Skip to content
Merged
Show file tree
Hide file tree
Changes from 20 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
Expand Up @@ -75,6 +75,7 @@
public class TestDynamicClient extends D2BaseTest {
private static final String D2_CONFIG_FILE = "d2_config_example.json";
private static final String ZK_HOST = "127.0.0.1";
private static final String ECHO_SERVER_HOST = "127.0.0.1";
private static final int ECHO_SERVER_PORT_START = 2851;
private static final int NUMBER_OF_HOSTS = 5;
private static final int NUMBER_OF_THREADS = 10;
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -40,7 +40,7 @@ public class MockClusterInfoProvider implements ClusterInfoProvider
public int getClusterCount(String clusterName, String scheme, int partitionId)
throws ServiceUnavailableException
{
return 0;
return clusterHttpsCount.getOrDefault(clusterName, 1);
}

@Override
Expand All @@ -50,6 +50,7 @@ public int getHttpsClusterCount(String clusterName)
return clusterHttpsCount.getOrDefault(clusterName, 1);
}


@Override
public DarkClusterConfigMap getDarkClusterConfigMap(String clusterName)
throws ServiceUnavailableException
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -27,9 +27,10 @@ public interface DarkClusterStrategyFactory
/**
* get retrieves the {@link DarkClusterStrategy} corresponding to the darkClusterName.
* @param darkClusterName darkClusterName to look up
* @param partitionId partition id to scope the strategy to
* @return {@link DarkClusterStrategy}
*/
DarkClusterStrategy get(String darkClusterName);
DarkClusterStrategy get(String darkClusterName, int partitionId);

/**
* Do any actions necessary to start the DarkClusterStrategyFactory.
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -55,12 +55,15 @@ public class ConstantQpsDarkClusterStrategy implements DarkClusterStrategy
private final ClusterInfoProvider _clusterInfoProvider;
private final ConstantQpsRateLimiter _rateLimiter;



private static final long ONE_SECOND_PERIOD = TimeUnit.SECONDS.toMillis(1);
private static final int NUM_REQUESTS_TO_SEND_PER_RATE_LIMITER_CYCLE = 1;

public ConstantQpsDarkClusterStrategy(@Nonnull String originalClusterName, @Nonnull String darkClusterName,
@Nonnull Float darkClusterPerHostQps, @Nonnull BaseDarkClusterDispatcher baseDarkClusterDispatcher,
@Nonnull Notifier notifier, @Nonnull ClusterInfoProvider clusterInfoProvider, @Nonnull ConstantQpsRateLimiter rateLimiter)
@Nonnull Notifier notifier, @Nonnull ClusterInfoProvider clusterInfoProvider,
@Nonnull ConstantQpsRateLimiter rateLimiter)
{
_originalClusterName = originalClusterName;
_darkClusterName = darkClusterName;
Expand All @@ -74,6 +77,7 @@ public ConstantQpsDarkClusterStrategy(@Nonnull String originalClusterName, @Nonn
@Override
public boolean handleRequest(RestRequest originalRequest, RestRequest darkRequest, RequestContext requestContext)
{
// ClusterInfoProvider is already partition-aware, so we can use it directly
float sendRate = getSendRate();
// set burst in such a way that requests are dispatched evenly across the ONE_SECOND_PERIOD
int burst = (int) Math.max(1, Math.ceil(sendRate / ONE_SECOND_PERIOD));
Expand Down Expand Up @@ -134,7 +138,6 @@ private float getSendRate()
{
return (numDarkClusterInstances * _darkClusterPerHostQps) / numSourceClusterInstances;
}

return 0F;
}
catch (ServiceUnavailableException e)
Expand Down Expand Up @@ -171,4 +174,6 @@ public void onSuccess(None result)
});
return true;
}


