private TopologyTestDriver(InternalTopologyBuilder builder, IStreamConfig config)
{
this.topologyBuilder = builder;
this.configuration = config;
// ONLY 1 thread for test driver
this.configuration.NumStreamThreads = 1;
this.configuration.Guarantee = ProcessingGuarantee.AT_LEAST_ONCE;
this.topicConfiguration = config.Clone();
this.topicConfiguration.ApplicationId = $"test-driver-{this.configuration.ApplicationId}";
var processID = Guid.NewGuid();
var clientId = string.IsNullOrEmpty(configuration.ClientId) ? $"{this.configuration.ApplicationId.ToLower()}-{processID}" : configuration.ClientId;
this.configuration.ClientId = clientId;
kafkaSupplier = new MockKafkaSupplier();
pipeBuilder = new PipeBuilder(kafkaSupplier);
this.processorTopology = this.topologyBuilder.BuildTopology();
this.threadTopology = StreamThread.Create(
$"{this.configuration.ApplicationId.ToLower()}-stream-thread-0",
clientId,
builder,
config,
kafkaSupplier,
kafkaSupplier.GetAdmin(configuration.ToAdminConfig($"{clientId}-admin")),
0);
RunDriver();
}
public StreamTask(string threadId, TaskId id, TopicPartition partition, ProcessorTopology processorTopology, IConsumer <byte[], byte[]> consumer, IStreamConfig configuration, IKafkaSupplier kafkaSupplier, IProducer <byte[], byte[]> producer)
: base(id, partition, processorTopology, consumer, configuration)
{
this.threadId = threadId;
this.kafkaSupplier = kafkaSupplier;
this.consumedOffsets = new Dictionary <TopicPartition, long>();
// eos enabled
if (producer == null)
{
this.producer = CreateEOSProducer();
InitializeTransaction();
eosEnabled = true;
}
else
{
this.producer = producer;
}
this.collector = new RecordCollector(logPrefix);
collector.Init(ref this.producer);
var sourceTimestampExtractor = (processorTopology.GetSourceProcessor(id.Topic) as ISourceProcessor).Extractor;
Context = new ProcessorContext(configuration, stateMgr).UseRecordCollector(collector);
processor = processorTopology.GetSourceProcessor(partition.Topic);
queue = new RecordQueue <ConsumeResult <byte[], byte[]> >(
100,
logPrefix,
$"record-queue-{id.Topic}-{id.Partition}",
sourceTimestampExtractor == null ? configuration.DefaultTimestampExtractor : sourceTimestampExtractor);
}
/// <summary>
/// Create a <see cref="KafkaStream"/> instance with your own <see cref="IKafkaSupplier" />
/// Please DO NOT FORGET to call Close to avoid resources leak !
/// </summary>
/// <param name="topology">the topology specifying the computational logic</param>
/// <param name="configuration">configuration about this stream</param>
/// <param name="kafkaSupplier">the Kafka clients supplier which provides underlying producer and consumer clients for the new <see cref="KafkaStream"/> instance</param>
public KafkaStream(Topology topology, IStreamConfig configuration, IKafkaSupplier kafkaSupplier)
{
this.topology = topology;
this.configuration = configuration;
this.kafkaSupplier = kafkaSupplier;
var processID = Guid.NewGuid();
clientId = string.IsNullOrEmpty(configuration.ClientId) ? $"{this.configuration.ApplicationId.ToLower()}-{processID}" : configuration.ClientId;
logPrefix = $"stream-application[{configuration.ApplicationId}] ";
// re-write the physical topology according to the config
topology.Builder.RewriteTopology(configuration);
// sanity check
this.processorTopology = topology.Builder.BuildTopology();
this.threads = new IThread[this.configuration.NumStreamThreads];
var threadState = new Dictionary <long, Processors.ThreadState>();
