" => CompressionType.NONE
case _ => CompressionType.SNAPPY
})
indexOptions.cacheSize(store.indexCacheSize)
indexOptions.logger(new Logger() {
val LOG = Log(factory.getClass.getName)
def log(msg: String) = LOG.debug("index: "+msg.stripSuffix("\n"))
})
log = createLog
log.logSize = store.logSize
log.on_log_rotate = ()=> {
post_log_rotate
}
}
def post_log_rotate ={
// We snapshot the index every time we rotate the logs.
writeExecutor {
snapshotIndex(false)
}
}
def replay_init() = {
// Find out what was the last snapshot.
val snapshots = find_sequence_files(directory, INDEX_SUFFIX)
var lastSnapshotIndex = snapshots.lastOption
lastIndexSnapshotPos = lastSnapshotIndex.map(_._1).getOrElse(0)
// Only keep the last snapshot..
snapshots.filterNot(_._1 == lastIndexSnapshotPos).foreach( _._2.recursiveDelete )
tempIndexFile.recursiveDelete
might_fail {
// Setup the plist index.
plistIndexFile.recursiveDelete
plistIndexFile.mkdirs()
plist = new RichDB(factory.open(plistIndexFile, indexOptions));
// Delete the dirty indexes
dirtyIndexFile.recursiveDelete
dirtyIndexFile.mkdirs()
for( (id, file)<- lastSnapshotIndex ) {
try {
copyIndex(file, dirtyIndexFile)
debug("Recovering from last index snapshot at: "+dirtyIndexFile)
} catch {
case e:Exception =>
warn(e, "Could not recover snapshot of the index: "+e)
lastSnapshotIndex = None
}
}
index = new RichDB(factory.open(dirtyIndexFile, indexOptions));
for(value <- index.get(DIRTY_INDEX_KEY) ) {
if( java.util.Arrays.equals(value, TRUE) ) {
warn("Recovering from a dirty index.")
}
}
index.put(DIRTY_INDEX_KEY, TRUE)
loadCounters
}
}
var replay_write_batch: WriteBatch = null
var indexRecoveryPosition = 0L
def replay_from(from:Long, limit:Long, print_progress:Boolean=true) = {
debug("Replay of journal from: %d to %d.", from, limit)
if( replay_write_batch==null ) {
replay_write_batch = index.db.createWriteBatch()
}
might_fail {
try {
// Update the index /w what was stored on the logs..
indexRecoveryPosition = from;
var last_reported_at = System.currentTimeMillis();
var showing_progress = false
var last_reported_pos = 0L
try {
while (indexRecoveryPosition < limit) {
if( print_progress ) {
val now = System.currentTimeMillis();
if( now > last_reported_at+1000 ) {
val at = indexRecoveryPosition-from
val total = limit-from
val rate = (indexRecoveryPosition-last_reported_pos)*1000.0 / (now - last_reported_at)
val eta = (total-at)/rate
val remaining = if(eta > 60*60) {
"%.2f hrs".format(eta/(60*60))
} else if(eta > 60) {
"%.2f mins".format(eta/60)
} else {
"%.0f secs".format(eta)
}
System.out.print("Replaying recovery log: %f%% done (%,d/%,d bytes) @ %,.2f kb/s, %s remaining. \r".format(
at*100.0/total, at, total, rate/1024, remaining))
showing_progress = true;
last_reported_at = now
last_reported_pos = indexRecoveryPosition
}
}
log.read(indexRecoveryPosition).map {
case (kind, data, nextPos) =>
kind match {
case LOG_DATA =>
val message = decodeMessage(data)
store.db.producerSequenceIdTracker.isDuplicate(message.getMessageId)
trace("Replay of LOG_DATA at %d, message id: ", indexRecoveryPosition, message.getMessageId)
case LOG_ADD_COLLECTION =>
val record= decodeCollectionRecord(data)
replay_write_batch.put(encodeLongKey(COLLECTION_PREFIX, record.getKey), data)
collectionMeta.put(record.getKey, new CollectionMeta)
trace("Replay of LOG_ADD_COLLECTION at %d, collection: %s", indexRecoveryPosition, record.getKey)
case LOG_REMOVE_COLLECTION =>
val record = decodeCollectionKeyRecord(data)
// Delete the entries in the collection.
