public async Task PlainPartitionedSource_Should_split_messages_by_partitions()
{
var topic = CreateTopic(1);
var group = CreateGroup(1);
var totalMessages = 100;
await ProduceStrings(topic, Enumerable.Range(1, totalMessages), ProducerSettings);
var consumerSettings = CreateConsumerSettings <string>(group);
var control = KafkaConsumer.PlainPartitionedSource(consumerSettings, Subscriptions.Topics(topic))
.SelectAsync(6, async tuple =>
{
var(topicPartition, source) = tuple;
Log.Info($"Sub-source for {topicPartition}");
var consumedPartitions = await source
.Select(m => m.TopicPartition.Partition)
.RunWith(Sink.Seq <Partition>(), Materializer);
// Return flag that all messages in child source are from the same, expected partition
return(consumedPartitions.All(partition => partition == topicPartition.Partition));
})
.As <Source <bool, IControl> >()
.ToMaterialized(Sink.Aggregate <bool, bool>(true, (result, childSourceIsValid) => result && childSourceIsValid), Keep.Both)
.MapMaterializedValue(tuple => DrainingControl <bool> .Create(tuple.Item1, tuple.Item2))
.Run(Materializer);
// Give it some time to consume all messages
await Task.Delay(5000);
var shutdown = control.DrainAndShutdown();
AwaitCondition(() => shutdown.IsCompleted, TimeSpan.FromSeconds(10));
shutdown.Result.Should().BeTrue();
}
public static async Task Main(string[] args)
{
Config fallbackConfig = ConfigurationFactory.ParseString(@"
akka.suppress-json-serializer-warning=true
akka.loglevel = DEBUG
").WithFallback(ConfigurationFactory.FromResource <ConsumerSettings <object, object> >("Akka.Streams.Kafka.reference.conf"));
var system = ActorSystem.Create("TestKafka", fallbackConfig);
var materializer = system.Materializer();
var consumerSettings = ConsumerSettings <Null, string> .Create(system, null, null)
.WithBootstrapServers($"{EventHubNamespace}.servicebus.windows.net:9093")
.WithGroupId(EventHubConsumerGroup)
.WithProperties(new Dictionary <string, string>
{
{ "security.protocol", "SASL_SSL" },
{ "sasl.mechanism", "PLAIN" },
{ "sasl.username", "$ConnectionString" },
{ "sasl.password", EventHubConnectionString },
});
var subscription = Subscriptions.Topics(EventHubName);
var committerDefaults = CommitterSettings.Create(system);
// Comment for simple no-commit consumer
DrainingControl <NotUsed> control = KafkaConsumer.CommittableSource(consumerSettings, subscription)
.SelectAsync(1, msg =>
Business(msg.Record).ContinueWith(done => (ICommittable)msg.CommitableOffset))
.ToMaterialized(
Committer.Sink(committerDefaults.WithMaxBatch(1)),
DrainingControl <NotUsed> .Create)
.Run(materializer);
// Uncomment for simple no-commit consumer
/*
* await KafkaConsumer.PlainSource(consumerSettings, subscription)
* .RunForeach(result =>
* {
* Console.WriteLine($"Consumer: {result.Topic}/{result.Partition} {result.Offset}: {result.Value}");
* }, materializer);
*/
Console.WriteLine("Press any key to stop consumer.");
Console.ReadKey();
// Comment for simple no-commit consumer
await control.Stop();
await system.Terminate();
}
public async Task PlainPartitionedSource_should_work()
{
var topic = CreateTopic(1);
var group = CreateGroup(1);
var totalMessages = 100;
var receivedMessages = new AtomicCounter(0);
var consumerSettings = CreateConsumerSettings <string>(group);
var control = KafkaConsumer.PlainPartitionedSource(consumerSettings, Subscriptions.Topics(topic))
.GroupBy(3, tuple => tuple.Item1)
.SelectAsync(8, async tuple =>
{
var(topicPartition, source) = tuple;
Log.Info($"Sub-source for {topicPartition}");
var sourceMessages = await source
