* Netdb Search:

- Follow all DSRM's, not just the last one, by moving the code
      from the match job to the selector
    - Update peer profile after SingleSearchJob
    - Cleanups, javadocs, log tweaks, final
This commit is contained in:
zzz
2011-07-29 20:43:52 +00:00
parent e7f449472c
commit dcc1861399
11 changed files with 102 additions and 66 deletions

View File

@ -1,3 +1,11 @@
2011-07-29 zzz
* Netdb Search:
- Follow all DSRM's, not just the last one, by moving the code
from the match job to the selector
- Update peer profile after SingleSearchJob
- Cleanups, javadocs, log tweaks, final
* ProfileOrganizer: Tweak fast tier size
2011-07-28 zzz
* Context: Split up big lock to avoid deadlocks
* Streaming: Avoid a rare exception on race

View File

@ -18,7 +18,7 @@ public class RouterVersion {
/** deprecated */
public final static String ID = "Monotone";
public final static String VERSION = CoreVersion.VERSION;
public final static long BUILD = 16;
public final static long BUILD = 17;
/** for example "-test" */
public final static String EXTRA = "";

View File

@ -14,7 +14,6 @@ import net.i2p.util.Log;
class FloodOnlyLookupMatchJob extends JobImpl implements ReplyJob {
private final Log _log;
private final FloodOnlySearchJob _search;
private DatabaseSearchReplyMessage _dsrm;
public FloodOnlyLookupMatchJob(RouterContext ctx, FloodOnlySearchJob job) {
super(ctx);
@ -28,19 +27,8 @@ class FloodOnlyLookupMatchJob extends JobImpl implements ReplyJob {
_log.info(_search.getJobId() + ": search match and found locally");
_search.success();
} else {
int remaining = _search.getLookupsRemaining();
if (_log.shouldLog(Log.INFO))
_log.info(_search.getJobId() + ": got a DatabaseSearchReply when we were looking for "
+ _search.getKey().toBase64() + ", with " + remaining + " outstanding searches");
// netDb reply pointing us at other people
// Only process if we don't know enough floodfills
// This only works if both reply, otherwise we aren't called - should be fixed
if (_search.shouldProcessDSRM() && _dsrm != null) {
if (_log.shouldLog(Log.INFO))
_log.info(_search.getJobId() + ": Processing DatabaseSearchReply");
// Chase the hashes from the reply
getContext().jobQueue().addJob(new SingleLookupJob(getContext(), _dsrm));
}
// In practice, we always have zero remaining when this is called,
// because the selector only returns true when there is zero remaining
_search.failed();
}
}
@ -49,11 +37,9 @@ class FloodOnlyLookupMatchJob extends JobImpl implements ReplyJob {
public void setMessage(I2NPMessage message) {
if (message instanceof DatabaseSearchReplyMessage) {
// DSRM processing now in FloodOnlyLookupSelector instead of here,
// a dsrm is only passed in when there are no more lookups remaining
// If more than one peer sent one, we only process the last one
// And sadly if the first peer sends a DRSM and the second one times out,
// this won't get called...
_dsrm = (DatabaseSearchReplyMessage) message;
// so that all DSRM's are processed, not just the last one.
_search.failed();
return;
}

View File

@ -38,14 +38,32 @@ class FloodOnlyLookupSelector implements MessageSelector {
} else if (message instanceof DatabaseSearchReplyMessage) {
DatabaseSearchReplyMessage dsrm = (DatabaseSearchReplyMessage)message;
if (_search.getKey().equals(dsrm.getSearchKey())) {
_search.decrementRemaining(dsrm.getFromHash());
// assume 0 old, all new, 0 invalid, 0 dup
// TODO - dsrm.getFromHash() can't be trusted - check against the list of
// those we sent the search to in _search ?
// assume 0 new, all old, 0 invalid, 0 dup
_context.profileManager().dbLookupReply(dsrm.getFromHash(), 0, dsrm.getNumReplies(), 0, 0,
System.currentTimeMillis()-_search.getCreated());
if (_search.getLookupsRemaining() <= 0)
return true; // ok, no more left, so time to fail
else
return false;
// Moved from FloodOnlyLookupMatchJob so it is called for all replies
// rather than just the last one
// Got a netDb reply pointing us at other floodfills...
// Only process if we don't know enough floodfills or are starting up
if (_search.shouldProcessDSRM()) {
if (_log.shouldLog(Log.INFO))
_log.info(_search.getJobId() + ": Processing DSRM via SingleLookupJob, apparently from " + dsrm.getFromHash());
// Chase the hashes from the reply
_context.jobQueue().addJob(new SingleLookupJob(_context, dsrm));
} else if (_log.shouldLog(Log.INFO)) {
int remaining = _search.getLookupsRemaining();
_log.info(_search.getJobId() + ": got a DSRM apparently from " + dsrm.getFromHash() + " when we were looking for "
+ _search.getKey() + ", with " + remaining + " outstanding searches");
}
// if no more left, time to fail
int remaining = _search.decrementRemaining(dsrm.getFromHash());
return remaining <= 0;
}
}
return false;

