You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
674 lines
32 KiB
674 lines
32 KiB
//plasmaWordIndexDistribution.java
|
|
//-------------------------------------------
|
|
//(C) by Michael Peter Christen; mc@anomic.de
|
|
//first published on http://www.anomic.de
|
|
//Frankfurt, Germany, 2005
|
|
//
|
|
//last major change: $LastChangedDate$ by $LastChangedBy$
|
|
//Revision: $LastChangedRevision$
|
|
//
|
|
//This program is free software; you can redistribute it and/or modify
|
|
//it under the terms of the GNU General Public License as published by
|
|
//the Free Software Foundation; either version 2 of the License, or
|
|
//(at your option) any later version.
|
|
//
|
|
//This program is distributed in the hope that it will be useful,
|
|
//but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
//MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
//GNU General Public License for more details.
|
|
//
|
|
//You should have received a copy of the GNU General Public License
|
|
//along with this program; if not, write to the Free Software
|
|
//Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
|
|
//
|
|
//Using this software in any meaning (reading, learning, copying, compiling,
|
|
//running) means that you agree that the Author(s) is (are) not responsible
|
|
//for cost, loss of data or any harm that may be caused directly or indirectly
|
|
//by usage of this softare or this documentation. The usage of this software
|
|
//is on your own risk. The installation and usage (starting/running) of this
|
|
//software may allow other people or application to access your computer and
|
|
//any attached devices and is highly dependent on the configuration of the
|
|
//software which must be done by the user of the software; the author(s) is
|
|
//(are) also not responsible for proper configuration and usage of the
|
|
//software, even if provoked by documentation provided together with
|
|
//the software.
|
|
//
|
|
//Any changes to this file according to the GPL as documented in the file
|
|
//gpl.txt aside this file in the shipment you received can be done to the
|
|
//lines that follows this copyright notice here, but changes must not be
|
|
//done inside the copyright notive above. A re-distribution must contain
|
|
//the intact and unchanged copyright notice.
|
|
//Contributions and changes to the program code must be marked as such.
|
|
|
|
package de.anomic.plasma;
|
|
|
|
import java.io.IOException;
|
|
import java.util.Enumeration;
|
|
import java.util.Vector;
|
|
import java.util.Iterator;
|
|
import java.util.HashSet;
|
|
import java.util.HashMap;
|
|
|
|
import de.anomic.yacy.yacyCore;
|
|
import de.anomic.yacy.yacySeed;
|
|
import de.anomic.yacy.yacySeedDB;
|
|
import de.anomic.yacy.yacyClient;
|
|
import de.anomic.yacy.yacyDHTAction;
|
|
import de.anomic.server.serverCodings;
|
|
import de.anomic.server.logging.serverLog;
|
|
import de.anomic.kelondro.kelondroException;
|
|
|
|
public class plasmaWordIndexDistribution {
|
|
|
|
// distributes parts of the index to other peers
|
|
// stops as soon as an error occurrs
|
|
|
|
private int indexCount;
|
|
private int juniorPeerCount, seniorPeerCount;
|
|
private long maxTime;
|
|
|
|
private plasmaURLPool urlPool;
|
|
private plasmaWordIndex wordIndex;
|
|
serverLog log;