.Scan(0, (i, message) => i + 1)
.Select(i =>
{
receivedMessages.IncrementAndGet();
return(LogReceivedMessages(topicPartition, i));
})
.RunWith(Sink.Last <long>(), Materializer);
Log.Info($"{topicPartition}: Received {sourceMessages} messages in total");
return(sourceMessages);
})
.MergeSubstreams()
.As <Source <long, IControl> >()
.Scan(0L, (i, subValue) => i + subValue)
.ToMaterialized(Sink.Last <long>(), Keep.Both)
.MapMaterializedValue(tuple => DrainingControl <long> .Create(tuple.Item1, tuple.Item2))
.Run(Materializer);
await ProduceStrings(topic, Enumerable.Range(1, totalMessages), ProducerSettings);
for (var i = 0; i < totalMessages; ++i)
{
await AwaitConditionAsync(() => receivedMessages.Current > i, TimeSpan.FromSeconds(10));
}
await Task.Delay(1000); // Wait for message handling finished after all messages received
var shutdown = control.DrainAndShutdown();
AwaitCondition(() => shutdown.IsCompleted, TimeSpan.FromSeconds(10));
shutdown.Result.Should().Be(totalMessages);
}
public async Task ProducerFlowWithContext_should_work_with_source_with_context()
{
bool Duplicate(string value) => value == "1";
bool Ignore(string value) => value == "2";
var consumerSettings = CreateConsumerSettings <string, string>(CreateGroup(1));
var topic1 = CreateTopic(1);
var topic2 = CreateTopic(2);
var topic3 = CreateTopic(3);
var topic4 = CreateTopic(4);
var producerSettings = BuildProducerSettings <string, string>();
var committerSettings = CommitterSettings;
var totalMessages = 10;
var totalConsumed = 0;
await ProduceStrings(topic1, Enumerable.Range(1, totalMessages), producerSettings);
var(control2, result) = KafkaConsumer.PlainSource(consumerSettings, Subscriptions.Topics(topic2, topic3, topic4))
.Scan(0, (c, _) => c + 1)
.Select(consumed =>
{
totalConsumed = consumed;
return(consumed);
})
.ToMaterialized(Sink.Last <int>(), Keep.Both)
.Run(Materializer);
var control = KafkaConsumer.SourceWithOffsetContext(consumerSettings, Subscriptions.Topics(topic1))
.Select(record =>
{
IEnvelope <string, string, NotUsed> output;
if (Duplicate(record.Message.Value))
{
output = ProducerMessage.Multi(new[]
{
new ProducerRecord <string, string>(topic2, record.Message.Key, record.Message.Value),
new ProducerRecord <string, string>(topic3, record.Message.Key, record.Message.Value)
}.ToImmutableSet());
}
else if (Ignore(record.Message.Value))
{
output = ProducerMessage.PassThrough <string, string>();
}
else
{
output = ProducerMessage.Single(new ProducerRecord <string, string>(topic4, record.Message.Key, record.Message.Value));
}
Log.Debug($"Giving message of type {output.GetType().Name}");
return(output);
})
.Via(KafkaProducer.FlowWithContext <string, string, ICommittableOffset>(producerSettings))
.AsSource()
.Log("Produced messages", r => $"Committing {r.Item2.Offset.Topic}:{r.Item2.Offset.Partition}[{r.Item2.Offset.Offset}]")
.ToMaterialized(Committer.SinkWithOffsetContext <IResults <string, string, ICommittableOffset> >(committerSettings), Keep.Both)
.MapMaterializedValue(tuple => DrainingControl <NotUsed> .Create(tuple.Item1, tuple.Item2))
.Run(Materializer);
// One by one, wait while all `totalMessages` will be consumed
for (var i = 1; i < totalMessages; ++i)
{
var consumedExpect = i;
Log.Info($"Waiting for {consumedExpect} to be consumed...");
try
{
await AwaitConditionAsync(() => totalConsumed >= consumedExpect, TimeSpan.FromSeconds(30));
}
finally
{
Log.Info($"Finished waiting for {consumedExpect} messages. Total: {totalConsumed}");
}
Log.Info($"Confirmed that {consumedExpect} messages are consumed");
}