index.cursorPrefixed(encodeLongKey(ENTRY_PREFIX, record.getKey), new ReadOptions) { (key, value)=>
val record = decodeEntryRecord(value)
val pos = if ( record.hasValueLocation ) {
Some(record.getValueLocation)
} else {
None
}
pos.foreach(logRefDecrement(_))
index.delete(key)
true
}
index.delete(data)
collectionMeta.remove(record.getKey)
trace("Replay of LOG_REMOVE_COLLECTION at %d, collection: %s", indexRecoveryPosition, record.getKey)
case LOG_ADD_ENTRY | LOG_UPDATE_ENTRY =>
val record = decodeEntryRecord(data)
val index_record = new EntryRecord.Bean()
index_record.setValueLocation(record.getValueLocation)
if( record.hasValueLength ) {
index_record.setValueLength(record.getValueLength)
}
val index_value = encodeEntryRecord(index_record.freeze()).toByteArray
replay_write_batch.put(encodeEntryKey(ENTRY_PREFIX, record.getCollectionKey, record.getEntryKey), index_value)
if( kind==LOG_ADD_ENTRY ) {
logRefIncrement(record.getValueLocation)
collectionIncrementSize(record.getCollectionKey, record.getEntryKey.toByteArray)
trace("Replay of LOG_ADD_ENTRY at %d, collection: %s, entry: %s", indexRecoveryPosition, record.getCollectionKey, record.getEntryKey)
} else {
trace("Replay of LOG_UPDATE_ENTRY at %d, collection: %s, entry: %s", indexRecoveryPosition, record.getCollectionKey, record.getEntryKey)
}
case LOG_REMOVE_ENTRY =>
val record = decodeEntryRecord(data)
// Figure out which log file this message reference is pointing at..
if ( record.hasValueLocation ) {
logRefDecrement(record.getValueLocation)
}
replay_write_batch.delete(encodeEntryKey(ENTRY_PREFIX, record.getCollectionKey, record.getEntryKey))
collectionDecrementSize( record.getCollectionKey)
trace("Replay of LOG_REMOVE_ENTRY collection: %s, entry: %s", indexRecoveryPosition, record.getCollectionKey, record.getEntryKey)
case LOG_TRACE =>
trace("Replay of LOG_TRACE, message: %s", indexRecoveryPosition, data.ascii())
case RecordLog.UOW_END_RECORD =>
trace("Replay of UOW_END_RECORD")
index.db.write(replay_write_batch)
replay_write_batch=index.db.createWriteBatch()
case kind => // Skip other records, they don't modify the index.
trace("Skipping replay of %d record kind at %d", kind, indexRecoveryPosition)
}
indexRecoveryPosition = nextPos
}
}
}
catch {
case e:Throwable => e.printStackTrace()
}
if(showing_progress) {
System.out.println("Replaying recovery log: 100% done ");
}
} catch {
case e:Throwable =>
// replay failed.. good thing we are in a retry block...
index.close
replay_write_batch = null
throw e;
} finally {
recoveryLogs = null
debug("Replay end")
}
}
}
private def logRefDecrement(pos: Long) {
for( key <- logRefKey(pos) ) {
logRefs.get(key) match {
case Some(counter) => counter.decrementAndGet() == 0
case None => warn("invalid: logRefDecrement: "+pos)
}
}
}
private def logRefIncrement(pos: Long) {
for( key <- logRefKey(pos) ) {
logRefs.getOrElseUpdate(key, new LongCounter(0)).incrementAndGet()
}
}
def logRefKey(pos: Long, log_info: RecordLog.LogInfo=null): Option[Long] = {
if( log_info!=null ) {
Some(log_info.position)
} else {
val rc = if( recoveryLogs !=null ) {
Option(recoveryLogs.floorKey(pos))
} else {
log.log_info(pos).map(_.position)
}
if( !rc.isDefined ) {
warn("Invalid log position: " + pos)
}
rc
}
}
private def collectionDecrementSize(key: Long) {
collectionMeta.get(key).foreach(_.size -= 1)
}
private def collectionIncrementSize(key: Long, last_key:Array[Byte]) {
collectionMeta.get(key).foreach{ x=>
x.size += 1
x.last_key = last_key
}
}
private def storeCounters = {
def storeMap[T <: AnyRef](key:Array[Byte], map:HashMap[Long, T]) {
val baos = new ByteArrayOutputStream()
val os = new ObjectOutputStream(baos);
os.writeInt(map.size);