|
|
boolean paused = false;
|
|
private boolean enabled;
|
|
private boolean enabledWhileCrawling;
|
|
private boolean closed;
|
|
|
|
public transferIndexThread transferIdxThread = null;
|
|
|
|
public plasmaWordIndexDistribution(plasmaURLPool urlPool, plasmaWordIndex wordIndex, serverLog log,
|
|
boolean enable, boolean enabledWhileCrawling) {
|
|
this.urlPool = urlPool;
|
|
this.wordIndex = wordIndex;
|
|
this.enabled = enable;
|
|
this.enabledWhileCrawling = enabledWhileCrawling;
|
|
this.log = log;
|
|
this.closed = false;
|
|
setCounts(100 /*indexCount*/, 1 /*juniorPeerCount*/, 3 /*seniorPeerCount*/, 8000);
|
|
}
|
|
|
|
public void enable() {
|
|
enabled = true;
|
|
}
|
|
|
|
public void disable() {
|
|
enabled = false;
|
|
}
|
|
|
|
public void enableWhileCrawling() {
|
|
this.enabledWhileCrawling = true;
|
|
}
|
|
|
|
public void disableWhileCrawling() {
|
|
this.enabledWhileCrawling = false;
|
|
}
|
|
|
|
public void close() {
|
|
closed = true;
|
|
if (transferIdxThread != null) {
|
|
stopTransferWholeIndex(false);
|
|
}
|
|
}
|
|
|
|
public boolean job() {
|
|
|
|
if (this.closed) {
|
|
log.logFine("no word distribution: closed");
|
|
return false;
|
|
}
|
|
if (yacyCore.seedDB == null) {
|
|
log.logFine("no word distribution: seedDB == null");
|
|
return false;
|
|
}
|
|
if (yacyCore.seedDB.mySeed == null) {
|
|
log.logFine("no word distribution: mySeed == null");
|
|
return false;
|
|
}
|
|
if (yacyCore.seedDB.mySeed.isVirgin()) {
|
|
log.logFine("no word distribution: status is virgin");
|
|
return false;
|
|
}
|
|
if (!(enabled)) {
|
|
log.logFine("no word distribution: not enabled");
|
|
return false;
|
|
}
|
|
if (paused) {
|
|
log.logFine("no word distribution: paused");
|
|
return false;
|
|
}
|
|
if (urlPool.loadedURL.size() < 10) {
|
|
log.logFine("no word distribution: loadedURL.size() = " + urlPool.loadedURL.size());
|
|
return false;
|
|
}
|
|
if (wordIndex.size() < 100) {
|
|
log.logFine("no word distribution: not enough words - wordIndex.size() = " + wordIndex.size());
|
|
return false;
|
|
}
|
|
if ((!enabledWhileCrawling) && (urlPool.noticeURL.stackSize() > 0)) {
|
|
log.logFine("no word distribution: crawl in progress - noticeURL.stackSize() = " + urlPool.noticeURL.stackSize());
|
|
return false;
|
|
}
|
|
|
|
// do the transfer
|
|
int peerCount = (yacyCore.seedDB.mySeed.isJunior()) ? juniorPeerCount : seniorPeerCount;
|
|
long starttime = System.currentTimeMillis();
|
|
int transferred = performTransferIndex(indexCount, peerCount, true);
|
|
|
|
if (transferred <= 0) {
|
|
log.logFine("no word distribution: transfer failed");
|
|
return false;
|
|
}
|
|
|
|
// adopt transfer count
|
|
if ((System.currentTimeMillis() - starttime) > (maxTime * peerCount))
|
|
indexCount--;
|
|
else
|
|
indexCount++;
|
|
if (indexCount < 50) indexCount = 50;
|
|
|
|
// show success
|
|
return true;
|
|
|
|
}
|
|
|
|
public void setCounts(int indexCount, int juniorPeerCount, int seniorPeerCount, long maxTimePerTransfer) {
|
|
this.maxTime = maxTimePerTransfer;
|
|
this.indexCount = indexCount;
|
|
if (indexCount < 30) indexCount = 30;
|
|
this.juniorPeerCount = juniorPeerCount;
|
|
this.seniorPeerCount = seniorPeerCount;
|
|
}
|
|
|
|
public int performTransferIndex(int indexCount, int peerCount, boolean delete) {
|
|