View File

@ -6,17 +6,16 @@ import net.i2p.util.Log;
class FloodOnlyLookupTimeoutJob extends JobImpl {
private final FloodSearchJob _search;
private final Log _log;
public FloodOnlyLookupTimeoutJob(RouterContext ctx, FloodOnlySearchJob job) {
super(ctx);
_search = job;
_log = ctx.logManager().getLog(getClass());
}
public void runJob() {
if (_log.shouldLog(Log.INFO))
_log.info(_search.getJobId() + ": search timed out");
Log log = getContext().logManager().getLog(getClass());
if (log.shouldLog(Log.INFO))
log.info(_search.getJobId() + ": search timed out");
_search.failed();
}

View File

@ -38,24 +38,24 @@ import net.i2p.util.Log;
*/
class FloodOnlySearchJob extends FloodSearchJob {
private volatile boolean _dead;
private final long _created;
protected final long _created;
private boolean _shouldProcessDSRM;
private final HashSet<Hash> _unheardFrom;
private final List<OutNetMessage> _out;
/** this is a marker to register with the MessageRegistry, it is never sent */
private OutNetMessage _out;
protected final MessageSelector _replySelector;
protected final ReplyJob _onReply;
protected final Job _onTimeout;
private static final int MIN_FOR_NO_DSRM = 4;
public FloodOnlySearchJob(RouterContext ctx, FloodfillNetworkDatabaseFacade facade, Hash key, Job onFind, Job onFailed, int timeoutMs, boolean isLease) {
super(ctx, facade, key, onFind, onFailed, timeoutMs, isLease);
// these override the settings in super
_log = ctx.logManager().getLog(FloodOnlySearchJob.class);
_timeoutMs = Math.min(timeoutMs, SearchJob.PER_FLOODFILL_PEER_TIMEOUT);
_expiration = _timeoutMs + ctx.clock().now();
_origExpiration = _timeoutMs + ctx.clock().now();
// do we need a synchronizedList, since we synch on _out everywhere below...
_out = Collections.synchronizedList(new ArrayList(2));
_unheardFrom = new HashSet(CONCURRENT_SEARCHES);
_replySelector = new FloodOnlyLookupSelector(getContext(), this);
_onReply = new FloodOnlyLookupMatchJob(getContext(), this);
@ -63,10 +63,10 @@ class FloodOnlySearchJob extends FloodSearchJob {
_created = System.currentTimeMillis();
}
/** System time, NOT context time */
public long getCreated() { return _created; }
public boolean shouldProcessDSRM() { return _shouldProcessDSRM; }
private static final int CONCURRENT_SEARCHES = 2;
private static final int MIN_FOR_NO_DSRM = 4;
@Override
public void runJob() {
@ -105,8 +105,10 @@ class FloodOnlySearchJob extends FloodSearchJob {
}
Collections.shuffle(floodfillPeers, getContext().random());
}
OutNetMessage out = getContext().messageRegistry().registerPending(_replySelector, _onReply, _onTimeout, _timeoutMs);
synchronized (_out) { _out.add(out); }
// This OutNetMessage is never used or sent (setMessage() is never called), it's only
// so we can register a reply selector.
_out = getContext().messageRegistry().registerPending(_replySelector, _onReply, _onTimeout, _timeoutMs);
/********
// We need to randomize our ff selection, else we stay with the same ones since
@ -194,11 +196,15 @@ class FloodOnlySearchJob extends FloodSearchJob {
@Override
public String getName() { return "NetDb flood search (phase 1)"; }
/** Note that we heard from the peer */
void decrementRemaining(Hash peer) {
decrementRemaining();
/**
* Note that we heard from the peer
*
* @return number remaining after decrementing
*/
int decrementRemaining(Hash peer) {
synchronized(_unheardFrom) {
_unheardFrom.remove(peer);
return decrementRemaining();
}
}
@ -208,12 +214,7 @@ class FloodOnlySearchJob extends FloodSearchJob {
if (_dead) return;
_dead = true;
}
List outBuf = null;
synchronized (_out) { outBuf = new ArrayList(_out); }
for (int i = 0; i < outBuf.size(); i++) {
OutNetMessage out = (OutNetMessage)outBuf.get(i);
getContext().messageRegistry().unregisterPending(out);
}
getContext().messageRegistry().unregisterPending(_out);
int timeRemaining = (int)(_origExpiration - getContext().clock().now());
if (_log.shouldLog(Log.INFO))
_log.info(getJobId() + ": Floodfill search for " + _key.toBase64() + " failed with " + timeRemaining + " remaining after " + (System.currentTimeMillis()-_created));

