org.apache.cassandra.hints.HintsDispatchExecutor 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.hints;
import java.io.File;
import java.util.Map;
import java.util.UUID;
import java.util.concurrent.*;
import java.util.concurrent.atomic.AtomicBoolean;
import java.util.function.BooleanSupplier;
import java.util.function.Predicate;
import java.util.function.Supplier;
import com.google.common.util.concurrent.RateLimiter;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.apache.cassandra.concurrent.JMXEnabledThreadPoolExecutor;
import org.apache.cassandra.concurrent.NamedThreadFactory;
import org.apache.cassandra.config.DatabaseDescriptor;
import org.apache.cassandra.io.FSReadError;
import org.apache.cassandra.locator.InetAddressAndPort;
import org.apache.cassandra.service.StorageService;
/**
* A multi-threaded (by default) executor for dispatching hints.
*
* Most of dispatch is triggered by {@link HintsDispatchTrigger} running every ~10 seconds.
*/
final class HintsDispatchExecutor
{
private static final Logger logger = LoggerFactory.getLogger(HintsDispatchExecutor.class);
private final File hintsDirectory;
private final ExecutorService executor;
private final AtomicBoolean isPaused;
private final Predicate isAlive;
private final Map scheduledDispatches;
HintsDispatchExecutor(File hintsDirectory, int maxThreads, AtomicBoolean isPaused, Predicate isAlive)
{
this.hintsDirectory = hintsDirectory;
this.isPaused = isPaused;
this.isAlive = isAlive;
scheduledDispatches = new ConcurrentHashMap<>();
executor = new JMXEnabledThreadPoolExecutor(maxThreads, 1, TimeUnit.MINUTES,
new LinkedBlockingQueue<>(),
new NamedThreadFactory("HintsDispatcher", Thread.MIN_PRIORITY),
"internal");
}
/*
* It's safe to terminate dispatch in process and to deschedule dispatch.
*/
void shutdownBlocking()
{
scheduledDispatches.clear();
executor.shutdownNow();
try
{
executor.awaitTermination(1, TimeUnit.MINUTES);
}
catch (InterruptedException e)
{
throw new AssertionError(e);
}
}
boolean isScheduled(HintsStore store)
{
return scheduledDispatches.containsKey(store.hostId);
}
Future dispatch(HintsStore store)
{
return dispatch(store, store.hostId);
}
Future dispatch(HintsStore store, UUID hostId)
{
/*
* It is safe to perform dispatch for the same host id concurrently in two or more threads,
* however there is nothing to win from it - so we don't.
*
* Additionally, having just one dispatch task per host id ensures that we'll never violate our per-destination
* rate limit, without having to share a ratelimiter between threads.
*
* It also simplifies reasoning about dispatch sessions.
*/
return scheduledDispatches.computeIfAbsent(hostId, uuid -> executor.submit(new DispatchHintsTask(store, hostId)));
}
Future transfer(HintsCatalog catalog, Supplier hostIdSupplier)
{
return executor.submit(new TransferHintsTask(catalog, hostIdSupplier));
}
void completeDispatchBlockingly(HintsStore store)
{
Future future = scheduledDispatches.get(store.hostId);
try
{
if (future != null)
future.get();
}
catch (ExecutionException | InterruptedException e)
{
throw new RuntimeException(e);
}
}
void interruptDispatch(UUID hostId)
{
Future future = scheduledDispatches.remove(hostId);
if (null != future)
future.cancel(true);
}
private final class TransferHintsTask implements Runnable
{
private final HintsCatalog catalog;
/*
* Supplies target hosts to stream to. Generally returns the one the DynamicSnitch thinks is closest.
* We use a supplier here to be able to get a new host if the current one dies during streaming.
*/
private final Supplier hostIdSupplier;
private TransferHintsTask(HintsCatalog catalog, Supplier hostIdSupplier)
{
this.catalog = catalog;
this.hostIdSupplier = hostIdSupplier;
}
@Override
public void run()
{
UUID hostId = hostIdSupplier.get();
InetAddressAndPort address = StorageService.instance.getEndpointForHostId(hostId);
logger.info("Transferring all hints to {}: {}", address, hostId);
if (transfer(hostId))
return;
logger.warn("Failed to transfer all hints to {}: {}; will retry in {} seconds", address, hostId, 10);
try
{
TimeUnit.SECONDS.sleep(10);
}
catch (InterruptedException e)
{
throw new RuntimeException(e);
}
hostId = hostIdSupplier.get();
logger.info("Transferring all hints to {}: {}", address, hostId);
if (!transfer(hostId))
{
logger.error("Failed to transfer all hints to {}: {}", address, hostId);
throw new RuntimeException("Failed to transfer all hints to " + hostId);
}
}
private boolean transfer(UUID hostId)
{
catalog.stores()
.map(store -> new DispatchHintsTask(store, hostId, true))
.forEach(Runnable::run);
return !catalog.hasFiles();
}
}
private final class DispatchHintsTask implements Runnable
{
private final HintsStore store;
private final UUID hostId;
private final RateLimiter rateLimiter;
DispatchHintsTask(HintsStore store, UUID hostId, boolean isTransfer)
{
this.store = store;
this.hostId = hostId;
// Rate limit is in bytes per second. Uses Double.MAX_VALUE if disabled (set to 0 in cassandra.yaml).
