ReadOnlyKeyValueStore from a KTable using spring-kafka - spring-boot

I am migrating a Kafka Streams implementation which uses pure Kafka apis to use spring-kafka instead as it's incorporated in a spring-boot application.
Everything works fine the Stream, GlobalKTable, branching that I have all works perfectly fine but I am having a hard time incorporating a ReadOnlyKeyValueStore. Based on the spring-kafka documentation here: https://docs.spring.io/spring-kafka/docs/2.6.10/reference/html/#streams-spring
It says:
If you need to perform some KafkaStreams operations directly, you can
access that internal KafkaStreams instance by using
StreamsBuilderFactoryBean.getKafkaStreams(). You can autowire
StreamsBuilderFactoryBean bean by type, but you should be sure to use
the full type in the bean definition.
Based on that I tried to incorporate it to my example as in the following fragments below:
#Bean(name = KafkaStreamsDefaultConfiguration.DEFAULT_STREAMS_CONFIG_BEAN_NAME)
public KafkaStreamsConfiguration defaultKafkaStreamsConfig() {
Map<String, Object> props = defaultStreamsConfigs();
props.put(StreamsConfig.APPLICATION_ID_CONFIG, "quote-stream");
props.put(StreamsConfig.DEFAULT_KEY_SERDE_CLASS_CONFIG, Serdes.String().getClass().getName());
props.put(StreamsConfig.DEFAULT_VALUE_SERDE_CLASS_CONFIG, SpecificAvroSerde.class);
props.put(ConsumerConfig.GROUP_ID_CONFIG, "stock-quotes-stream-group");
return new KafkaStreamsConfiguration(props);
}
#Bean(name = KafkaStreamsDefaultConfiguration.DEFAULT_STREAMS_BUILDER_BEAN_NAME)
public StreamsBuilderFactoryBean defaultKafkaStreamsBuilder(KafkaStreamsConfiguration defaultKafkaStreamsConfig) {
return new StreamsBuilderFactoryBean(defaultKafkaStreamsConfig);
}
...
final GlobalKTable<String, LeveragePrice> leverageBySymbolGKTable = streamsBuilder
.globalTable(KafkaConfiguration.LEVERAGE_PRICE_TOPIC,
Materialized.<String, LeveragePrice, KeyValueStore<Bytes, byte[]>>as("leverage-by-symbol-table")
.withKeySerde(Serdes.String())
.withValueSerde(leveragePriceSerde));
leveragePriceView = myKStreamsBuilder.getKafkaStreams().store("leverage-by-symbol-table", QueryableStoreTypes.keyValueStore());
But adding the StreamsBuilderFactoryBean(which seems to be needed to get a reference to KafkaStreams) definition causes an error:
The bean 'defaultKafkaStreamsBuilder', defined in class path resource [com/resona/springkafkastream/repository/KafkaConfiguration.class], could not be registered. A bean with that name has already been defined in class path resource [org/springframework/kafka/annotation/KafkaStreamsDefaultConfiguration.class] and overriding is disabled.
The issue is I don't want to control the lifecycle of the stream that's what I get with the plain Kafka APIs so I would like to get a reference to the default managed one as I want spring to manage it but whenever I try to expose the bean it gives the error. Any ideas on what's the correct approach to that using spring-kafka?
P.S - I am not interested in solutions using spring-cloud-stream I am looking for implementations of spring-kafka.

You don't need to define any new beans; something like this should work...
spring.application.name=quote-stream
spring.kafka.streams.properties.default.key.serde=org.apache.kafka.common.serialization.Serdes$StringSerde
spring.kafka.streams.properties.default.value.serde=org.apache.kafka.common.serialization.Serdes$StringSerde
#SpringBootApplication
#EnableKafkaStreams
public class So69669791Application {
public static void main(String[] args) {
SpringApplication.run(So69669791Application.class, args);
}
#Bean
GlobalKTable<String, String> leverageBySymbolGKTable(StreamsBuilder sb) {
return sb.globalTable("gkTopic",
Materialized.<String, String, KeyValueStore<Bytes, byte[]>> as("leverage-by-symbol-table"));
}
private ReadOnlyKeyValueStore<String, String> leveragePriceView;
#Bean
StreamsBuilderFactoryBean.Listener afterStart(StreamsBuilderFactoryBean sbfb,
GlobalKTable<String, String> leverageBySymbolGKTable) {
StreamsBuilderFactoryBean.Listener listener = new StreamsBuilderFactoryBean.Listener() {
#Override
public void streamsAdded(String id, KafkaStreams streams) {
leveragePriceView = streams.store("leverage-by-symbol-table", QueryableStoreTypes.keyValueStore());
}
};
sbfb.addListener(listener);
return listener;
}
#Bean
KStream<String, String> stream(StreamsBuilder builder) {
KStream<String, String> stream = builder.stream("someTopic");
stream.to("otherTopic");
return stream;
}
}

