Spring Cloud Stream header enricher circular reference and NPE

Question

Using Spring Cloud Data Flow Server and shell 1.3.0.M1, the Kafka binder, and loading apps from the Bacon bit-ly URL (1.2.0.RELEASE).

I created/deployed a stream in SCDF shell:

source | httpclient <args> | header-enricher --headers=\"key=payload\" | log

and received exception in the header-enricher log:

2017-08-29 16:37:16,991  WARN main o.s.b.c.e.AnnotationConfigEmbeddedWebApplicationContext:550 - 
Exception encountered during context initialization - cancelling refresh attempt: 
org.springframework.beans.factory.BeanCreationException: Error creating bean with name
'org.springframework.cloud.stream.app.header.enricher.processor.
 HeaderEnricherProcessorConfiguration': Initialization of bean failed; nested exception is org.springframework.beans.factory.BeanCreationException: Error creating bean with name 'headerEnricher' defined in org.springframework.cloud.stream.app.header.enricher.processor.HeaderEnricherProcessorConfiguration: Bean instantiation via factory method failed; nested exception is 
org.springframework.beans.BeanInstantiationException: Failed to instantiate 
[org.springframework.integration.transformer.HeaderEnricher]: Circular reference involving 
containing bean 
'org.springframework.cloud.stream.app.header.enricher.processor.
 HeaderEnricherProcessorConfiguration' - consider declaring the factory method as static 
 for independence from its containing instance. Factory method 'headerEnricher' 
 threw exception; nested exception is java.lang.NullPointerException

Also tried header-enricher --headers=token='\"foo\"' (string value) with the same result.

Am I doing something wrong?

Answer 1

I unregistered the header-enricher app from the Bacon.RELEASE (1.2.0) and replaced it with the latest snapshot (1.3.0.BUILD-SNAPSHOT) using these shell commands:

dataflow:>app unregister --name header-enricher --type processor
dataflow:>app register --name header-enricher --type processor 
  --uri maven://org.springframework.cloud.stream.app:header-enricher-processor-kafka-10:1.3.0.BUILD-SNAPSHOT

When I register my test stream using --headers='key=payload', and deploy it like this:

dataflow:>stream deploy test --properties "app.header-enricher.spring.cloud.stream.kafka.binder.headers=key"

all works as expected. The circular reference exception is gone, and I see the key header in the message received by the log sink. Perhaps the circular reference is a bug that has been corrected.

---- Update ----

Whatever the issue was appears to be corrected with the Celsius.M1 milestone release. The circular reference exception no longer occurs when I use that version of the app starters.

Answer 2

The header-enricher's option headers is of type Properties and you can specify the option as --headers='key=payload'