for( (k,v) <- map ) {
os.writeLong(k)
os.writeObject(v)
}
os.close()
try {
index.put(key, baos.toByteArray)
}
catch {
case e : Throwable => throw e
}
}
def storeList[T <: AnyRef](key:Array[Byte], list:Array[Long]) {
val baos = new ByteArrayOutputStream()
val os = new ObjectOutputStream(baos);
os.writeInt(list.size);
for( k <- list ) {
os.writeLong(k)
}
os.close()
try {
index.put(key, baos.toByteArray)
}
catch {
case e : Throwable => throw e
}
}
def storeObject(key:Array[Byte], o:Object) = {
val baos = new ByteArrayOutputStream()
val os = new ObjectOutputStream(baos);
os.writeObject(o)
os.close()
index.put(key, baos.toByteArray)
}
storeMap(LOG_REF_INDEX_KEY, logRefs)
storeMap(COLLECTION_META_KEY, collectionMeta)
storeList(LOGS_INDEX_KEY, log.log_file_positions)
storeObject(PRODUCER_IDS_INDEX_KEY, store.db.producerSequenceIdTracker)
}
private def loadCounters = {
def loadMap[T <: AnyRef](key:Array[Byte], map:HashMap[Long, T]) {
map.clear()
index.get(key, new ReadOptions).foreach { value=>
val bais = new ByteArrayInputStream(value)
val is = new ObjectInputStream(bais);
var remaining = is.readInt()
while(remaining > 0 ) {
map.put(is.readLong(), is.readObject().asInstanceOf[T])
remaining-=1
}
}
}
def loadList[T <: AnyRef](key:Array[Byte]) = {
index.get(key, new ReadOptions).map { value=>
val rc = ListBuffer[Long]()
val bais = new ByteArrayInputStream(value)
val is = new ObjectInputStream(bais);
var remaining = is.readInt()
while(remaining > 0 ) {
rc.append(is.readLong())
remaining-=1
}
rc
}
}
def loadObject(key:Array[Byte]) = {
index.get(key, new ReadOptions).map { value=>
val bais = new ByteArrayInputStream(value)
val is = new ObjectInputStream(bais);
is.readObject();
}
}
loadMap(LOG_REF_INDEX_KEY, logRefs)
loadMap(COLLECTION_META_KEY, collectionMeta)
for( list <- loadList(LOGS_INDEX_KEY) ) {
recoveryLogs = new java.util.TreeMap[Long, Void]()
for( k <- list ) {
recoveryLogs.put(k, null)
}
}
for( audit <- loadObject(PRODUCER_IDS_INDEX_KEY) ) {
store.db.producerSequenceIdTracker = audit.asInstanceOf[ActiveMQMessageAuditNoSync]
}
}
var stored_wal_append_position = 0L
def wal_append_position = this.synchronized {
if (log!=null && log.isOpen) {
log.appender_limit
} else {
stored_wal_append_position
}
}
def dirty_stop = this.synchronized {
def ingorefailure(func: =>Unit) = try { func } catch { case e:Throwable=> }
ingorefailure(index.close)
ingorefailure(log.close)
ingorefailure(plist.close)
ingorefailure(might_fail(throw new IOException("non-clean close")))
}
def stop():Unit = {
var executorToShutdown:ExecutorService = null
this synchronized {
if (writeExecutor != null) {
executorToShutdown = writeExecutor
writeExecutor = null
}
}
if (executorToShutdown != null) {
executorToShutdown.shutdown
executorToShutdown.awaitTermination(60, TimeUnit.SECONDS)
// this blocks until all io completes..
snapshotRwLock.writeLock().lock()
try {
// Suspend also deletes the index.
if( index!=null ) {
storeCounters
index.put(DIRTY_INDEX_KEY, FALSE, new WriteOptions().sync(true))
index.close
index = null
debug("Gracefuly closed the index")
copyDirtyIndexToSnapshot
}
this synchronized {
if (log!=null && log.isOpen) {
log.close
stored_wal_append_position = log.appender_limit
log = null
}
}
if( plist!=null ) {
plist.close
plist=null
}
} finally {
snapshotRwLock.writeLock().unlock()
}
}
}
def usingIndex[T](func: =>T):T = {
val lock = snapshotRwLock.readLock();
lock.lock()
try {
func
} finally {
lock.unlock()
}
}
def might_fail_using_index[T](func: =>T):T = might_fail(usingIndex( func ))
/**
* TODO: expose this via management APIs, handy if you want to
* do a file system level snapshot and want the data to be consistent.