if ((yacyCore.seedDB == null) || (yacyCore.seedDB.sizeConnected() == 0)) return -1;
|
|
|
|
// collect index
|
|
String startPointHash = selectTransferStart();
|
|
log.logFine("Selected hash " + startPointHash + " as start point for index distribution, distance = " + yacyDHTAction.dhtDistance(yacyCore.seedDB.mySeed.hash, startPointHash));
|
|
Object[] selectResult = selectTransferIndexes(startPointHash, indexCount);
|
|
plasmaWordIndexEntity[] indexEntities = (plasmaWordIndexEntity[]) selectResult[0];
|
|
HashMap urlCache = (HashMap) selectResult[1]; // String (url-hash) / plasmaCrawlLURL.Entry
|
|
if ((indexEntities == null) || (indexEntities.length == 0)) {
|
|
log.logFine("No index available for index transfer, hash start-point " + startPointHash);
|
|
return -1;
|
|
}
|
|
// count the indexes again, can be smaller as expected
|
|
indexCount = 0;
|
|
for (int i = 0; i < indexEntities.length; i++) {
|
|
indexCount += indexEntities[i].size();
|
|
}
|
|
|
|
// find start point for DHT-selection
|
|
String keyhash = indexEntities[indexEntities.length - 1].wordHash(); // DHT targets must have greater hashes
|
|
|
|
// iterate over DHT-peers and send away the indexes
|
|
yacySeed seed;
|
|
int hc = 0;
|
|
Enumeration e = yacyCore.dhtAgent.getAcceptRemoteIndexSeeds(keyhash);
|
|
String error;
|
|
String peerNames = "";
|
|
double avdist;
|
|
long start;
|
|
while ((e.hasMoreElements()) && (hc < peerCount)) {
|
|
if (closed) {
|
|
log.logSevere("Index distribution interrupted by close, nothing deleted locally.");
|
|
return -1; // interrupted
|
|
}
|
|
seed = (yacySeed) e.nextElement();
|
|
if ((seed != null) &&
|
|
((avdist = (yacyDHTAction.dhtDistance(seed.hash, indexEntities[0].wordHash()) +
|
|
yacyDHTAction.dhtDistance(seed.hash, indexEntities[indexEntities.length-1].wordHash())) / 2.0) < 0.3)) {
|
|
start = System.currentTimeMillis();
|
|
error = yacyClient.transferIndex(seed, indexEntities, urlCache);
|
|
if (error == null) {
|
|
log.logInfo("Index transfer of " + indexCount + " words [" + indexEntities[0].wordHash() + " .. " + indexEntities[indexEntities.length-1].wordHash() + "]/" +
|
|
avdist + " to peer " + seed.getName() + ":" + seed.hash + " in " +
|
|
((System.currentTimeMillis() - start) / 1000) + " seconds successfull (" +
|
|
(1000 * indexCount / (System.currentTimeMillis() - start + 1)) + " words/s)");
|
|
peerNames += ", " + seed.getName();
|
|
hc++;
|
|
} else {
|
|
log.logWarning("Index transfer to peer " + seed.getName() + ":" + seed.hash + " failed:'" + error + "', disconnecting peer");
|
|
yacyCore.peerActions.peerDeparture(seed);
|
|
}
|
|
}
|
|
}
|
|
if (peerNames.length() > 0) peerNames = peerNames.substring(2); // remove comma
|
|
|
|
// clean up and finish with deletion of indexes
|
|
if (hc >= peerCount) {
|
|
// success
|
|
if (delete) {
|
|
try {
|
|
if (deleteTransferIndexes(indexEntities)) {
|
|
log.logFine("Deleted all " + indexEntities.length + " transferred whole-word indexes locally");
|
|
return indexCount;
|
|
} else {
|
|
log.logSevere("Deleted not all transferred whole-word indexes");
|
|
return -1;
|
|
}
|
|
} catch (IOException ee) {
|
|
log.logSevere("Deletion of indexes not possible:" + ee.getMessage(), ee);
|
|
return -1;
|
|
}
|
|
} else {
|
|
// simply close the indexEntities
|
|
for (int i = 0; i < indexEntities.length; i++) try {