List <StreamThreadStateStoreProvider> stateStoreProviders = new List <StreamThreadStateStoreProvider>();
for (int i = 0; i < this.configuration.NumStreamThreads; ++i)
{
var threadId = $"{this.configuration.ApplicationId.ToLower()}-stream-thread-{i}";
adminClient = this.kafkaSupplier.GetAdmin(configuration.ToAdminConfig(StreamThread.GetSharedAdminClientId(clientId)));
this.threads[i] = StreamThread.Create(
threadId,
clientId,
this.topology.Builder,
configuration,
this.kafkaSupplier,
adminClient,
i);
threadState.Add(this.threads[i].Id, this.threads[i].State);
stateStoreProviders.Add(new StreamThreadStateStoreProvider(this.threads[i], this.topology.Builder));
}
var manager = new StreamStateManager(this, threadState);
foreach (var t in threads)
{
t.StateChanged += manager.OnChange;
}
this.queryableStoreProvider = new QueryableStoreProvider(stateStoreProviders);
StreamState = State.CREATED;
}
public GlobalStreamThreadFactory(ProcessorTopology topology,
string threadClientId,
IConsumer <byte[], byte[]> globalConsumer,
IStreamConfig configuration,
IAdminClient adminClient)
{
this.adminClient = adminClient;
this.topology = topology;
this.threadClientId = threadClientId;
this.configuration = configuration;
this.globalConsumer = globalConsumer;
}
public void SetUp()
{
mockKafkaSupplier = new SyncKafkaSupplier();
var consumerConfig = new ConsumerConfig();
consumerConfig.GroupId = "global-consulmer";
var globalConsumer = mockKafkaSupplier.GetConsumer(consumerConfig, null);
streamConfigMock = new Mock <IStreamConfig>();
streamConfigMock.Setup(c => c.StateDir).Returns($"./{Guid.NewGuid().ToString()}");
streamConfigMock.Setup(c => c.ApplicationId).Returns("app");
kvStoreMock = CreateMockStore <IKeyValueStore <object, object> >(kvStoreName);
otherStoreMock = CreateMockStore <IKeyValueStore <object, object> >(otherStoreName);
var globalStateStores = new Dictionary <string, IStateStore>()
{
{ kvStoreMock.Object.Name, kvStoreMock.Object },
{ otherStoreMock.Object.Name, otherStoreMock.Object }
};
var storesToTopics = new Dictionary <string, string>()
{
{ kvStoreMock.Object.Name, kvStoreTopic },
{ otherStoreMock.Object.Name, otherStoreTopic }
};
topology = new ProcessorTopology(
null,
new Dictionary <string, IProcessor>(),
new Dictionary <string, IProcessor>(),
new Dictionary <string, IProcessor>(),
new Dictionary <string, IStateStore>(),
globalStateStores,
storesToTopics,
new List <string>());
adminClientMock = new Mock <IAdminClient>();
RegisterPartitionInAdminClient(kvStoreTopic);
RegisterPartitionInAdminClient(otherStoreTopic);
stateManager = new GlobalStateManager(globalConsumer, topology,
adminClientMock.Object,
streamConfigMock.Object
);
context = new GlobalProcessorContext(
streamConfigMock.Object,
stateManager,
new StreamMetricsRegistry());
stateManager.SetGlobalProcessorContext(context);
}
protected AbstractTask(TaskId id, TopicPartition partition, ProcessorTopology topology, IConsumer <byte[], byte[]> consumer, IStreamConfig config)
{
log = Logger.GetLogger(GetType());
logPrefix = $"stream-task[{id.Topic}|{id.Partition}] ";
Partition = partition;
Id = id;
Topology = topology;
this.consumer = consumer;
configuration = config;
stateMgr = new ProcessorStateManager(id, partition);
}
public TaskSynchronousTopologyDriver(string clientId, InternalTopologyBuilder topologyBuilder,
IStreamConfig configuration, IStreamConfig topicConfiguration, IKafkaSupplier supplier,
CancellationToken token)
{
this.configuration = configuration;
this.configuration.ClientId = clientId;
this.topicConfiguration = topicConfiguration;