Related

Validate Consumer Group is created on topic or not

I am working on project where I need to validate consumer group is created on topic or not. Is there any way in boldSpring Kafkabold to validate it
Currently, I haven't seen describeConsumerGroups supported in Spring-Kafka KafkaAdmin. So, you may need to create a Kafka AdminClient and call the method by yourself.
E.g: Here, I took advantage of the auto-configuration property class KafkaProperties and autowired it to the service.
#Service
public class KafkaBrokerService implements BrokerService {
private Map<String, Object> configs;
public KafkaBrokerService(KafkaProperties kafkaProperties) { // Autowired
this.configs = kafkaProperties.buildAdminProperties();
}
private AdminClient createAdmin() {
Map<String, Object> configs2 = new HashMap<>(this.configs);
return AdminClient.create(configs2);
}
public SomeDto consumerGroupDescription(String groupId) {
try (AdminClient adminClient = createAdmin()) {
// ConsumerGroup's members
ConsumerGroupDescription consumerGroupDescription = adminClient.describeConsumerGroups(Collections.singletonList(groupId))
.describedGroups().get(groupId).get();
// ConsumerGroup's partitions and the committed offset in each partition
Map<TopicPartition, OffsetAndMetadata> offsets = adminClient.listConsumerGroupOffsets(groupId).partitionsToOffsetAndMetadata().get();
// When you get the information, you can validate it here.
...
} catch (ExecutionException | InterruptedException e) {
//
}
}
}

StateStore is never added on Spring cloud

Any Help how can I add state store on Spring cloud
I always receive this error "nested exception is org.springframework.kafka.KafkaException: Could not start stream: ; nested exception is org.apache.kafka.streams.errors.TopologyException: Invalid topology: StateStore myStore is not added yet."
Here is the bean definition however it never works
#Bean
public StoreBuilder storeBuilder() {
KeyValueBytesStoreSupplier storeSupplier = Stores.persistentKeyValueStore("mystore");
StoreBuilder<KeyValueStore<String, MyData>> storeBuilder = Stores.keyValueStoreBuilder(storeSupplier, Serdes.String(), StreamsSerde.MyDataSerde());
return storeBuilder;
}
Here is the Serde
public static final class MyDataSerde extends Serdes.WrapperSerde<MyData> {
public MyDataSerde() {
super(new JsonSerializer<>(), new JsonDeserializer<>(MyData.class));
}
}
Here is the data class
public class MyData {
private String name;
private String course;
}
Here is the spring cloud dependencies
springBootVersion = "2.2.5.RELEASE"
set('springCloudVersion', "Hoxton.SR3")
implementation group:"org.springframework.cloud", name: "spring-cloud-stream"
implementation group: "org.springframework.cloud", name: "spring-cloud-stream-binder-kafka-streams"
implementation group: "org.springframework.cloud", name: "spring-cloud-starter-stream-kafka"
You need to add state stores like this when you have to use the lower level processor or transformer API. Did you try to add the state store to your process or transform method call? Here is a test that works. Take a look at the process call and the way the state stores are passed along.
I found a solution to add the store programmatically on this article
public void initializeStateStores() throws Exception {
StreamsBuilderFactoryBean streamsBuilderFactoryBean =
applicationContext.getBean("&stream-builder-requestListener", StreamsBuilderFactoryBean.class);
StreamsBuilder streamsBuilder = streamsBuilderFactoryBean.getObject();
StoreBuilder<KeyValueStore<String, Long>> keyValueStoreBuilder = Stores.keyValueStoreBuilder(Stores.persistentKeyValueStore(stateStoreName), Serdes.String(), Serdes.Long());
streamsBuilder.addStateStore(keyValueStoreBuilder);
}
https://medium.com/#daniyaryeralin/utilizing-kafka-streams-processor-api-and-implementing-custom-aggregator-6cb23d00eaa7