}
Original file line number Diff line number Diff line change
Expand Up @@ -144,13 +144,14 @@ public boolean handleDarkRequest(RestRequest originalRequest, RequestContext ori
RestRequest reqCopy = originalRequest.builder().build();
RequestContext newRequestContext = new RequestContext(originalRequestContext);
DarkClusterConfigMap configMap = _facilities.getClusterInfoProvider().getDarkClusterConfigMap(_sourceClusterName);
int partitionId = getPartitionId(originalRequest);
for (String darkClusterName : configMap.keySet())
{
if (_darkGateKeeper.shouldDispatchToDark(originalRequest, originalRequestContext, darkClusterName))
{
RestRequest newD2Request = rewriteRequest(reqCopy, darkClusterName);
// now find the strategy appropriate for each dark cluster
DarkClusterStrategy strategy = _darkClusterStrategyFactory.get(darkClusterName);
DarkClusterStrategy strategy = _darkClusterStrategyFactory.get(darkClusterName, partitionId);
darkRequestSent |= strategy.handleRequest(reqCopy, newD2Request, newRequestContext);
}
}
Expand All @@ -164,6 +165,20 @@ public boolean handleDarkRequest(RestRequest originalRequest, RequestContext ori
return darkRequestSent;
}

private int getPartitionId(RestRequest request)
{
try
{
String serviceName = com.linkedin.d2.balancer.util.LoadBalancerUtil.getServiceNameFromUri(request.getURI());
com.linkedin.d2.balancer.util.partitions.PartitionAccessor accessor = _facilities.getPartitionInfoProvider().getPartitionAccessor(serviceName);
return accessor.getPartitionId(request.getURI());
}
catch (Throwable t)
{
return com.linkedin.d2.balancer.util.partitions.DefaultPartitionAccessor.DEFAULT_PARTITION_ID;
}
}

/**
* isSafe returns true if the underlying HttpMethod has the expectation of only doing retrieval with no side effects. For further details,
* see {@link HttpMethod}
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,8 @@
import java.util.Random;
import java.util.Set;
import java.util.concurrent.ConcurrentHashMap;
import com.linkedin.d2.balancer.ServiceUnavailableException;
import com.linkedin.d2.balancer.clusterfailout.FailoutConfig;

import java.util.function.Supplier;
import javax.annotation.Nonnull;
Expand Down Expand Up @@ -62,7 +64,9 @@ public class DarkClusterStrategyFactoryImpl implements DarkClusterStrategyFactor
private final DarkClusterDispatcher _darkClusterDispatcher;
private final Notifier _notifier;

private final Map<String, DarkClusterStrategy> _darkStrategyMap;
// Map of partition ID to dark cluster name to list of dark cluster strategies for that partition
private final Map<Integer, Map<String, DarkClusterStrategy>> _partitionToDarkStrategyMap;
private volatile boolean _sourceClusterPresent = false;
private final Random _random;
private final LoadBalancerClusterListener _clusterListener;
private final DarkClusterVerifierManager _verifierManager;
Expand All @@ -79,7 +83,7 @@ public DarkClusterStrategyFactoryImpl(@Nonnull Facilities facilities,
_facilities = facilities;
_sourceClusterName = sourceClusterName;
_notifier = notifier;
_darkStrategyMap = new ConcurrentHashMap<>();
_partitionToDarkStrategyMap = new ConcurrentHashMap<>();
_random = random;
_darkClusterDispatcher = darkClusterDispatcher;
_verifierManager = verifierManager;
Expand Down Expand Up @@ -138,16 +142,44 @@ public void shutdown()
* @return darkClusterStrategy to use.
*/
@Override
public DarkClusterStrategy get(@Nonnull String darkClusterName)
public DarkClusterStrategy get(@Nonnull String darkClusterName, int partitionId)
{
return _darkStrategyMap.getOrDefault(darkClusterName, NO_OP_DARK_CLUSTER_STRATEGY);
Map<String, DarkClusterStrategy> darkMap = _partitionToDarkStrategyMap.computeIfAbsent(partitionId, k -> new ConcurrentHashMap<>());
DarkClusterStrategy existing = darkMap.get(darkClusterName);

if (existing != null)
{
return existing;
}

try
{
// Lazily create the strategy if it doesn't exist.
DarkClusterConfigMap darkClusterConfigMap = _facilities.getClusterInfoProvider().getDarkClusterConfigMap(_sourceClusterName);
if (darkClusterConfigMap != null && darkClusterConfigMap.containsKey(darkClusterName))
{
DarkClusterConfig config = darkClusterConfigMap.get(darkClusterName);
DarkClusterStrategy strategy = createStrategy(darkClusterName, config, partitionId);
darkMap.put(darkClusterName, strategy);
return strategy;
}
}
catch (Throwable t)
{
LOG.warn("Unable to get DarkClusterConfigMap for source cluster: " + _sourceClusterName, t);
}
return NO_OP_DARK_CLUSTER_STRATEGY;
}