AssertTaskCompletesWithin(TimeSpan.FromSeconds(10), control.DrainAndShutdown());
AssertTaskCompletesWithin(TimeSpan.FromSeconds(10), control2.Shutdown());
AssertTaskCompletesWithin(TimeSpan.FromSeconds(10), result).Should().Be(totalConsumed);
}
public static async Task <int> Main(string[] args)
{
// Setup
await SetupKafkaAsync();
await SetupAkkaAsync();
List <CpuUsage> usageBeforeLoad;
List <CpuUsage> usageAfterLoad;
try
{
_log = Logging.GetLogger(ConsumerSystem, nameof(Program));
// Create topic on Kafka server
var builder = new AdminClientBuilder(new AdminClientConfig
{
BootstrapServers = Benchmark.Docker.KafkaAddress
});
using (var client = builder.Build())
{
await client.CreateTopicsAsync(new[] { new TopicSpecification
{
Name = KafkaTopic,
NumPartitions = 3,
ReplicationFactor = 1
} });
}
// Set up consumer
var consumerSettings = ConsumerSettings <string, string> .Create(ConsumerSystem, null, null)
.WithBootstrapServers(Benchmark.Docker.KafkaAddress)
.WithStopTimeout(TimeSpan.FromSeconds(1))
.WithProperty("auto.offset.reset", "earliest")
.WithGroupId(KafkaGroup);
var control = KafkaConsumer.PlainPartitionedSource(consumerSettings, Subscriptions.Topics(KafkaTopic))
.GroupBy(3, tuple => tuple.Item1)
.SelectAsync(8, async tuple =>
{
var(topicPartition, source) = tuple;
_log.Info($"Sub-source for {topicPartition}");
var sourceMessages = await source
.Scan(0, (i, message) => i + 1)
.Select(i =>
{
ReceivedMessage.IncrementAndGet();
return(LogReceivedMessages(topicPartition, i));
})
.RunWith(Sink.Last <long>(), ConsumerSystem.Materializer());
_log.Info($"{topicPartition}: Received {sourceMessages} messages in total");
return(sourceMessages);
})
.MergeSubstreams()
.AsInstanceOf <Source <long, IControl> >()
.Scan(0L, (i, subValue) => i + subValue)
.ToMaterialized(Sink.Last <long>(), Keep.Both)
.MapMaterializedValue(tuple => DrainingControl <long> .Create(tuple.Item1, tuple.Item2))
.Run(ConsumerSystem.Materializer());
// Delay before benchmark
await Task.Delay(TimeSpan.FromSeconds(DefaultDelay));
// Warmup
await CollectSamplesAsync(DefaultWarmUpRepeat, DefaultSampleDuration, "[Warmup]");
// Collect CPU usage before load
usageBeforeLoad = await CollectSamplesAsync(DefaultRepeat, DefaultSampleDuration, "[CPU Usage Before Load]");
// Create load
var producerSettings = ProducerSettings <string, string> .Create(ConsumerSystem, null, null)
.WithBootstrapServers(Benchmark.Docker.KafkaAddress);
await Source
.From(Enumerable.Range(1, DefaultMessageCount))
.Select(elem => new ProducerRecord <string, string>(KafkaTopic, "key", elem.ToString()))
.RunWith(KafkaProducer.PlainSink(producerSettings), ConsumerSystem.Materializer());
// Wait until consumer consumed all messages
var stopwatch = Stopwatch.StartNew();
while (stopwatch.Elapsed.TotalSeconds < DefaultTimeout && ReceivedMessage.Current < DefaultMessageCount)
{
await Task.Delay(100);
}
stopwatch.Stop();
if (stopwatch.Elapsed.TotalSeconds > DefaultTimeout)
{
throw new Exception($"Timed out while waiting consumer to process {DefaultMessageCount} messages");
}
// Delay before benchmark
await Task.Delay(TimeSpan.FromSeconds(DefaultDelay));
// Collect CPU usage after load
usageAfterLoad = await CollectSamplesAsync(DefaultRepeat, DefaultSampleDuration, "[CPU Usage After Load]");
}
finally
{
// Tear down
await TearDownAkkaAsync();
await TearDownKafkaAsync();
}
Console.WriteLine("CPU Benchmark complete.");
await GenerateReportAsync(usageBeforeLoad, "BeforeLoad", DefaultSampleDuration, DefaultRepeat);
await GenerateReportAsync(usageAfterLoad, "AfterLoad", DefaultSampleDuration, DefaultRepeat);
return(0);
}