View File

@ -29,7 +29,7 @@ import net.i2p.util.Log;
* Note that this does NOT extend SearchJob.
*/
public class FloodSearchJob extends JobImpl {
protected Log _log;
protected final Log _log;
protected final FloodfillNetworkDatabaseFacade _facade;
protected final Hash _key;
protected final List<Job> _onFind;
@ -43,24 +43,22 @@ public class FloodSearchJob extends JobImpl {
public FloodSearchJob(RouterContext ctx, FloodfillNetworkDatabaseFacade facade, Hash key, Job onFind, Job onFailed, int timeoutMs, boolean isLease) {
super(ctx);
_log = ctx.logManager().getLog(FloodSearchJob.class);
_log = ctx.logManager().getLog(getClass());
_facade = facade;
_key = key;
_onFind = new ArrayList();
_onFind = new ArrayList(4);
_onFind.add(onFind);
_onFailed = new ArrayList();
_onFailed = new ArrayList(4);
_onFailed.add(onFailed);
int timeout = -1;
timeout = timeoutMs / FLOOD_SEARCH_TIME_FACTOR;
int timeout = timeoutMs / FLOOD_SEARCH_TIME_FACTOR;
if (timeout < timeoutMs)
timeout = timeoutMs;
_timeoutMs = timeout;
_expiration = timeout + ctx.clock().now();
_origExpiration = timeoutMs + ctx.clock().now();
_isLease = isLease;
_lookupsRemaining = 0;
_dead = false;
}
void addDeferred(Job onFind, Job onFailed, long timeoutMs, boolean isLease) {
if (_dead) {
getContext().jobQueue().addJob(onFailed);
@ -69,10 +67,14 @@ public class FloodSearchJob extends JobImpl {
if (onFailed != null) synchronized (_onFailed) { _onFailed.add(onFailed); }
}
}
/** using context clock */
public long getExpiration() { return _expiration; }
private static final int CONCURRENT_SEARCHES = 2;
protected static final int CONCURRENT_SEARCHES = 2;
private static final int FLOOD_SEARCH_TIME_FACTOR = 2;
private static final int FLOOD_SEARCH_TIME_MIN = 30*1000;
public void runJob() {
// pick some floodfill peers and send out the searches
List floodfillPeers = _facade.getFloodfillPeers();
@ -120,10 +122,21 @@ public class FloodSearchJob extends JobImpl {
_facade.searchFull(_key, _onFind, _onFailed, _timeoutMs*FLOOD_SEARCH_TIME_FACTOR, _isLease);
}
}
public String getName() { return "NetDb search (phase 1)"; }
protected Hash getKey() { return _key; }
protected void decrementRemaining() { if (_lookupsRemaining > 0) _lookupsRemaining--; }
/**
* TODO AtomicInteger?
* @return number remaining after decrementing
*/
protected int decrementRemaining() {
if (_lookupsRemaining > 0)
return (--_lookupsRemaining);
return 0;
}
protected int getLookupsRemaining() { return _lookupsRemaining; }
void failed() {
@ -144,6 +157,7 @@ public class FloodSearchJob extends JobImpl {
getContext().jobQueue().addJob(removed.remove(0));
}
}
void success() {
if (_dead) return;
if (_log.shouldLog(Log.INFO))
@ -166,8 +180,8 @@ public class FloodSearchJob extends JobImpl {
_search = job;
}
public void runJob() {
_search.decrementRemaining();
if (_search.getLookupsRemaining() <= 0)
int remaining = _search.decrementRemaining();
if (remaining <= 0)
_search.failed();
}
public String getName() { return "NetDb search (phase 1) timeout"; }