// Max rate is scaled by the number of nodes in the cluster (CASSANDRA-5272), unless we are transferring
// hints during decomission rather than dispatching them to their final destination.
// The goal is to bound maximum hints traffic going towards a particular node from the rest of the cluster,
// not total outgoing hints traffic from this node. This is why the rate limiter is not shared between
// all the dispatch tasks (as there will be at most one dispatch task for a particular host id at a time).
int nodesCount = isTransfer ? 1 : Math.max(1, StorageService.instance.getTokenMetadata().getAllEndpoints().size() - 1);
double throttleInBytes = DatabaseDescriptor.getHintedHandoffThrottleInKB() * 1024.0 / nodesCount;
this.rateLimiter = RateLimiter.create(throttleInBytes == 0 ? Double.MAX_VALUE : throttleInBytes);
}
DispatchHintsTask(HintsStore store, UUID hostId)
{
this(store, hostId, false);
}
public void run()
{
try
{
dispatch();
}
finally
{
scheduledDispatches.remove(hostId);
}
}
private void dispatch()
{
while (true)
{
if (isPaused.get())
break;
HintsDescriptor descriptor = store.poll();
if (descriptor == null)
break;
try
{
if (!dispatch(descriptor))
break;
}
catch (FSReadError e)
{
logger.error(String.format("Failed to dispatch hints file %s: file is corrupted", descriptor.fileName()), e);
store.cleanUp(descriptor);
store.markCorrupted(descriptor);
throw e;
}
}
}
/*
* Will return true if dispatch was successful, false if we hit a failure (destination node went down, for example).
*/
private boolean dispatch(HintsDescriptor descriptor)
{
logger.trace("Dispatching hints file {}", descriptor.fileName());
InetAddressAndPort address = StorageService.instance.getEndpointForHostId(hostId);
if (address != null)
return deliver(descriptor, address);
// address == null means the target no longer exist; find new home for each hint entry.
convert(descriptor);
return true;
}
private boolean deliver(HintsDescriptor descriptor, InetAddressAndPort address)
{
File file = new File(hintsDirectory, descriptor.fileName());
InputPosition offset = store.getDispatchOffset(descriptor);
BooleanSupplier shouldAbort = () -> !isAlive.test(address) || isPaused.get();
try (HintsDispatcher dispatcher = HintsDispatcher.create(file, rateLimiter, address, descriptor.hostId, shouldAbort))
{
if (offset != null)
dispatcher.seek(offset);
if (dispatcher.dispatch())
{
store.delete(descriptor);
store.cleanUp(descriptor);
logger.info("Finished hinted handoff of file {} to endpoint {}: {}", descriptor.fileName(), address, hostId);
return true;
}
else
{
store.markDispatchOffset(descriptor, dispatcher.dispatchPosition());
store.offerFirst(descriptor);
logger.info("Finished hinted handoff of file {} to endpoint {}: {}, partially", descriptor.fileName(), address, hostId);
return false;
}
}
}
// for each hint in the hints file for a node that isn't part of the ring anymore, write RF hints for each replica
private void convert(HintsDescriptor descriptor)
{
File file = new File(hintsDirectory, descriptor.fileName());
try (HintsReader reader = HintsReader.open(file, rateLimiter))
{
reader.forEach(page -> page.hintsIterator().forEachRemaining(HintsService.instance::writeForAllReplicas));
store.delete(descriptor);
store.cleanUp(descriptor);
logger.info("Finished converting hints file {}", descriptor.fileName());
}
}
}
public boolean isPaused()
{
return isPaused.get();
}
public boolean hasScheduledDispatches()
{
return !scheduledDispatches.isEmpty();
}
}
© 2015 - 2024 Weber Informatics LLC | Privacy Policy