*/
def suspend() = {
// Make sure we are the only ones accessing the index. since
// we will be closing it to create a consistent snapshot.
snapshotRwLock.writeLock().lock()
storeCounters
index.put(DIRTY_INDEX_KEY, FALSE, new WriteOptions().sync(true))
// Suspend the index so that it's files are not changed async on us.
index.db.suspendCompactions()
}
/**
* TODO: expose this via management APIs, handy if you want to
* do a file system level snapshot and want the data to be consistent.
*/
def resume() = {
// re=open it..
index.db.resumeCompactions()
snapshotRwLock.writeLock().unlock()
}
def nextIndexSnapshotPos:Long = wal_append_position
def copyDirtyIndexToSnapshot:Unit = {
if( nextIndexSnapshotPos == lastIndexSnapshotPos ) {
// no need to snapshot again...
return
}
copyDirtyIndexToSnapshot(nextIndexSnapshotPos)
}
def copyDirtyIndexToSnapshot(walPosition:Long):Unit = {
debug("Taking a snapshot of the current index: "+snapshotIndexFile(walPosition))
// Where we start copying files into. Delete this on
// restart.
val tmpDir = tempIndexFile
tmpDir.mkdirs()
try {
// Copy the index to the tmp dir.
copyIndex(dirtyIndexFile, tmpDir)
// Rename to signal that the snapshot is complete.
tmpDir.renameTo(snapshotIndexFile(walPosition))
replaceLatestSnapshotDirectory(walPosition)
} catch {
case e: Exception =>
// if we could not snapshot for any reason, delete it as we don't
// want a partial check point..
warn(e, "Could not snapshot the index: " + e)
tmpDir.recursiveDelete
}
}
def replaceLatestSnapshotDirectory(newSnapshotIndexPos: Long) {
snapshotIndexFile(lastIndexSnapshotPos).recursiveDelete
lastIndexSnapshotPos = newSnapshotIndexPos
}
def snapshotIndex(sync:Boolean=false):Unit = {
suspend()
try {
if( sync ) {
log.current_appender.force
}
copyDirtyIndexToSnapshot
} finally {
resume()
}
}
def purge() = {
suspend()
try{
log.close
locked_purge
} finally {
might_fail {
log.open()
}
resume()
}
}
def locked_purge {
for( x <- logDirectory.list_files) {
if (x.getName.endsWith(".log")) {
x.delete()
}
}
for( x <- directory.list_files) {
if (x.getName.endsWith(".index")) {
x.recursiveDelete
}
}
}
def addCollection(record: CollectionRecord.Buffer) = {
assert_write_thread_executing
val key = encodeLongKey(COLLECTION_PREFIX, record.getKey)
val value = record.toUnframedBuffer
might_fail_using_index {
log.appender { appender =>
appender.append(LOG_ADD_COLLECTION, value)
index.put(key, value.toByteArray)
}
}
collectionMeta.put(record.getKey, new CollectionMeta)
}
def getLogAppendPosition = log.appender_limit
def listCollections: Seq[(Long, CollectionRecord.Buffer)] = {
val rc = ListBuffer[(Long, CollectionRecord.Buffer)]()
might_fail_using_index {
val ro = new ReadOptions
ro.verifyChecksums(verifyChecksums)
ro.fillCache(false)
index.cursorPrefixed(COLLECTION_PREFIX_ARRAY, ro) { (key, value) =>
rc.append(( decodeLongKey(key)._2, CollectionRecord.FACTORY.parseUnframed(value) ))
true // to continue cursoring.