|
|
indexEntities[i].close();
|
|
} catch (IOException ee) {}
|
|
}
|
|
return indexCount;
|
|
} else {
|
|
log.logSevere("Index distribution failed. Too less peers (" + hc + ") received the index, not deleted locally.");
|
|
return -1;
|
|
}
|
|
}
|
|
|
|
private String selectTransferStart() {
|
|
String startPointHash;
|
|
// first try to select with increasing probality a good start point
|
|
if (Math.round(Math.random() * 6) != 4) for (int i = 9; i > 0; i--) {
|
|
startPointHash = serverCodings.encodeMD5B64(Long.toString(i + System.currentTimeMillis()), true).substring(2, 2 + yacySeedDB.commonHashLength);
|
|
if (yacyDHTAction.dhtDistance(yacyCore.seedDB.mySeed.hash, startPointHash) > ((double) i / (double) 10)) return startPointHash;
|
|
}
|
|
// if that fails, take simply the best start point (this is usually avoided, since that leads to always the same target peers)
|
|
startPointHash = yacyCore.seedDB.mySeed.hash.substring(0, 11) + "z";
|
|
return startPointHash;
|
|
}
|
|
|
|
Object[] /* of {plasmaWordIndexEntity[], HashMap(String, plasmaCrawlLURL.Entry)}*/
|
|
selectTransferIndexes(String hash, int count) {
|
|
// the hash is a start hash from where the indexes are picked
|
|
Vector tmpEntities = new Vector();
|
|
String nexthash = "";
|
|
try {
|
|
Iterator wordHashIterator = wordIndex.wordHashes(hash, true, true);
|
|
plasmaWordIndexEntity indexEntity, tmpEntity;
|
|
Enumeration urlEnum;
|
|
Iterator hashIter;
|
|
plasmaWordIndexEntry indexEntry;
|
|
plasmaCrawlLURL.Entry lurl;
|
|
HashSet unknownURLEntries;
|
|
HashMap knownURLs = new HashMap();
|
|
while ((count > 0) && (wordHashIterator.hasNext()) &&
|
|
((nexthash = (String) wordHashIterator.next()) != null) && (nexthash.trim().length() > 0)) {
|
|
indexEntity = wordIndex.getEntity(nexthash, true);
|
|
if (indexEntity.size() == 0) {
|
|
indexEntity.deleteComplete();
|
|
} else if ((indexEntity.size() <= count)|| // if we havn't exceeded the limit
|
|
(Math.abs(indexEntity.size() - count) <= 10)){ // or there are only at most 10 entries left
|
|
// take the whole entity
|
|
try {
|
|
// fist check if we know all urls
|
|
urlEnum = indexEntity.elements(true);
|
|
unknownURLEntries = new HashSet();
|
|
while (urlEnum.hasMoreElements()) {
|
|
indexEntry = (plasmaWordIndexEntry) urlEnum.nextElement();
|
|
lurl = urlPool.loadedURL.getEntry(indexEntry.getUrlHash());
|
|
if ((lurl == null) || (lurl.toString() == null)) {
|
|
unknownURLEntries.add(indexEntry.getUrlHash());
|
|
} else {
|
|
if (lurl.toString() == null) {
|
|
urlPool.loadedURL.remove(indexEntry.getUrlHash());
|
|
unknownURLEntries.add(indexEntry.getUrlHash());
|
|
} else {
|
|
knownURLs.put(indexEntry.getUrlHash(), lurl);
|
|
}
|
|
}
|
|
}
|
|
// now delete all entries that have no url entry
|
|
hashIter = unknownURLEntries.iterator();
|
|
while (hashIter.hasNext()) {
|
|
indexEntity.removeEntry((String) hashIter.next(), false);
|
|
}
|
|
// use whats remaining
|
|
tmpEntities.add(indexEntity);
|
|
log.logFine("Selected whole index (" + indexEntity.size() + " URLs, " + unknownURLEntries.size() + " not bound) for word " + indexEntity.wordHash());
|
|
count -= indexEntity.size();
|
|
} catch (kelondroException e) {
|
|
log.logSevere("plasmaWordIndexDistribution/1: deleted DB for word " + indexEntity.wordHash(), e);