Spring Kafka global transaction ID stays open after program ends

I am creating a Kafka Spring producer under Spring Boot which will send data to Kafka and then write to a database; I want all that work to be in one transaction. I am new to Kafka and no expert on Spring, and am having some difficulty. Any pointers much appreciated.
So far my code writes to Kafka successfully in a loop. I have not yet set up
the DB, but have proceeded to set up global transactioning by adding a transactionIdPrefix to the producerFactory in the configuration:
producerFactory.setTransactionIdPrefix("MY_SERVER");
and added #Transactional to the method that does the Kafka send. Eventually I plan to do my DB work in that same method.
Problem: the code runs great the first time. But if I stop the program, even cleanly, I find that the code hangs the 2nd time I run it as soon as it enters the #Transactional method. If I comment out the #Transactional, it enters the method but hangs on the kafa template send().
The problem seems to be the transaction ID. If I change the prefix and rerun, the program runs fine again the first time but hangs when I run it again, until a new prefix is chosen. Since after a restart the trans ID counter starts at zero, if the trans ID prefix does not change then the same trans ID will be used upon restart.
It seems to me that the original transID is still open on the server, and was never committed. (I can read the data off the topic using the console-consumer, but that will read uncommitted). But if that is the case, how do I get spring to commit the trans? I am thinking my coniguration must be wrong. Or-- is the issue possibly that trans ID's can never be reused? (In which case, how does one solve that?)
Here is my relevant code. Config is:
#SpringBootApplication
public class MYApplication {
#Autowired
private static ChangeSweeper changeSweeper;
#Value("${kafka.bootstrap-servers}")
private String bootstrapServers;
#Bean
public ProducerFactory<String, String> producerFactory() {
Map<String, Object> configProps = new HashMap<>();
configProps.put(ProducerConfig.BOOTSTRAP_SERVERS_CONFIG, bootstrapServers);
configProps.put(ProducerConfig.KEY_SERIALIZER_CLASS_CONFIG, StringSerializer.class);
configProps.put(ProducerConfig.VALUE_SERIALIZER_CLASS_CONFIG, StringSerializer.class);
DefaultKafkaProducerFactory<String, String> producerFactory=new DefaultKafkaProducerFactory<>(configProps);
producerFactory.setTransactionIdPrefix("MY_SERVER");
return producerFactory;
}
#Bean
public KafkaTransactionManager<String, String> KafkaTransactionManager() {
return new KafkaTransactionManager<String, String>((producerFactory()));
}
#Bean(name="kafkaProducerTemplate")
public KafkaTemplate<String, String> kafkaProducerTemplate() {
return new KafkaTemplate<>(producerFactory());
}
And the method that does the transaction is:
#Transactional
public void send( final List<Record> records) {
logger.debug("sending {} records; batchSize={}; topic={}", records.size(),batchSize, kafkaTopic);
// Divide the record set into batches of size batchSize and send each batch with a kafka transaction:
for (int batchStartIndex = 0; batchStartIndex < records.size(); batchStartIndex += batchSize ) {
int batchEndIndex=Math.min(records.size()-1, batchStartIndex+batchSize-1);
List<Record> nextBatch = records.subList(batchStartIndex, batchEndIndex);
logger.debug("## batch is from " + batchStartIndex + " to " + batchEndIndex);
for (Record record : nextBatch) {
kafkaProducerTemplate.send( kafkaTopic, record.getKey().toString(), record.getData().toString());
logger.debug("Sending> " + record);
}
// I will put the DB writes here
}
This works fine for me no matter how many times I run it (but I have to run 3 broker instances on my local machine because transactions require that by default)...
#SpringBootApplication
#EnableTransactionManagement
public class So47817034Application {
public static void main(String[] args) {
SpringApplication.run(So47817034Application.class, args).close();
}
private final CountDownLatch latch = new CountDownLatch(2);
#Bean
public ApplicationRunner runner(Foo foo) {
return args -> {
foo.send("foo");
foo.send("bar");
this.latch.await(10, TimeUnit.SECONDS);
};
}
#Bean
public KafkaTransactionManager<Object, Object> KafkaTransactionManager(KafkaProperties properties) {
return new KafkaTransactionManager<Object, Object>(kafkaProducerFactory(properties));
}
#Bean
public ProducerFactory<Object, Object> kafkaProducerFactory(KafkaProperties properties) {
DefaultKafkaProducerFactory<Object, Object> factory =
new DefaultKafkaProducerFactory<Object, Object>(properties.buildProducerProperties());
factory.setTransactionIdPrefix("foo-");
return factory;
}
#KafkaListener(id = "foo", topics = "so47817034")
public void listen(String in) {
System.out.println(in);
this.latch.countDown();
}
#Component
public static class Foo {
#Autowired
private KafkaTemplate<Object, Object> template;
#Transactional
public void send(String go) {
this.template.send("so47817034", go);
}
}
}