}
}
rc
}
def removeCollection(collectionKey: Long) = {
assert_write_thread_executing
val key = encodeLongKey(COLLECTION_PREFIX, collectionKey)
val value = encodeVLong(collectionKey)
val entryKeyPrefix = encodeLongKey(ENTRY_PREFIX, collectionKey)
collectionMeta.remove(collectionKey)
might_fail_using_index {
log.appender { appender =>
appender.append(LOG_REMOVE_COLLECTION, new Buffer(value))
}
val ro = new ReadOptions
ro.fillCache(false)
ro.verifyChecksums(verifyChecksums)
index.cursorPrefixed(entryKeyPrefix, ro) { (key, value)=>
val record = decodeEntryRecord(value)
val pos = if ( record.hasValueLocation ) {
Some(record.getValueLocation)
} else {
None
}
pos.foreach(logRefDecrement(_))
index.delete(key)
true
}
index.delete(key)
}
}
def collectionEmpty(collectionKey: Long) = {
assert_write_thread_executing
val key = encodeLongKey(COLLECTION_PREFIX, collectionKey)
val value = encodeVLong(collectionKey)
val entryKeyPrefix = encodeLongKey(ENTRY_PREFIX, collectionKey)
val meta = collectionMeta.getOrElseUpdate(collectionKey, new CollectionMeta)
meta.size = 0
meta.last_key = null
might_fail_using_index {
index.get(key).foreach { collectionData =>
log.appender { appender =>
appender.append(LOG_REMOVE_COLLECTION, new Buffer(value))
appender.append(LOG_ADD_COLLECTION, new Buffer(collectionData))
}
val ro = new ReadOptions
ro.fillCache(false)
ro.verifyChecksums(verifyChecksums)
index.cursorPrefixed(entryKeyPrefix, ro) { (key, value)=>
val record = decodeEntryRecord(value)
val pos = if ( record.hasValueLocation ) {
Some(record.getValueLocation)
} else {
None
}
pos.foreach(logRefDecrement(_))
index.delete(key)
true
}
}
}
}
def decodeQueueEntryMeta(value:EntryRecord.Getter):Int= {
if( value.hasMeta ) {
val is = new DataByteArrayInputStream(value.getMeta);
val metaVersion = is.readVarInt()
metaVersion match {
case 1 =>
return is.readVarInt()
case _ =>
}
}
return 0
}
def getDeliveryCounter(collectionKey: Long, seq:Long):Int = {
val ro = new ReadOptions
ro.fillCache(true)
ro.verifyChecksums(verifyChecksums)
val key = encodeEntryKey(ENTRY_PREFIX, collectionKey, encodeLong(seq))
var rc = 0
might_fail_using_index {
for( v <- index.get(key, ro) ) {
rc = decodeQueueEntryMeta(EntryRecord.FACTORY.parseUnframed(v))
}
}
return rc
}
def queueCursor(collectionKey: Long, seq:Long, endSeq:Long)(func: (Message)=>Boolean) = {
collectionCursor(collectionKey, encodeLong(seq), encodeLong(endSeq)) { (key, value) =>
val seq = decodeLong(key)
var locator = DataLocator(store, value.getValueLocation, value.getValueLength)
val msg = getMessage(locator)
if( msg !=null ) {
msg.getMessageId().setEntryLocator(EntryLocator(collectionKey, seq))
msg.getMessageId().setDataLocator(locator)
msg.setRedeliveryCounter(decodeQueueEntryMeta(value))
func(msg)
} else {
warn("Could not load message seq: "+seq+" from "+locator)
true
}
}
}
def transactionCursor(collectionKey: Long)(func: (AnyRef)=>Boolean) = {
collectionCursor(collectionKey, encodeLong(0), encodeLong(Long.MaxValue)) { (key, value) =>
val seq = decodeLong(key)
if( value.getMeta != null ) {
val is = new DataByteArrayInputStream(value.getMeta);
val log = is.readLong()
val offset = is.readInt()
val qid = is.readLong()
val seq = is.readLong()
val sub = is.readLong()
val ack = store.wireFormat.unmarshal(is).asInstanceOf[MessageAck]
ack.getLastMessageId.setDataLocator(DataLocator(store, log, offset))
ack.getLastMessageId.setEntryLocator(EntryLocator(qid, seq))
func(XaAckRecord(collectionKey, seq, ack, sub))
} else {
var locator = DataLocator(store, value.getValueLocation, value.getValueLength)
val msg = getMessage(locator)
if( msg !=null ) {
msg.getMessageId().setEntryLocator(EntryLocator(collectionKey, seq))
msg.getMessageId().setDataLocator(locator)
func(msg)
} else {
warn("Could not load XA message seq: "+seq+" from "+locator)
true
}
}
}
}
def getAckPosition(subKey: Long): Long = {
might_fail_using_index {
index.get(encodeEntryKey(ENTRY_PREFIX, subKey, ACK_POSITION)).map{ value=>
val record = decodeEntryRecord(value)
record.getValueLocation()
}.getOrElse(0L)
}
}
def getMessage(locator:AnyRef):Message = {
assert(locator!=null)
val buffer = locator match {
case x:MessageRecord =>
// Encoded form is still in memory..