View File

@ -33,7 +33,7 @@ import net.i2p.util.Log;
*/
public class FloodfillNetworkDatabaseFacade extends KademliaNetworkDatabaseFacade {
public static final char CAPABILITY_FLOODFILL = 'f';
private final Map _activeFloodQueries;
private final Map<Hash, FloodSearchJob> _activeFloodQueries;
private boolean _floodfillEnabled;
/** for testing, see isFloodfill() below */
private static String _alwaysQuery;
@ -250,7 +250,7 @@ public class FloodfillNetworkDatabaseFacade extends KademliaNetworkDatabaseFacad
boolean isNew = false;
FloodSearchJob searchJob = null;
synchronized (_activeFloodQueries) {
searchJob = (FloodSearchJob)_activeFloodQueries.get(key);
searchJob = _activeFloodQueries.get(key);
if (searchJob == null) {
//if (SearchJob.onlyQueryFloodfillPeers(_context)) {
searchJob = new FloodOnlySearchJob(_context, this, key, onFindJob, onFailedLookupJob, (int)timeoutMs, isLease);
@ -325,6 +325,7 @@ public class FloodfillNetworkDatabaseFacade extends KademliaNetworkDatabaseFacad
}
}
}
void complete(Hash key) {
synchronized (_activeFloodQueries) { _activeFloodQueries.remove(key); }
}

View File

@ -43,6 +43,7 @@ class FloodfillPeerSelector extends PeerSelector {
* Puts the floodfill peers that are directly connected first in the list.
* List will not include our own hash.
*
* @param key the ROUTING key (NOT the original key)
* @param peersToIgnore can be null
* @return List of Hash for the peers selected
*/
@ -57,6 +58,7 @@ class FloodfillPeerSelector extends PeerSelector {
* Does not prefer the floodfill peers that are directly connected.
* List will not include our own hash.
*
* @param key the ROUTING key (NOT the original key)
* @param peersToIgnore can be null
* @return List of Hash for the peers selected
*/
@ -70,6 +72,7 @@ class FloodfillPeerSelector extends PeerSelector {
* after they're complete, sort via kademlia.
* List will not include our own hash.
*
* @param key the ROUTING key (NOT the original key)
* @param peersToIgnore can be null
* @return List of Hash for the peers selected
*/
@ -133,7 +136,7 @@ class FloodfillPeerSelector extends PeerSelector {
* List will not include our own hash.
*
* @return floodfills closest to the key that are not shitlisted forever
* @param key the routing key
* @param key the ROUTING key (NOT the original key)
* @param maxNumRouters max to return
* Sorted by closest to the key if > maxNumRouters, otherwise not
* The list is in 3 groups - sorted by routing key within each group.
@ -159,6 +162,7 @@ class FloodfillPeerSelector extends PeerSelector {
/**
* See above for description
* List will not include our own hash
* @param key the ROUTING key (NOT the original key)
* @param toIgnore can be null
*/
List<Hash> selectFloodfillParticipants(Hash key, int howMany, Set<Hash> toIgnore, KBucketSet kbuckets) {
@ -175,6 +179,7 @@ class FloodfillPeerSelector extends PeerSelector {
/**
* See above for description
* List MAY CONTAIN our own hash unless included in toIgnore
* @param key the ROUTING key (NOT the original key)
* @param toIgnore can be null
*/
private List<Hash> selectFloodfillParticipantsIncludingUs(Hash key, int howMany, Set<Hash> toIgnore, KBucketSet kbuckets) {
@ -271,6 +276,7 @@ class FloodfillPeerSelector extends PeerSelector {
/**
* Warning - may return our router hash - add to toIgnore if necessary
* @param key the ROUTING key (NOT the original key)
* @param toIgnore can be null
*/
public FloodfillSelectionCollector(Hash key, Set<Hash> toIgnore, int wanted) {

View File

@ -5,7 +5,7 @@ import net.i2p.data.RouterInfo;
import net.i2p.data.i2np.DatabaseSearchReplyMessage;
import net.i2p.router.JobImpl;
import net.i2p.router.RouterContext;
import net.i2p.util.Log;
//import net.i2p.util.Log;
/**
* Ask the peer who sent us the DSRM for the RouterInfos.
@ -20,12 +20,12 @@ import net.i2p.util.Log;
*
*/
class SingleLookupJob extends JobImpl {
private final Log _log;
//private final Log _log;
private final DatabaseSearchReplyMessage _dsrm;
public SingleLookupJob(RouterContext ctx, DatabaseSearchReplyMessage dsrm) {
super(ctx);
_log = ctx.logManager().getLog(getClass());
//_log = ctx.logManager().getLog(getClass());
_dsrm = dsrm;
}

View File

@ -56,8 +56,11 @@ class SingleSearchJob extends FloodOnlySearchJob {
@Override
void failed() {
getContext().messageRegistry().unregisterPending(_onm);
getContext().profileManager().dbLookupFailed(_to);
}
@Override
void success() {}
void success() {
getContext().profileManager().dbLookupSuccessful(_to, System.currentTimeMillis()-_created);
}
}