|
|
try {indexEntity.deleteComplete();} catch (IOException ee) {}
|
|
}
|
|
} else {
|
|
// make an on-the-fly entity and insert values
|
|
tmpEntity = new plasmaWordIndexEntity(indexEntity.wordHash());
|
|
try {
|
|
urlEnum = indexEntity.elements(true);
|
|
unknownURLEntries = new HashSet();
|
|
while ((urlEnum.hasMoreElements()) && (count > 0)) {
|
|
indexEntry = (plasmaWordIndexEntry) urlEnum.nextElement();
|
|
lurl = urlPool.loadedURL.getEntry(indexEntry.getUrlHash());
|
|
if (lurl == null) {
|
|
unknownURLEntries.add(indexEntry.getUrlHash());
|
|
} else {
|
|
if (lurl.toString() == null) {
|
|
urlPool.loadedURL.remove(indexEntry.getUrlHash());
|
|
unknownURLEntries.add(indexEntry.getUrlHash());
|
|
} else {
|
|
knownURLs.put(indexEntry.getUrlHash(), lurl);
|
|
tmpEntity.addEntry(indexEntry);
|
|
count--;
|
|
}
|
|
}
|
|
}
|
|
// now delete all entries that have no url entry
|
|
hashIter = unknownURLEntries.iterator();
|
|
while (hashIter.hasNext()) {
|
|
indexEntity.removeEntry((String) hashIter.next(), true);
|
|
}
|
|
// use whats remaining
|
|
log.logFine("Selected partial index (" + tmpEntity.size() + " from " + indexEntity.size() +" URLs, " + unknownURLEntries.size() + " not bound) for word " + tmpEntity.wordHash());
|
|
tmpEntities.add(tmpEntity);
|
|
} catch (kelondroException e) {
|
|
log.logSevere("plasmaWordIndexDistribution/2: deleted DB for word " + indexEntity.wordHash(), e);
|
|
try {indexEntity.deleteComplete();} catch (IOException ee) {}
|
|
}
|
|
indexEntity.close(); // important: is not closed elswhere and cannot be deleted afterwards
|
|
indexEntity = null;
|
|
}
|
|
|
|
}
|
|
// transfer to array
|
|
plasmaWordIndexEntity[] indexEntities = new plasmaWordIndexEntity[tmpEntities.size()];
|
|
for (int i = 0; i < tmpEntities.size(); i++) indexEntities[i] = (plasmaWordIndexEntity) tmpEntities.elementAt(i);
|
|
return new Object[]{indexEntities, knownURLs};
|
|
} catch (IOException e) {
|
|
log.logSevere("selectTransferIndexes IO-Error (hash=" + nexthash + "): " + e.getMessage(), e);
|
|
return new Object[]{new plasmaWordIndexEntity[0], new HashMap()};
|
|
} catch (kelondroException e) {
|
|
log.logSevere("selectTransferIndexes database corrupted: " + e.getMessage(), e);
|
|
return new Object[]{new plasmaWordIndexEntity[0], new HashMap()};
|
|
}
|
|
}
|
|
|
|
boolean deleteTransferIndexes(plasmaWordIndexEntity[] indexEntities) throws IOException {
|
|
String wordhash;
|
|
Enumeration urlEnum;
|
|
plasmaWordIndexEntry indexEntry;
|
|
plasmaWordIndexEntity indexEntity;
|
|
String[] urlHashes;
|
|
int sz;
|
|
boolean success = true;
|
|
for (int i = 0; i < indexEntities.length; i++) {
|
|
if (indexEntities[i].isTMPEntity()) {
|
|
// delete entries separately
|
|
int c = 0;
|
|
urlHashes = new String[indexEntities[i].size()];
|
|
urlEnum = indexEntities[i].elements(true);
|
|
while (urlEnum.hasMoreElements()) {
|
|
indexEntry = (plasmaWordIndexEntry) urlEnum.nextElement();
|
|
urlHashes[c++] = indexEntry.getUrlHash();
|
|
}
|
|
wordIndex.removeEntries(indexEntities[i].wordHash(), urlHashes, true);
|
|
indexEntity = wordIndex.getEntity(indexEntities[i].wordHash(), true);
|
|
sz = indexEntity.size();
|
|
indexEntity.close();
|
|
log.logFine("Deleted partial index (" + c + " URLs) for word " + indexEntities[i].wordHash() + "; " + sz + " entries left");