Some(x.data)
case DataLocator(store, pos, len) =>
// Load the encoded form from disk.
log.read(pos, len).map(new Buffer(_))
}
// Lets decode
buffer.map(decodeMessage(_)).getOrElse(null)
}
def decodeMessage(x: Buffer): Message = {
var data = if (store.snappyCompressLogs) {
Snappy.uncompress(x)
} else {
x
}
store.wireFormat.unmarshal(new ByteSequence(data.data, data.offset, data.length)).asInstanceOf[Message]
}
def collectionCursor(collectionKey: Long, cursorPosition:Buffer, endCursorPosition:Buffer)(func: (Buffer, EntryRecord.Buffer)=>Boolean) = {
val ro = new ReadOptions
ro.fillCache(true)
ro.verifyChecksums(verifyChecksums)
val start = encodeEntryKey(ENTRY_PREFIX, collectionKey, cursorPosition)
val end = encodeEntryKey(ENTRY_PREFIX, collectionKey, endCursorPosition)
might_fail_using_index {
index.cursorRange(start, end, ro) { case (key, value) =>
func(key.buffer.moveHead(9), EntryRecord.FACTORY.parseUnframed(value))
}
}
}
def collectionSize(collectionKey: Long) = {
collectionMeta.get(collectionKey).map(_.size).getOrElse(0L)
}
def collectionIsEmpty(collectionKey: Long) = {
val entryKeyPrefix = encodeLongKey(ENTRY_PREFIX, collectionKey)
var empty = true
might_fail_using_index {
val ro = new ReadOptions
ro.fillCache(false)
ro.verifyChecksums(verifyChecksums)
index.cursorKeysPrefixed(entryKeyPrefix, ro) { key =>
empty = false
false
}
}
empty
}
val max_write_message_latency = TimeMetric()
val max_write_enqueue_latency = TimeMetric()
val max_index_write_latency = TimeMetric()
def store(uows: Array[DelayableUOW]) {
assert_write_thread_executing
might_fail_using_index {
log.appender { appender =>
val syncNeeded = index.write(new WriteOptions, max_index_write_latency) { batch =>
write_uows(uows, appender, batch)
}
if( syncNeeded && sync ) {
appender.force
}
} // end of log.appender { block }
// now that data is logged.. locate message from the data in the logs
for( uow <- uows ) {
for((msg, action) <- uow.actions ){
val messageRecord = action.messageRecord
if (messageRecord != null) {
messageRecord.id.setDataLocator(messageRecord.locator)
}
}
}
}
}
def write_uows(uows: Array[DelayableUOW], appender: RecordLog#LogAppender, batch: WriteBatch) = {
var syncNeeded = false
var write_message_total = 0L
var write_enqueue_total = 0L
for( uow <- uows ) {
for( (msg, action) <- uow.actions ) {
val messageRecord = action.messageRecord
var log_info: LogInfo = null
var dataLocator: DataLocator = null
if (messageRecord != null && messageRecord.locator == null) {
store.db.producerSequenceIdTracker.isDuplicate(messageRecord.id)
val start = System.nanoTime()
val p = appender.append(LOG_DATA, messageRecord.data)
log_info = p._2
dataLocator = DataLocator(store, p._1, messageRecord.data.length)
messageRecord.locator = dataLocator
// println("msg: "+messageRecord.id+" -> "+dataLocator)
write_message_total += System.nanoTime() - start
}
for( entry <- action.dequeues) {
val keyLocation = entry.id.getEntryLocator.asInstanceOf[EntryLocator]
val key = encodeEntryKey(ENTRY_PREFIX, keyLocation.qid, keyLocation.seq)
if (dataLocator == null) {
dataLocator = entry.id.getDataLocator match {
case x: DataLocator => x
case x: MessageRecord => x.locator
case _ => throw new RuntimeException("Unexpected locator type: " + dataLocator)
}
}
// println("deq: "+entry.id+" -> "+dataLocator)
val log_record = new EntryRecord.Bean()
log_record.setCollectionKey(entry.queueKey)
log_record.setEntryKey(new Buffer(key, 9, 8))
log_record.setValueLocation(dataLocator.pos)
appender.append(LOG_REMOVE_ENTRY, encodeEntryRecord(log_record.freeze()))
batch.delete(key)
logRefDecrement(dataLocator.pos)
collectionDecrementSize(entry.queueKey)
}
for( entry<- action.enqueues) {
if (dataLocator == null) {
dataLocator = entry.id.getDataLocator match {
case x: DataLocator => x
case x: MessageRecord => x.locator
case _ =>
throw new RuntimeException("Unexpected locator type")
}
}
// println("enq: "+entry.id+" -> "+dataLocator)
val start = System.nanoTime()
val key = encodeEntryKey(ENTRY_PREFIX, entry.queueKey, entry.queueSeq)
assert(entry.id.getDataLocator() != null)
val log_record = new EntryRecord.Bean()
log_record.setCollectionKey(entry.queueKey)
log_record.setEntryKey(new Buffer(key, 9, 8))
log_record.setValueLocation(dataLocator.pos)
log_record.setValueLength(dataLocator.len)
val kind = if (entry.deliveries==0) LOG_ADD_ENTRY else LOG_UPDATE_ENTRY
appender.append(kind, encodeEntryRecord(log_record.freeze()))
val index_record = new EntryRecord.Bean()
index_record.setValueLocation(dataLocator.pos)
index_record.setValueLength(dataLocator.len)
// Store the delivery counter.