metricsRegistry = new StreamMetricsRegistry(clientId, MetricsRecordingLevel.DEBUG);
this.token = token;
builder = topologyBuilder;
this.supplier = supplier ?? new SyncKafkaSupplier();
this.supplier.MetricsRegistry = metricsRegistry;
producer = this.supplier.GetProducer(configuration.ToProducerConfig()) as SyncProducer;
foreach (var sourceTopic in builder
.GetSourceTopics())
{
var part = new TopicPartition(sourceTopic, 0);
var taskId = builder.GetTaskIdFromPartition(part);
if (partitionsByTaskId.ContainsKey(taskId))
{
partitionsByTaskId[taskId].Add(part);
}
else
{
partitionsByTaskId.Add(taskId, new List <TopicPartition> {
part
});
}
}
ProcessorTopology globalTaskTopology = topologyBuilder.BuildGlobalStateTopology();
hasGlobalTopology = globalTaskTopology != null;
if (hasGlobalTopology)
{
var globalConsumer =
this.supplier.GetGlobalConsumer(configuration.ToGlobalConsumerConfig($"{clientId}-global-consumer"));
var adminClient = this.supplier.GetAdmin(configuration.ToAdminConfig($"{clientId}-admin"));
var stateManager =
new GlobalStateManager(globalConsumer, globalTaskTopology, adminClient, configuration);
globalProcessorContext = new GlobalProcessorContext(configuration, stateManager, metricsRegistry);
stateManager.SetGlobalProcessorContext(globalProcessorContext);
globalTask = new GlobalStateUpdateTask(stateManager, globalTaskTopology, globalProcessorContext);
globalTask.Initialize();
}
}
protected AbstractTask(TaskId id, IEnumerable <TopicPartition> partition, ProcessorTopology topology, IConsumer <byte[], byte[]> consumer, IStreamConfig config)
{
log = Logger.GetLogger(GetType());
logPrefix = $"stream-task[{id.Id}|{id.Partition}] ";
var topicPartitions = partition.ToList();
Partition = topicPartitions;
Id = id;
Topology = topology;
this.consumer = consumer;
configuration = config;
stateMgr = new ProcessorStateManager(id, topicPartitions);
}
/// <summary>
/// Create a <see cref="KafkaStream"/> instance.
/// Please DO NOT FORGET to call Close to avoid resources leak !
/// </summary>
/// <param name="topology">the topology specifying the computational logic</param>
/// <param name="configuration">configuration about this stream</param>
public KafkaStream(Topology topology, IStreamConfig configuration)
{
this.topology = topology;
this.configuration = configuration;
this.kafkaSupplier = new DefaultKafkaClientSupplier(new KafkaLoggerAdapter(configuration));
var processID = Guid.NewGuid();
clientId = string.IsNullOrEmpty(configuration.ClientId) ? $"{this.configuration.ApplicationId.ToLower()}-{processID}" : configuration.ClientId;
logPrefix = $"stream-application[{configuration.ApplicationId}] ";
// sanity check
this.processorTopology = topology.Builder.BuildTopology();
this.threads = new IThread[this.configuration.NumStreamThreads];
var threadState = new Dictionary <long, Processors.ThreadState>();
for (int i = 0; i < this.configuration.NumStreamThreads; ++i)
{
var threadId = $"{this.configuration.ApplicationId.ToLower()}-stream-thread-{i}";
adminClient = this.kafkaSupplier.GetAdmin(configuration.ToAdminConfig(StreamThread.GetSharedAdminClientId(clientId)));
this.threads[i] = StreamThread.Create(
threadId,
clientId,
this.topology.Builder,
configuration,
this.kafkaSupplier,
adminClient,
i);
threadState.Add(this.threads[i].Id, this.threads[i].State);
}
var manager = new StreamStateManager(this, threadState);
foreach (var t in threads)
{
t.StateChanged += manager.OnChange;
}
StreamState = State.CREATED;
}
public void Setup()
{