|
|
// DEBUG: now try to delete the remaining index. If this works, this routine is fine
|
|
/*
|
|
if (wordIndex.getEntity(indexEntities[i].wordHash()).deleteComplete())
|
|
System.out.println("DEBUG: trial delete of partial word index " + indexEntities[i].wordHash() + " SUCCESSFULL");
|
|
else
|
|
System.out.println("DEBUG: trial delete of partial word index " + indexEntities[i].wordHash() + " FAILED");
|
|
*/
|
|
// end debug
|
|
indexEntities[i].close();
|
|
} else {
|
|
// delete complete file
|
|
if (indexEntities[i].deleteComplete()) {
|
|
indexEntities[i].close();
|
|
} else {
|
|
indexEntities[i].close();
|
|
// have another try...
|
|
if (!(plasmaWordIndexEntity.wordHash2path(wordIndex.getRoot() /*PLASMADB*/, indexEntities[i].wordHash()).delete())) {
|
|
success = false;
|
|
log.logSevere("Could not delete whole index for word " + indexEntities[i].wordHash());
|
|
}
|
|
}
|
|
}
|
|
indexEntities[i] = null;
|
|
}
|
|
return success;
|
|
}
|
|
|
|
|
|
public void startTransferWholeIndex(yacySeed seed, boolean delete) {
|
|
if (transferIdxThread == null) {
|
|
this.transferIdxThread = new transferIndexThread(seed,delete);
|
|
this.transferIdxThread.start();
|
|
}
|
|
}
|
|
|
|
public void stopTransferWholeIndex(boolean wait) {
|
|
if ((transferIdxThread != null) && (transferIdxThread.isAlive()) && (!transferIdxThread.isFinished())) {
|
|
try {
|
|
this.transferIdxThread.stopIt(wait);
|
|
} catch (InterruptedException e) { }
|
|
}
|
|
}
|
|
|
|
public void abortTransferWholeIndex(boolean wait) {
|
|
if (transferIdxThread != null) {
|
|
if (!transferIdxThread.isFinished())
|
|
try {
|
|
this.transferIdxThread.stopIt(wait);
|
|
} catch (InterruptedException e) { }
|
|
transferIdxThread = null;
|
|
}
|
|
}
|
|
|
|
|
|
public class transferIndexThread extends Thread {
|
|
private yacySeed seed = null;
|
|
private boolean delete = false;
|
|
private boolean finished = false;
|
|
private int transferedIndexCount = 0;
|
|
private String status = "Running";
|
|
private String oldStartingPointHash = "------------", startPointHash = "------------";
|
|
private int wordsDBSize = 0;
|
|
private int chunkSize = 500;
|
|
|
|
public transferIndexThread(yacySeed seed, boolean delete) {
|
|
super(new ThreadGroup("TransferIndexThreadGroup"),"TransferIndex_" + seed.getName());
|
|
this.seed = seed;
|
|
this.delete = delete;
|
|
this.wordsDBSize = plasmaSwitchboard.getSwitchboard().wordIndex.size();
|
|
}
|
|
|
|
public void run() {
|
|
performTransferWholeIndex();
|
|
}
|
|
|
|
public void stopIt(boolean wait) throws InterruptedException {
|
|
this.finished = true;
|
|
this.join();
|
|
}
|
|
|
|
public boolean isFinished() {
|
|
return this.finished;
|
|
}
|
|
|
|
public boolean deleteIndex() {
|
|
return this.delete;
|
|
}
|
|
|
|
public int getChunkSize() {
|
|
return this.chunkSize;
|
|
}
|
|
|
|
public int getTransferedIndexCount() {
|
|
return this.transferedIndexCount;
|
|
}
|
|
|
|
public float getTransferedIndexPercent() {
|
|
if (wordsDBSize == 0) return 100;
|
|
else return (float)(this.transferedIndexCount*100/wordsDBSize);
|
|
}
|
|
|
|
public yacySeed getSeed() {
|
|
return this.seed;
|
|
}
|
|
|
|
public String getStatus() {
|
|
return this.status;
|
|
}
|
|
|
|
public String getRange() {
|
|
