org.apache.cassandra.locator.DynamicEndpointSnitch Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of cassandra-all Show documentation
Show all versions of cassandra-all Show documentation
The Apache Cassandra Project develops a highly scalable second-generation distributed database, bringing together Dynamo's fully distributed design and Bigtable's ColumnFamily-based data model.
/**
* 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.cassandra.locator;
import java.net.InetAddress;
import java.net.UnknownHostException;
import java.util.*;
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.TimeUnit;
import com.codahale.metrics.ExponentiallyDecayingReservoir;
import com.codahale.metrics.Snapshot;
import org.apache.cassandra.concurrent.ScheduledExecutors;
import org.apache.cassandra.config.DatabaseDescriptor;
import org.apache.cassandra.net.MessagingService;
import org.apache.cassandra.service.StorageService;
import org.apache.cassandra.utils.FBUtilities;
import org.apache.cassandra.utils.MBeanWrapper;
/**
* A dynamic snitch that sorts endpoints by latency with an adapted phi failure detector
*/
public class DynamicEndpointSnitch extends AbstractEndpointSnitch implements ILatencySubscriber, DynamicEndpointSnitchMBean
{
private static final boolean USE_SEVERITY = !Boolean.getBoolean("cassandra.ignore_dynamic_snitch_severity");
private static final double ALPHA = 0.75; // set to 0.75 to make EDS more biased to towards the newer values
private static final int WINDOW_SIZE = 100;
private final int UPDATE_INTERVAL_IN_MS = DatabaseDescriptor.getDynamicUpdateInterval();
private final int RESET_INTERVAL_IN_MS = DatabaseDescriptor.getDynamicResetInterval();
private final double BADNESS_THRESHOLD = DatabaseDescriptor.getDynamicBadnessThreshold();
// the score for a merged set of endpoints must be this much worse than the score for separate endpoints to
// warrant not merging two ranges into a single range
private double RANGE_MERGING_PREFERENCE = 1.5;
private String mbeanName;
private boolean registered = false;
private volatile HashMap scores = new HashMap<>();
private final ConcurrentHashMap samples = new ConcurrentHashMap<>();
public final IEndpointSnitch subsnitch;
public DynamicEndpointSnitch(IEndpointSnitch snitch)
{
this(snitch, null);
}
public DynamicEndpointSnitch(IEndpointSnitch snitch, String instance)
{
mbeanName = "org.apache.cassandra.db:type=DynamicEndpointSnitch";
if (instance != null)
mbeanName += ",instance=" + instance;
subsnitch = snitch;
Runnable update = new Runnable()
{
public void run()
{
updateScores();
}
};
Runnable reset = new Runnable()
{
public void run()
{
// we do this so that a host considered bad has a chance to recover, otherwise would we never try
// to read from it, which would cause its score to never change
reset();
}
};
ScheduledExecutors.scheduledTasks.scheduleWithFixedDelay(update, UPDATE_INTERVAL_IN_MS, UPDATE_INTERVAL_IN_MS, TimeUnit.MILLISECONDS);
ScheduledExecutors.scheduledTasks.scheduleWithFixedDelay(reset, RESET_INTERVAL_IN_MS, RESET_INTERVAL_IN_MS, TimeUnit.MILLISECONDS);
registerMBean();
}
private void registerMBean()
{
MBeanWrapper.instance.registerMBean(this, mbeanName);
}
public void unregisterMBean()
{
MBeanWrapper.instance.unregisterMBean(mbeanName);
}
@Override
public void gossiperStarting()
{
subsnitch.gossiperStarting();
}
public String getRack(InetAddress endpoint)
{
return subsnitch.getRack(endpoint);
}
public String getDatacenter(InetAddress endpoint)
{
return subsnitch.getDatacenter(endpoint);
}
public List getSortedListByProximity(final InetAddress address, Collection addresses)
{
List list = new ArrayList(addresses);
sortByProximity(address, list);
return list;
}
@Override
public void sortByProximity(final InetAddress address, List addresses)
{
assert address.equals(FBUtilities.getBroadcastAddress()); // we only know about ourself
if (BADNESS_THRESHOLD == 0)
{
sortByProximityWithScore(address, addresses);
}
else
{
sortByProximityWithBadness(address, addresses);
}
}
private void sortByProximityWithScore(final InetAddress address, List addresses)
{
// Scores can change concurrently from a call to this method. But Collections.sort() expects
// its comparator to be "stable", that is 2 endpoint should compare the same way for the duration
// of the sort() call. As we copy the scores map on write, it is thus enough to alias the current
// version of it during this call.
final HashMap scores = this.scores;
Collections.sort(addresses, new Comparator()
{
public int compare(InetAddress a1, InetAddress a2)
{
return compareEndpoints(address, a1, a2, scores);
}
});
}
private void sortByProximityWithBadness(final InetAddress address, List addresses)
{
if (addresses.size() < 2)
return;
subsnitch.sortByProximity(address, addresses);
HashMap scores = this.scores; // Make sure the score don't change in the middle of the loop below
// (which wouldn't really matter here but its cleaner that way).
ArrayList subsnitchOrderedScores = new ArrayList<>(addresses.size());
for (InetAddress inet : addresses)
{
Double score = scores.get(inet);
if (score == null)
continue;
subsnitchOrderedScores.add(score);
}
// Sort the scores and then compare them (positionally) to the scores in the subsnitch order.