contextMock = new Mock <GlobalProcessorContext>(null, null, new StreamMetricsRegistry());
stateManagerMock = new Mock <IGlobalStateManager>();
processorMock = new Mock <IProcessor>();
sourceProcessorMock = new Mock <ISourceProcessor>();
otherSourceProcessorMock = new Mock <ISourceProcessor>();
sourceProcessorMock.Setup(x => x.TopicName).Returns("topic1");
otherSourceProcessorMock.Setup(x => x.TopicName).Returns("topic2");
var sourceProcessors = new Dictionary <string, IProcessor>()
{
{ "source1", sourceProcessorMock.Object }, { "source2", otherSourceProcessorMock.Object }
};
var processors = new Dictionary <string, IProcessor>()
{
{ "processor", processorMock.Object }
};
var storesToTopics = new Dictionary <string, string>()
{
{ "store1", "topic1" },
{ "store2", "topic2" },
};
stateManagerMock.Setup(x => x.Initialize()).Returns(new HashSet <string>()
{
"store1", "store2"
});
var topology = new ProcessorTopology(
null,
sourceProcessors,
null,
processors,
null,
null,
storesToTopics,
null);
globalStateUpdateTask = new GlobalStateUpdateTask(stateManagerMock.Object, topology, contextMock.Object);
}
public void SetUp()
{
streamConfigMock = new Mock <IStreamConfig>();
streamConfigMock.Setup(x => x.MetadataRequestTimeoutMs).Returns(1);
kvStoreMock = CreateMockStore <IKeyValueStore <object, object> >(kvStoreName);
otherStoreMock = CreateMockStore <IKeyValueStore <object, object> >(otherStoreName);
var globalStateStores = new Dictionary <string, IStateStore>()
{
{ kvStoreMock.Object.Name, kvStoreMock.Object },
{ otherStoreMock.Object.Name, otherStoreMock.Object }
};
var storesToTopics = new Dictionary <string, string>()
{
{ kvStoreMock.Object.Name, kvStoreTopic },
{ otherStoreMock.Object.Name, otherStoreTopic }
};
topology = new ProcessorTopology(
null,
new Dictionary <string, IProcessor>(),
new Dictionary <string, IProcessor>(),
new Dictionary <string, IProcessor>(),
new Dictionary <string, IStateStore>(),
globalStateStores,
storesToTopics);
adminClientMock = new Mock <IAdminClient>();
RegisterPartitionInAdminClient(kvStoreTopic);
RegisterPartitionInAdminClient(otherStoreTopic);
stateManager = new GlobalStateManager(
topology,
adminClientMock.Object,
streamConfigMock.Object
);
}
protected AbstractTask(TaskId id, IEnumerable <TopicPartition> partition, ProcessorTopology topology, IConsumer <byte[], byte[]> consumer, IStreamConfig config, IChangelogRegister changelogRegister)
{
log = Logger.GetLogger(GetType());
logPrefix = $"stream-task[{id.Id}|{id.Partition}] ";
Partition = partition;
Id = id;
Topology = topology;
this.consumer = consumer;
configuration = config;
var offsetCheckpointMngt = config.OffsetCheckpointManager
?? new OffsetCheckpointFile(Path.Combine(config.StateDir, config.ApplicationId, $"{id.Id}-{id.Partition}"));
offsetCheckpointMngt.Configure(config, id);
stateMgr = new ProcessorStateManager(
id,
partition,
topology.StoresToTopics,
changelogRegister,
offsetCheckpointMngt);
}
UnassignedStreamTask(string threadId, TaskId id, IEnumerable <TopicPartition> partitions, ProcessorTopology processorTopology, IConsumer <byte[], byte[]> consumer, IStreamConfig configuration, IKafkaSupplier kafkaSupplier, IProducer <byte[], byte[]> producer)
: base(threadId, id, partitions, processorTopology, consumer, configuration, kafkaSupplier, producer, null, new StreamMetricsRegistry())
{
config = configuration;
}
/// <summary>
/// Create a <see cref="KafkaStream"/> instance with your own <see cref="IKafkaSupplier" />
/// Please DO NOT FORGET to call Close to avoid resources leak !