return "[" + oldStartingPointHash + " .. " + startPointHash + "]";
|
|
}
|
|
|
|
public void performTransferWholeIndex() {
|
|
try {
|
|
plasmaWordIndexDistribution.this.paused = true;
|
|
|
|
// initial startingpoint of intex transfer is "------------"
|
|
plasmaWordIndexDistribution.this.log.logFine("Selected hash " + startPointHash + " as start point for index distribution of whole index");
|
|
|
|
/* Loop until we have
|
|
* - finished transfer of whole index
|
|
* - detected a server shutdown or user interruption
|
|
* - detected a failure
|
|
*/
|
|
long start, retryCount = 0, iteration = 0;
|
|
while (!finished && !Thread.currentThread().isInterrupted()) {
|
|
iteration++;
|
|
int idxCount = 0;
|
|
start = System.currentTimeMillis();
|
|
|
|
// selecting 500 words to transfer
|
|
this.status = "Running: Selection of chunk " + iteration;
|
|
Object[] selectResult = selectTransferIndexes(startPointHash, chunkSize);
|
|
plasmaWordIndexEntity[] indexEntities = (plasmaWordIndexEntity[]) selectResult[0];
|
|
|
|
HashMap urlCache = (HashMap) selectResult[1]; // String (url-hash) / plasmaCrawlLURL.Entry
|
|
if ((indexEntities == null) || (indexEntities.length == 0)) {
|
|
plasmaWordIndexDistribution.this.log.logFine("No index available for index transfer, hash start-point " + startPointHash);
|
|
this.status = "Finished. " + iteration + " chunks transfered.";
|
|
return;
|
|
}
|
|
// count the indexes again, can be smaller as expected
|
|
for (int i = 0; i < indexEntities.length; i++) idxCount += indexEntities[i].size();
|
|
|
|
// getting start point for next DHT-selection
|
|
oldStartingPointHash = startPointHash;
|
|
startPointHash = indexEntities[indexEntities.length - 1].wordHash(); // DHT targets must have greater hashes
|
|
|
|
plasmaWordIndexDistribution.this.log.logInfo("Index selection of " + idxCount + " words [" + indexEntities[0].wordHash() + " .. " + indexEntities[indexEntities.length-1].wordHash() + "]" +
|
|
" in " +
|
|
((System.currentTimeMillis() - start) / 1000) + " seconds (" +
|
|
(1000 * idxCount / (System.currentTimeMillis() - start + 1)) + " words/s)");
|
|
|
|
/* loop until we
|
|
* - have successfully transfered the words list or
|
|
* - the retry counter limit was exceeded
|
|
*/
|
|
start = System.currentTimeMillis();
|
|
while (true) {
|
|
// testing if we wer aborted
|
|
if (isAborted()) return;
|
|
|
|
// transfering seleted words to remote peer
|
|
this.status = "Running: Transfer of chunk " + iteration;
|
|
String error = yacyClient.transferIndex(seed, indexEntities, urlCache);
|
|
if (error == null) {
|
|
// words successfully transfered
|
|
long transferTime = System.currentTimeMillis() - start;
|
|
plasmaWordIndexDistribution.this.log.logInfo("Index transfer of " + idxCount + " words [" + indexEntities[0].wordHash() + " .. " + indexEntities[indexEntities.length-1].wordHash() + "]" +
|
|
" to peer " + seed.getName() + ":" + seed.hash + " in " + (transferTime/1000) + " seconds successfull (" +
|
|
(1000 * idxCount / (transferTime + 1)) + " words/s)");
|
|
retryCount = 0;
|
|
|
|
if (transferTime > 30000) {
|
|
if (chunkSize>100) chunkSize-=50;
|
|
} else {
|
|
chunkSize+=50;
|
|
}
|
|
|
|
break;
|
|
} else {
|
|
// worts transfer failed
|
|
|
|
// inc retry counter
|
|
retryCount++;
|
|
|
|
// testing if we were aborted ...