JAXBElement: providing codec (/converter?) for class java.lang.Class

I have been evaluating to adopt spring-data-mongodb for a project. In summary, my aim is:
Using existing XML schema files to generate Java classes.
This is achieved using JAXB xjc
The root class is TSDProductDataType and is further modeled as below:
The thing to note here is that ExtensionType contains protected List<Object> any; allowing it to store Objects of any class. In my case, it is amongst the classes named TSDModule_Name_HereModuleType and can be browsed here
Use spring-data-mongodb as persistence store
This is achieved using a simple ProductDataRepository
#RepositoryRestResource(collectionResourceRel = "product", path = "product")
public interface ProductDataRepository extends MongoRepository<TSDProductDataType, String> {
TSDProductDataType queryByGtin(#Param("gtin") String gtin);
}
The unmarshalled TSDProductDataType, however, contains JAXBElement which spring-data-mongodb doesn't seem to handle by itself and throws a CodecConfigurationException org.bson.codecs.configuration.CodecConfigurationException: Can't find a codec for class java.lang.Class.
Here is the faulty statement:
TSDProductDataType tsdProductDataType = jaxbElement.getValue();
repository.save(tsdProductDataType);
I tried playing around with Converters for spring-data-mongodb as explained here, however, it seems I am missing something since the exception is about "Codecs" and not "Converters".
Any help is appreciated.
EDIT:
Adding converters for JAXBElement
Note: Works with version 1.5.6.RELEASE of org.springframework.boot::spring-boot-starter-parent. With version 2.0.0.M3, hell breaks loose
It seems that I missed something while trying to add converter earlier. So, I added it like below for testing:
#Component
#ReadingConverter
public class JAXBElementReadConverter implements Converter<DBObject, JAXBElement> {
//#Autowired
//MongoConverter converter;
#Override
public JAXBElement convert(DBObject dbObject) {
Class declaredType, scope;
QName name = qNameFromString((String)dbObject.get("name"));
Object rawValue = dbObject.get("value");
try {
declaredType = Class.forName((String)dbObject.get("declaredType"));
} catch (ClassNotFoundException e) {
if (rawValue.getClass().isArray()) declaredType = List.class;
else declaredType = LinkedHashMap.class;
}
try {
scope = Class.forName((String) dbObject.get("scope"));
} catch (ClassNotFoundException e) {
scope = JAXBElement.GlobalScope.class;
}
//Object value = rawValue instanceof DBObject ? converter.read(declaredType, (DBObject) rawValue) : rawValue;
Object value = "TODO";
return new JAXBElement(name, declaredType, scope, value);
}
QName qNameFromString(String s) {
String[] parts = s.split("[{}]");
if (parts.length > 2) return new QName(parts[1], parts[2], parts[0]);
if (parts.length == 1) return new QName(parts[0]);
return new QName("undef");
}
}
#Component
#WritingConverter
public class JAXBElementWriteConverter implements Converter<JAXBElement, DBObject> {
//#Autowired
//MongoConverter converter;
#Override
public DBObject convert(JAXBElement jaxbElement) {
DBObject dbObject = new BasicDBObject();
dbObject.put("name", qNameToString(jaxbElement.getName()));
dbObject.put("declaredType", jaxbElement.getDeclaredType().getName());
dbObject.put("scope", jaxbElement.getScope().getCanonicalName());
//dbObject.put("value", converter.convertToMongoType(jaxbElement.getValue()));
dbObject.put("value", "TODO");
dbObject.put("_class", JAXBElement.class.getName());
return dbObject;
}
public String qNameToString(QName name) {
if (name.getNamespaceURI() == XMLConstants.NULL_NS_URI) return name.getLocalPart();
return name.getPrefix() + '{' + name.getNamespaceURI() + '}' + name.getLocalPart();
}
}
#SpringBootApplication
public class TsdApplication {
public static void main(String[] args) {
SpringApplication.run(TsdApplication.class, args);
}
#Bean
public CustomConversions customConversions() {
return new CustomConversions(Arrays.asList(
new JAXBElementReadConverter(),
new JAXBElementWriteConverter()
));
}
}
So far so good. However, how do I instantiate MongoConverter converter;?
MongoConverter is an interface so I guess I need an instantiable class adhering to this interface. Any suggestions?
I understand the desire for convenience in being able to just map an existing domain object to the database layer with no boilerplate, but even if you weren't having the JAXB class structure issue, I would still be recommending away from using it verbatim. Unless this is a simple one-off project, you almost definitely will hit a point where your domain models will need to change but your persisted data need to remain in an existing state. If you are just straight persisting the data, you have no mechanism to convert between a newer domain schema and an older persisted data scheme. Versioning of the persisted data scheme would be wise too.
The link you posted for writing the customer converters is one way to achieve this and fits in nicely with the Spring ecosystem. That method should also solve the issue you are experiencing (about the underlying messy JAXB data structure not converting cleanly).
Are you unable to get that method working? Ensure you are loading them into the Spring context with #Component plus auto-class scanning or manually via some Configuration class.
EDIT to address your EDIT:
Add the following to each of your converters:
private final MongoConverter converter;
public JAXBElement____Converter(MongoConverter converter) {
this.converter = converter;
}
Try changing your bean definition to:
#Bean
public CustomConversions customConversions(#Lazy MongoConverter converter) {
return new CustomConversions(Arrays.asList(
new JAXBElementReadConverter(converter),
new JAXBElementWriteConverter(converter)
));
}