/// </summary>
/// <param name="topology">the topology specifying the computational logic</param>
/// <param name="configuration">configuration about this stream</param>
/// <param name="kafkaSupplier">the Kafka clients supplier which provides underlying producer and consumer clients for the new <see cref="KafkaStream"/> instance</param>
public KafkaStream(Topology topology, IStreamConfig configuration, IKafkaSupplier kafkaSupplier)
{
this.topology = topology;
this.kafkaSupplier = kafkaSupplier;
// check if ApplicationId & BootstrapServers has been set
if (string.IsNullOrEmpty(configuration.ApplicationId) || string.IsNullOrEmpty(configuration.BootstrapServers))
{
throw new StreamConfigException($"Stream configuration is not correct. Please set ApplicationId and BootstrapServers as minimal.");
}
var processID = Guid.NewGuid();
clientId = string.IsNullOrEmpty(configuration.ClientId) ? $"{configuration.ApplicationId.ToLower()}-{processID}" : configuration.ClientId;
logPrefix = $"stream-application[{configuration.ApplicationId}] ";
logger.Info($"{logPrefix} Start creation of the stream application with this configuration: {configuration}");
// re-write the physical topology according to the config
topology.Builder.RewriteTopology(configuration);
// sanity check
var processorTopology = topology.Builder.BuildTopology();
int numStreamThreads = topology.Builder.HasNoNonGlobalTopology ? 0 : configuration.NumStreamThreads;
threads = new IThread[numStreamThreads];
var threadState = new Dictionary <long, Processors.ThreadState>();
ProcessorTopology globalTaskTopology = topology.Builder.BuildGlobalStateTopology();
bool hasGlobalTopology = globalTaskTopology != null;
if (numStreamThreads == 0 && !hasGlobalTopology)
{
throw new TopologyException("Topology has no stream threads and no global threads, " +
"must subscribe to at least one source topic or global table.");
}
GlobalThreadState globalThreadState = null;
if (hasGlobalTopology)
{
string globalThreadId = $"{clientId}-GlobalStreamThread";
GlobalStreamThreadFactory globalStreamThreadFactory = new GlobalStreamThreadFactory(globalTaskTopology,
globalThreadId,
kafkaSupplier.GetGlobalConsumer(configuration.ToGlobalConsumerConfig(globalThreadId)),
configuration,
kafkaSupplier.GetAdmin(configuration.ToAdminConfig(clientId)));
globalStreamThread = globalStreamThreadFactory.GetGlobalStreamThread();
globalThreadState = globalStreamThread.State;
}
List <StreamThreadStateStoreProvider> stateStoreProviders = new List <StreamThreadStateStoreProvider>();
for (int i = 0; i < numStreamThreads; ++i)
{
var threadId = $"{configuration.ApplicationId.ToLower()}-stream-thread-{i}";
var adminClient = this.kafkaSupplier.GetAdmin(configuration.ToAdminConfig(StreamThread.GetSharedAdminClientId(clientId)));
threads[i] = StreamThread.Create(
threadId,
clientId,
this.topology.Builder,
configuration,
this.kafkaSupplier,
adminClient,
i);
threadState.Add(threads[i].Id, threads[i].State);
stateStoreProviders.Add(new StreamThreadStateStoreProvider(threads[i], this.topology.Builder));
}
var manager = new StreamStateManager(this, threadState, globalThreadState);
if (hasGlobalTopology)
{
globalStreamThread.StateChanged += manager.OnGlobalThreadStateChange;
}
foreach (var t in threads)
{
t.StateChanged += manager.OnChange;
}
var globalStateStoreProvider = new GlobalStateStoreProvider(topology.Builder.GlobalStateStores);
queryableStoreProvider = new QueryableStoreProvider(stateStoreProviders, globalStateStoreProvider);
StreamState = State.CREATED;
}
public GlobalStateManager(ProcessorTopology topology, IAdminClient adminClient, IStreamConfig config)
{
this.topology = topology;
this.adminClient = adminClient;
this.config = config;
}
public GlobalStateUpdateTask(IGlobalStateManager globalStateManager, ProcessorTopology topology, ProcessorContext context)
{
this.globalStateManager = globalStateManager;
this.topology = topology;
this.context = context;
}
public ProcessorTopology BuildTopology()
{
var topology = new ProcessorTopology(root, sourceOperators, sinkOperators, processorOperators, stateStores);
return(topology);
}