|
|
if (isAborted()) return;
|
|
|
|
// we have lost the connection to the remote peer. Adding peer to disconnected list
|
|
plasmaWordIndexDistribution.this.log.logWarning("Index transfer to peer " + seed.getName() + ":" + seed.hash + " failed:'" + error + "', disconnecting peer");
|
|
yacyCore.peerActions.peerDeparture(seed);
|
|
|
|
// if the retry counter limit was not exceeded we'll retry it in a few seconds
|
|
this.status = "Disconnected peer: " + ((retryCount > 5)? error + ". Transfer aborted":"Retry " + retryCount);
|
|
if (retryCount > 5) return;
|
|
Thread.sleep(retryCount*5000);
|
|
|
|
/* loop until
|
|
* - we have successfully done a peer ping or
|
|
* - the retry counter limit was exceeded
|
|
*/
|
|
while (true) {
|
|
// testing if we were aborted ...
|
|
if (isAborted()) return;
|
|
|
|
// doing a peer ping to the remote seed
|
|
int added = yacyClient.publishMySeed(seed.getAddress(), seed.hash);
|
|
if (added < 0) {
|
|
// inc. retry counter
|
|
retryCount++;
|
|
this.status = "Disconnected peer: Peer ping failed. " + ((retryCount > 5)?"Transfer aborted.":"Retry " + retryCount);
|
|
if (retryCount > 5) return;
|
|
Thread.sleep(retryCount*5000);
|
|
continue;
|
|
} else {
|
|
yacyCore.seedDB.getConnected(seed.hash);
|
|
this.status = "running";
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
// deleting transfered words from index
|
|
if (delete) {
|
|
this.status = "Running: Deletion of chunk " + iteration;
|
|
try {
|
|
if (deleteTransferIndexes(indexEntities)) {
|
|
plasmaWordIndexDistribution.this.log.logFine("Deleted all " + indexEntities.length + " transferred whole-word indexes locally");
|
|
transferedIndexCount += idxCount;
|
|
} else {
|
|
plasmaWordIndexDistribution.this.log.logSevere("Deleted not all transferred whole-word indexes");
|
|
}
|
|
} catch (IOException ee) {
|
|
plasmaWordIndexDistribution.this.log.logSevere("Deletion of indexes not possible:" + ee.getMessage(), ee);
|
|
}
|
|
} else {
|
|
// simply close the indexEntities
|
|
for (int i = 0; i < indexEntities.length; i++) try {
|
|
indexEntities[i].close();
|
|
} catch (IOException ee) {}
|
|
transferedIndexCount += idxCount;
|
|
}
|
|
}
|
|
|
|
// if we reach this point we were aborted by the user or by server shutdown
|
|
this.status = "aborted";
|
|
} catch (Exception e) {
|
|
this.status = "Error: " + e.getMessage();
|
|
plasmaWordIndexDistribution.this.log.logWarning("Index transfer to peer " + seed.getName() + ":" + seed.hash + " failed:'" + e.getMessage() + "'",e);
|
|
} finally {
|
|
plasmaWordIndexDistribution.this.paused = false;
|
|
}
|
|
}
|
|
|
|
private boolean isAborted() {
|
|
if (finished || Thread.currentThread().isInterrupted()) {
|
|
this.status = "aborted";
|
|
return true;
|
|
}
|
|
return false;
|
|
}
|
|
}
|
|
}
|