if( entry.deliveries!=0 ) {
val os = new DataByteArrayOutputStream()
os.writeVarInt(1) // meta data format version
os.writeVarInt(entry.deliveries)
index_record.setMeta(os.toBuffer)
}
val index_data = encodeEntryRecord(index_record.freeze()).toByteArray
batch.put(key, index_data)
if( kind==LOG_ADD_ENTRY ) {
logRefIncrement(dataLocator.pos)
collectionIncrementSize(entry.queueKey, log_record.getEntryKey.toByteArray)
}
write_enqueue_total += System.nanoTime() - start
}
for( entry <- action.xaAcks ) {
val ack = entry.ack
if (dataLocator == null) {
dataLocator = ack.getLastMessageId.getDataLocator match {
case x: DataLocator => x
case x: MessageRecord => x.locator
case _ =>
throw new RuntimeException("Unexpected locator type")
}
}
// println(dataLocator)
val el = ack.getLastMessageId.getEntryLocator.asInstanceOf[EntryLocator];
val os = new DataByteArrayOutputStream()
os.writeLong(dataLocator.pos)
os.writeInt(dataLocator.len)
os.writeLong(el.qid)
os.writeLong(el.seq)
os.writeLong(entry.sub)
store.wireFormat.marshal(ack, os)
var ack_encoded = os.toBuffer
val key = encodeEntryKey(ENTRY_PREFIX, entry.container, entry.seq)
val log_record = new EntryRecord.Bean()
log_record.setCollectionKey(entry.container)
log_record.setEntryKey(new Buffer(key, 9, 8))
log_record.setMeta(ack_encoded)
appender.append(LOG_ADD_ENTRY, encodeEntryRecord(log_record.freeze()))
val index_record = new EntryRecord.Bean()
index_record.setMeta(ack_encoded)
batch.put(key, encodeEntryRecord(log_record.freeze()).toByteArray)
}
}
for( entry <- uow.subAcks ) {
val key = encodeEntryKey(ENTRY_PREFIX, entry.subKey, ACK_POSITION)
val log_record = new EntryRecord.Bean()
log_record.setCollectionKey(entry.subKey)
log_record.setEntryKey(ACK_POSITION)
log_record.setValueLocation(entry.ackPosition)
appender.append(LOG_UPDATE_ENTRY, encodeEntryRecord(log_record.freeze()))
val index_record = new EntryRecord.Bean()
index_record.setValueLocation(entry.ackPosition)
batch.put(key, encodeEntryRecord(index_record.freeze()).toByteArray)
}
if (uow.syncNeeded) {
syncNeeded = true
}
}
max_write_message_latency.add(write_message_total)
max_write_enqueue_latency.add(write_enqueue_total)
syncNeeded
}
def getCollectionEntries(collectionKey: Long, firstSeq:Long, lastSeq:Long): Seq[(Buffer, EntryRecord.Buffer)] = {
var rc = ListBuffer[(Buffer, EntryRecord.Buffer)]()
val ro = new ReadOptions
ro.verifyChecksums(verifyChecksums)
ro.fillCache(true)
might_fail_using_index {
index.snapshot { snapshot =>
ro.snapshot(snapshot)
val start = encodeEntryKey(ENTRY_PREFIX, collectionKey, firstSeq)
val end = encodeEntryKey(ENTRY_PREFIX, collectionKey, lastSeq+1)
index.cursorRange( start, end, ro ) { (key, value) =>
val (_, _, seq) = decodeEntryKey(key)
rc.append((seq, EntryRecord.FACTORY.parseUnframed(value)))
true
}
}
}
rc
}
def getLastQueueEntrySeq(collectionKey: Long): Long = {
getLastCollectionEntryKey(collectionKey).map(_.bigEndianEditor().readLong()).getOrElse(0L)
}
def getLastCollectionEntryKey(collectionKey: Long): Option[Buffer] = {
collectionMeta.get(collectionKey).flatMap(x=> Option(x.last_key)).map(new Buffer(_))
}
// APLO-245: lets try to detect when leveldb needs a compaction..