/**
* In the future, additional strategies can be added, and the logic here can choose the appropriate one based on the config values.
*/
private DarkClusterStrategy createStrategy(String darkClusterName, DarkClusterConfig darkClusterConfig)
private DarkClusterStrategy createStrategy(String darkClusterName, DarkClusterConfig darkClusterConfig, int partitionId)
{
// Create partition-aware ClusterInfoProvider that filters cluster information for this specific partition
com.linkedin.d2.balancer.util.ClusterInfoProvider partitionAwareProvider =
new PartitionAwareClusterInfoProvider(_facilities.getClusterInfoProvider(), partitionId);

if (darkClusterConfig.hasDarkClusterStrategyPrioritizedList())
{
DarkClusterStrategyNameArray strategyList = darkClusterConfig.getDarkClusterStrategyPrioritizedList();
Expand All @@ -162,7 +194,8 @@ private DarkClusterStrategy createStrategy(String darkClusterName, DarkClusterCo
new BaseDarkClusterDispatcherImpl(darkClusterName, _darkClusterDispatcher, _notifier, _verifierManager);
return new RelativeTrafficMultiplierDarkClusterStrategy(_sourceClusterName, darkClusterName,
darkClusterConfig.getMultiplier(), baseDarkClusterDispatcher,
_notifier, _facilities.getClusterInfoProvider(), _random);
_notifier, partitionAwareProvider,
_random);
}
break;
case IDENTICAL_TRAFFIC:
Expand All @@ -172,7 +205,8 @@ private DarkClusterStrategy createStrategy(String darkClusterName, DarkClusterCo
new BaseDarkClusterDispatcherImpl(darkClusterName, _darkClusterDispatcher, _notifier, _verifierManager);
return new IdenticalTrafficMultiplierDarkClusterStrategy(_sourceClusterName, darkClusterName,
darkClusterConfig.getMultiplier(), baseDarkClusterDispatcher,
_notifier, _facilities.getClusterInfoProvider(), _random);
_notifier, partitionAwareProvider,
_random);
}
break;
case CONSTANT_QPS:
Expand All @@ -191,7 +225,7 @@ private DarkClusterStrategy createStrategy(String darkClusterName, DarkClusterCo
rateLimiter.setBufferTtl(darkClusterConfig.getDispatcherBufferedRequestExpiryInSeconds(), ChronoUnit.SECONDS);
return new ConstantQpsDarkClusterStrategy(_sourceClusterName, darkClusterName,
darkClusterConfig.getDispatcherOutboundTargetRate(), baseDarkClusterDispatcher,
_notifier, _facilities.getClusterInfoProvider(), rateLimiter);
_notifier, partitionAwareProvider, rateLimiter);
}
break;
default:
Expand All @@ -215,6 +249,7 @@ public void onClusterAdded(String updatedClusterName)
// pertinent dark cluster strategy properties are contained there.
if (_sourceClusterName.equals(updatedClusterName))
{
_sourceClusterPresent = true;
_facilities.getClusterInfoProvider().getDarkClusterConfigMap(_sourceClusterName, new Callback<DarkClusterConfigMap>()
{
@Override
Expand All @@ -227,23 +262,23 @@ public void onError(Throwable e)
@Override
public void onSuccess(DarkClusterConfigMap updatedDarkConfigMap)
{
Set<String> oldDarkStrategySet = _darkStrategyMap.keySet();
Set<String> updatedDarkClusterConfigKeySet = updatedDarkConfigMap.keySet();
// Any old strategy entry that isn't in the "updated" set should be removed from the strategyMap.
oldDarkStrategySet.removeAll(updatedDarkClusterConfigKeySet);
for (String darkClusterToRemove : oldDarkStrategySet)
// Determine partitions to (re)build. If none exist yet, ensure default partition is initialized.
java.util.Set<Integer> partitions = new java.util.HashSet<>(_partitionToDarkStrategyMap.keySet());
if (partitions.isEmpty())
{
_darkStrategyMap.remove(darkClusterToRemove);
LOG.info("Removed dark cluster strategy for dark cluster: " + darkClusterToRemove + ", source cluster: " + _sourceClusterName);
partitions.add(com.linkedin.d2.balancer.util.partitions.DefaultPartitionAccessor.DEFAULT_PARTITION_ID);
}