// If any of the subsnitch-ordered scores exceed the optimal/sorted score by BADNESS_THRESHOLD, use
// the score-sorted ordering instead of the subsnitch ordering.
ArrayList sortedScores = new ArrayList<>(subsnitchOrderedScores);
Collections.sort(sortedScores);
Iterator sortedScoreIterator = sortedScores.iterator();
for (Double subsnitchScore : subsnitchOrderedScores)
{
if (subsnitchScore > (sortedScoreIterator.next() * (1.0 + BADNESS_THRESHOLD)))
{
sortByProximityWithScore(address, addresses);
return;
}
}
}
// Compare endpoints given an immutable snapshot of the scores
private int compareEndpoints(InetAddress target, InetAddress a1, InetAddress a2, Map scores)
{
Double scored1 = scores.get(a1);
Double scored2 = scores.get(a2);
if (scored1 == null)
{
scored1 = 0.0;
}
if (scored2 == null)
{
scored2 = 0.0;
}
if (scored1.equals(scored2))
return subsnitch.compareEndpoints(target, a1, a2);
if (scored1 < scored2)
return -1;
else
return 1;
}
public int compareEndpoints(InetAddress target, InetAddress a1, InetAddress a2)
{
// That function is fundamentally unsafe because the scores can change at any time and so the result of that
// method is not stable for identical arguments. This is why we don't rely on super.sortByProximity() in
// sortByProximityWithScore().
throw new UnsupportedOperationException("You shouldn't wrap the DynamicEndpointSnitch (within itself or otherwise)");
}
public void receiveTiming(InetAddress host, long latency) // this is cheap
{
ExponentiallyDecayingReservoir sample = samples.get(host);
if (sample == null)
{
ExponentiallyDecayingReservoir maybeNewSample = new ExponentiallyDecayingReservoir(WINDOW_SIZE, ALPHA);
sample = samples.putIfAbsent(host, maybeNewSample);
if (sample == null)
sample = maybeNewSample;
}
sample.update(latency);
}
private void updateScores() // this is expensive
{
if (!StorageService.instance.isInitialized())
return;
if (!registered)
{
if (MessagingService.instance() != null)
{
MessagingService.instance().register(this);
registered = true;
}
}
double maxLatency = 1;
Map snapshots = new HashMap<>(samples.size());
for (Map.Entry entry : samples.entrySet())
{
snapshots.put(entry.getKey(), entry.getValue().getSnapshot());
}
// We're going to weight the latency for each host against the worst one we see, to
// arrive at sort of a 'badness percentage' for them. First, find the worst for each:
HashMap newScores = new HashMap<>();
for (Map.Entry entry : snapshots.entrySet())
{
double mean = entry.getValue().getMedian();
if (mean > maxLatency)
maxLatency = mean;
}
// now make another pass to do the weighting based on the maximums we found before
for (Map.Entry entry : snapshots.entrySet())
{
double score = entry.getValue().getMedian() / maxLatency;
// finally, add the severity without any weighting, since hosts scale this relative to their own load and the size of the task causing the severity.
// "Severity" is basically a measure of compaction activity (CASSANDRA-3722).
if (USE_SEVERITY)
score += StorageService.instance.getSeverity(entry.getKey());
// lowest score (least amount of badness) wins.
newScores.put(entry.getKey(), score);
}
scores = newScores;
}
private void reset()
{
samples.clear();
}
public Map getScores()
{
return scores;
}
public int getUpdateInterval()
{
return UPDATE_INTERVAL_IN_MS;
}
public int getResetInterval()
{
return RESET_INTERVAL_IN_MS;
}
public double getBadnessThreshold()
{
return BADNESS_THRESHOLD;
}
public String getSubsnitchClassName()
{
return subsnitch.getClass().getName();
}
public List dumpTimings(String hostname) throws UnknownHostException
{
InetAddress host = InetAddress.getByName(hostname);
ArrayList timings = new ArrayList();
ExponentiallyDecayingReservoir sample = samples.get(host);
if (sample != null)
{
for (double time: sample.getSnapshot().getValues())
timings.add(time);
}
return timings;
}
public void setSeverity(double severity)
{
StorageService.instance.reportManualSeverity(severity);
}
public double getSeverity()
{
return StorageService.instance.getSeverity(FBUtilities.getBroadcastAddress());
}
public boolean isWorthMergingForRangeQuery(List merged, List l1, List l2)
{
if (!subsnitch.isWorthMergingForRangeQuery(merged, l1, l2))
return false;
// skip checking scores in the single-node case
if (l1.size() == 1 && l2.size() == 1 && l1.get(0).equals(l2.get(0)))
return true;
// Make sure we return the subsnitch decision (i.e true if we're here) if we lack too much scores
double maxMerged = maxScore(merged);
double maxL1 = maxScore(l1);
double maxL2 = maxScore(l2);
if (maxMerged < 0 || maxL1 < 0 || maxL2 < 0)
return true;
return maxMerged <= (maxL1 + maxL2) * RANGE_MERGING_PREFERENCE;
}
// Return the max score for the endpoint in the provided list, or -1.0 if no node have a score.
private double maxScore(List endpoints)
{
double maxScore = -1.0;
for (InetAddress endpoint : endpoints)
{
Double score = scores.get(endpoint);
if (score == null)
continue;
if (score > maxScore)
maxScore = score;
}
return maxScore;
}
}
© 2015 - 2024 Weber Informatics LLC | Privacy Policy