private def detect_if_compact_needed:Unit = {
// auto compaction might be disabled...
if ( store.autoCompactionRatio <= 0 ) {
return
}
// How much space is the dirty index using??
var index_usage = 0L
for( file <- dirtyIndexFile.recursiveList ) {
if(!file.isDirectory && file.getName.endsWith(".sst") ) {
index_usage += file.length()
}
}
// Lets use the log_refs to get a rough estimate on how many entries are store in leveldb.
var index_queue_entries=0L
for ( (_, count) <- logRefs ) {
index_queue_entries += count.get()
}
// Don't force compactions until level 0 is full.
val SSL_FILE_SIZE = 1024*1024*4L
if( index_usage > SSL_FILE_SIZE*10 ) {
if ( index_queue_entries > 0 ) {
val ratio = (index_usage*1.0f/index_queue_entries)
// println("usage: index_usage:%d, index_queue_entries:%d, ratio: %f".format(index_usage, index_queue_entries, ratio))
// lets compact if we go way over the healthy ratio.
if( ratio > store.autoCompactionRatio ) {
index.compact_needed = true
}
} else {
// at most the index should have 1 full level file.
index.compact_needed = true
}
}
}
def gc(topicPositions:Seq[(Long, Long)]):Unit = {
// Delete message refs for topics who's consumers have advanced..
if( !topicPositions.isEmpty ) {
might_fail_using_index {
index.write(new WriteOptions, max_index_write_latency) { batch =>
for( (topic, first) <- topicPositions ) {
val ro = new ReadOptions
ro.fillCache(true)
ro.verifyChecksums(verifyChecksums)
val start = encodeEntryKey(ENTRY_PREFIX, topic, 0)
val end = encodeEntryKey(ENTRY_PREFIX, topic, first)
debug("Topic: %d GC to seq: %d", topic, first)
index.cursorRange(start, end, ro) { case (key, value) =>
val entry = EntryRecord.FACTORY.parseUnframed(value)
batch.delete(key)
logRefDecrement(entry.getValueLocation)
true
}
}
}
}
}
detect_if_compact_needed
// Lets compact the leveldb index if it looks like we need to.
if( index.compact_needed ) {
val start = System.nanoTime()
index.compact
val duration = System.nanoTime() - start;
info("Compacted the leveldb index at: %s in %.2f ms", dirtyIndexFile, (duration / 1000000.0))
}
import collection.JavaConversions._
// drop the logs that are no longer referenced.
for( (x,y) <- logRefs.toSeq ) {
if( y.get() <= 0 ) {
if( y.get() < 0 ) {
warn("Found a negative log reference for log: "+x)
}
debug("Log no longer referenced: %x", x)
logRefs.remove(x)
}
}
val emptyJournals = log.log_infos.keySet.toSet -- logRefs.keySet
// We don't want to delete any journals that the index has not snapshot'ed or
// the the
var limit = oldest_retained_snapshot
val deleteLimit = logRefKey(limit).getOrElse(limit).min(log.appender_start)
emptyJournals.foreach { id =>
if ( id < deleteLimit ) {
debug("Deleting log at %x", id)
log.delete(id)
}
}
}
def oldest_retained_snapshot = lastIndexSnapshotPos
def removePlist(collectionKey: Long) = {
val entryKeyPrefix = encodeLong(collectionKey)
collectionMeta.remove(collectionKey)
might_fail {
val ro = new ReadOptions
ro.fillCache(false)
ro.verifyChecksums(false)
plist.cursorPrefixed(entryKeyPrefix, ro) { (key, value)=>
plist.delete(key)
true
}
}
}
def plistPut(key:Array[Byte], value:Array[Byte]) = plist.put(key, value, PLIST_WRITE_OPTIONS)
def plistDelete(key:Array[Byte]) = plist.delete(key, PLIST_WRITE_OPTIONS)
def plistGet(key:Array[Byte]) = plist.get(key)
def plistIterator = plist.db.iterator()
}