How to make writer initiated for every job instance in spring batch

I am writing a spring batch job. I am implementing custom writer using KafkaClientWriter extends AbstractItemStreamItemWriter<ProducerMessage>
I have fields which need to be unique for each instance. But I could see this class initiated only once. Rest jobs have same instance of writer class.
Where as my custom readers and processors are getting initiated for each job.
Below is my job configurations. How can I achieve the same behavior for writer as well?
#Bean
#Scope("job")
public ZipMultiResourceItemReader reader(#Value("#{jobParameters[fileName]}") String fileName, #Value("#{jobParameters[s3SourceFolderPrefix]}") String s3SourceFolderPrefix, #Value("#{jobParameters[timeStamp]}") long timeStamp, com.fastretailing.catalogPlatformSCMProducer.service.ConfigurationService confService) {
FlatFileItemReader faltFileReader = new FlatFileItemReader();
ZipMultiResourceItemReader zipReader = new ZipMultiResourceItemReader();
Resource[] resArray = new Resource[1];
resArray[0] = new FileSystemResource(new File(fileName));
zipReader.setArchives(resArray);
DefaultLineMapper<ProducerMessage> lineMapper = new DefaultLineMapper<ProducerMessage>();
lineMapper.setLineTokenizer(new DelimitedLineTokenizer());
CSVFieldMapper csvFieldMapper = new CSVFieldMapper(fileName, s3SourceFolderPrefix, timeStamp, confService);
lineMapper.setFieldSetMapper(csvFieldMapper);
faltFileReader.setLineMapper(lineMapper);
zipReader.setDelegate(faltFileReader);
return zipReader;
}
#Bean
#Scope("job")
public ItemProcessor<ProducerMessage, ProducerMessage> processor(#Value("#{jobParameters[timeStamp]}") long timeStamp) {
ProducerProcessor processor = new ProducerProcessor();
processor.setS3FileTimeStamp(timeStamp);
return processor;
}
#Bean
#ConfigurationProperties
public ItemWriter<ProducerMessage> writer() {
return new KafkaClientWriter();
}
#Bean
public Step step1(StepBuilderFactory stepBuilderFactory,
ItemReader reader, ItemWriter writer,
ItemProcessor processor, #Value("${reader.chunkSize}")
int chunkSize) {
LOGGER.info("Step configuration loaded with chunk size {}", chunkSize);
return stepBuilderFactory.get("step1")
.chunk(chunkSize).reader(reader)
.processor(processor).writer(writer)
.build();
}
#Bean
public StepScope stepScope() {
final StepScope stepScope = new StepScope();
stepScope.setAutoProxy(true);
return stepScope;
}
#Bean
public JobScope jobScope() {
final JobScope jobScope = new JobScope();
return jobScope;
}
#Bean
public Configuration configuration() {
return new Configuration();
}
I tried making the writer with job scope. But in that case open is not getting called. This is where I am doing some initializations.
When using java based configuration and a scoped proxy what happens is that the return type of the method is detected and for that a proxy is created. So when you return ItemWriter you will get a JDK proxy only implementing ItemWriter, whereas your open method is on the ItemStream interface. Because that interface isn't included on the proxy there is no way to call the method.
Either change the return type to KafkaClientWriter or ItemStreamWriter< ProducerMessage> (assuming the KafkaCLientWriter implements that method). Next add #Scope("job") and you should have your open method called again with a properly scoped writer.

Resources