// Now update/add the dark clusters.
for (Map.Entry<String, DarkClusterConfig> entry : updatedDarkConfigMap.entrySet())
for (int partitionId : partitions)
{
String darkClusterToAdd = entry.getKey();
// For simplicity, we refresh all strategies since we expect cluster updates to be rare and refresh to be cheap.
_darkStrategyMap.put(darkClusterToAdd, createStrategy(darkClusterToAdd, entry.getValue()));
LOG.info("Created new strategy for dark cluster: " + darkClusterToAdd + ", source cluster: " + _sourceClusterName);
Map<String, DarkClusterStrategy> darkStrategyMap = new ConcurrentHashMap<>();
for (Map.Entry<String, DarkClusterConfig> entry : updatedDarkConfigMap.entrySet())
{
String darkClusterToAdd = entry.getKey();
darkStrategyMap.put(darkClusterToAdd, createStrategy(darkClusterToAdd, entry.getValue(), partitionId));
LOG.info("Created new strategy for dark cluster: " + darkClusterToAdd + ", partition: " + partitionId + ", source cluster: " + _sourceClusterName);
}
_partitionToDarkStrategyMap.put(partitionId, darkStrategyMap);
}
}
});
Expand All @@ -258,8 +293,58 @@ public void onClusterRemoved(String clusterName)
{
if (_sourceClusterName.equals(clusterName))
{
_darkStrategyMap.clear();
_partitionToDarkStrategyMap.clear();
_sourceClusterPresent = false;
}
}
}

/**
* Partition-aware wrapper around a {@link ClusterInfoProvider} that filters cluster information
* for a specific partition before forwarding to the strategies.
*/
private static final class PartitionAwareClusterInfoProvider implements com.linkedin.d2.balancer.util.ClusterInfoProvider {
private final com.linkedin.d2.balancer.util.ClusterInfoProvider _delegate;
private final int _partitionId;

PartitionAwareClusterInfoProvider(com.linkedin.d2.balancer.util.ClusterInfoProvider delegate, int partitionId) {
_delegate = delegate;
_partitionId = partitionId;
}

@Override
public int getHttpsClusterCount(String clusterName) throws ServiceUnavailableException {
return getClusterCount(clusterName, com.linkedin.d2.balancer.properties.PropertyKeys.HTTPS_SCHEME, _partitionId);
}

@Override
public int getClusterCount(String clusterName, String scheme, int partitionId) throws ServiceUnavailableException {
return _delegate.getClusterCount(clusterName, scheme, partitionId);
}

@Override
public DarkClusterConfigMap getDarkClusterConfigMap(String clusterName) throws ServiceUnavailableException {
return _delegate.getDarkClusterConfigMap(clusterName);
}

@Override
public void getDarkClusterConfigMap(String clusterName, Callback<DarkClusterConfigMap> callback) {
_delegate.getDarkClusterConfigMap(clusterName, callback);
}

@Override
public void registerClusterListener(LoadBalancerClusterListener clusterListener) {
_delegate.registerClusterListener(clusterListener);
}

@Override
public void unregisterClusterListener(LoadBalancerClusterListener clusterListener) {
_delegate.unregisterClusterListener(clusterListener);
}

@Override
public FailoutConfig getFailoutConfig(String clusterName) {
return _delegate.getFailoutConfig(clusterName);
}
}
}
Original file line number Diff line number Diff line change
Expand Up @@ -107,8 +107,7 @@ private int getNumDuplicateRequests(RequestContext requestContext)
{
try
{
// Only support https for now. http support can be added later if truly needed, but would be non-ideal
// because potentially both dark and source would have to be configured.
// ClusterInfoProvider is already partition-aware, so we use the default partition ID
int numDarkClusterInstances = _clusterInfoProvider.getHttpsClusterCount(_darkClusterName);
int numSourceClusterInstances = _clusterInfoProvider.getHttpsClusterCount(_originalClusterName);
float randomNumber;
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -110,8 +110,6 @@ private int getNumDuplicateRequests()
{
try
{
// Only support https for now. http support can be added later if truly needed, but would be non-ideal
// because potentially both dark and source would have to be configured.
int numDarkClusterInstances = _clusterInfoProvider.getHttpsClusterCount(_darkClusterName);
int numSourceClusterInstances = _clusterInfoProvider.getHttpsClusterCount(_originalClusterName);
if (numSourceClusterInstances != 0)
Expand Down
Loading
Loading