public StreamTask(string threadId, TaskId id, IEnumerable <TopicPartition> partitions, ProcessorTopology processorTopology, IConsumer <byte[], byte[]> consumer, IStreamConfig configuration, IKafkaSupplier kafkaSupplier, IProducer <byte[], byte[]> producer)
: base(id, partitions, processorTopology, consumer, configuration)
{
this.threadId = threadId;
this.kafkaSupplier = kafkaSupplier;
consumedOffsets = new Dictionary <TopicPartition, long>();
maxTaskIdleMs = configuration.MaxTaskIdleMs;
maxBufferedSize = configuration.BufferedRecordsPerPartition;
idleStartTime = -1;
// eos enabled
if (producer == null)
{
this.producer = CreateEOSProducer();
InitializeTransaction();
eosEnabled = true;
}
else
{
this.producer = producer;
}
collector = new RecordCollector(logPrefix);
collector.Init(ref this.producer);
Context = new ProcessorContext(this, configuration, stateMgr).UseRecordCollector(collector);
var partitionsQueue = new Dictionary <TopicPartition, RecordQueue>();
foreach (var p in partitions)
{
var sourceProcessor = processorTopology.GetSourceProcessor(p.Topic);
var sourceTimestampExtractor = sourceProcessor.Extractor ?? configuration.DefaultTimestampExtractor;
var queue = new RecordQueue(
logPrefix,
$"record-queue-{p.Topic}-{id.Id}-{id.Partition}",
sourceTimestampExtractor,
sourceProcessor);
partitionsQueue.Add(p, queue);
processors.Add(sourceProcessor);
}
partitionGrouper = new PartitionGrouper(partitionsQueue);
}
public StreamTask(string threadId, TaskId id, IEnumerable <TopicPartition> partitions,
ProcessorTopology processorTopology, IConsumer <byte[], byte[]> consumer, IStreamConfig configuration,
IKafkaSupplier kafkaSupplier, IProducer <byte[], byte[]> producer, IChangelogRegister changelogRegister,
StreamMetricsRegistry streamMetricsRegistry)
: base(id, partitions, processorTopology, consumer, configuration, changelogRegister)
{
this.threadId = threadId;
this.kafkaSupplier = kafkaSupplier;
this.streamMetricsRegistry = streamMetricsRegistry;
consumedOffsets = new Dictionary <TopicPartition, long>();
maxTaskIdleMs = configuration.MaxTaskIdleMs;
maxBufferedSize = configuration.BufferedRecordsPerPartition;
followMetadata = configuration.FollowMetadata;
idleStartTime = -1;
// eos enabled
if (producer == null)
{
this.producer = CreateEOSProducer();
InitializeTransaction();
eosEnabled = true;
}
else
{
this.producer = producer;
}
var droppedRecordsSensor = TaskMetrics.DroppedRecordsSensor(this.threadId, Id, this.streamMetricsRegistry);
collector = new RecordCollector(logPrefix, configuration, id, droppedRecordsSensor);
collector.Init(ref this.producer);
Context = new ProcessorContext(this, configuration, stateMgr, streamMetricsRegistry)
.UseRecordCollector(collector);
Context.FollowMetadata = followMetadata;
var partitionsQueue = new Dictionary <TopicPartition, RecordQueue>();
foreach (var p in partitions)
{
var sourceProcessor = processorTopology.GetSourceProcessor(p.Topic);
sourceProcessor.SetTaskId(id);
var sourceTimestampExtractor = sourceProcessor.Extractor ?? configuration.DefaultTimestampExtractor;
var queue = new RecordQueue(
logPrefix,
$"record-queue-{p.Topic}-{id.Id}-{id.Partition}",
sourceTimestampExtractor,
p,
sourceProcessor,
droppedRecordsSensor);
partitionsQueue.Add(p, queue);
processors.Add(sourceProcessor);
}
partitionGrouper = new PartitionGrouper(partitionsQueue);
closeTaskSensor = ThreadMetrics.ClosedTaskSensor(this.threadId, streamMetricsRegistry);
activeBufferedRecordSensor = TaskMetrics.ActiveBufferedRecordsSensor(this.threadId, Id, streamMetricsRegistry);
processSensor = TaskMetrics.ProcessSensor(this.threadId, Id, streamMetricsRegistry);
processLatencySensor = TaskMetrics.ProcessLatencySensor(this.threadId, Id, streamMetricsRegistry);
enforcedProcessingSensor = TaskMetrics.EnforcedProcessingSensor(this.threadId, Id, streamMetricsRegistry);
commitSensor = TaskMetrics.CommitSensor(this.threadId, Id, streamMetricsRegistry);
activeRestorationSensor = TaskMetrics.ActiveRestorationSensor(this.threadId, Id, streamMetricsRegistry);
restorationRecordsSendsor = TaskMetrics.RestorationRecordsSensor(this.threadId, Id, streamMetricsRegistry);
}