Confluent Schema Registry cannot be restarted

[jhamm...@squarespace.com]

Good evening Confluent Platform Google Group,

I am encountering a strange error in my development environment. I have a single zookeeper server, and a single Kafka broker. The daemon launches with no error messages and I have a listener on port 8081. If I stop the server with schema-registry-stop, subsequent attempts to start the daemon fail. The error log output is:

[2017-12-15 23:12:55,557] INFO Initializing KafkaStore with broker endpoints: PLAINTEXT://<kafka host>t:9092 (io.confluent.kafka.schemaregistry.storage.KafkaStore)

[2017-12-15 23:12:55,822] INFO Validating schemas topic <schema topic> (io.confluent.kafka.schemaregistry.storage.KafkaStore)

[2017-12-15 23:12:55,831] WARN The replication factor of the schema topic <schema topic> is less than the desired one of 3. If this is a production environment, it's crucial to add more brokers and increase the replication factor of the topic. (io.confluent.kafka.schemaregistry.storage.KafkaStore)

[2017-12-15 23:12:55,838] INFO Shutting down schema registry (io.confluent.kafka.schemaregistry.storage.KafkaSchemaRegistry)

[2017-12-15 23:12:55,839] ERROR Server died unexpectedly:  (io.confluent.kafka.schemaregistry.rest.SchemaRegistryMain)

java.lang.NullPointerException

        at io.confluent.kafka.schemaregistry.storage.KafkaStore.close(KafkaStore.java:366)

        at io.confluent.kafka.schemaregistry.storage.KafkaSchemaRegistry.close(KafkaSchemaRegistry.java:720)

        at io.confluent.kafka.schemaregistry.rest.SchemaRegistryRestApplication.onShutdown(SchemaRegistryRestApplication.java:111)

        at io.confluent.kafka.schemaregistry.rest.SchemaRegistryRestApplication.setupResources(SchemaRegistryRestApplication.java:66)

        at io.confluent.kafka.schemaregistry.rest.SchemaRegistryRestApplication.setupResources(SchemaRegistryRestApplication.java:42)

        at io.confluent.rest.Application.createServer(Application.java:157)

        at io.confluent.kafka.schemaregistry.rest.SchemaRegistryMain.main(SchemaRegistryMain.java:43)

Is it possible that there are topic elements that are left null when the daemon performs it's cleanup routine? This bug report seems to be related:

https://github.com/confluentinc/schema-registry/issues/693

OS: CentOS 7.4

Confluent version: confluent-platform-oss-2.11-4.0.0-1.noarch

Kafka version: 2.10

Thanks in advance for any insight or troubleshooting assistance this mesage board can provide!

Joseph Hammerman

[jhamm...@squarespace.com]

Good morning Confluent Platform Google Group,

I was able to resolve this by downgrading the confluent-schema-registry package to the 3.1 branch. If anyone on the board would like this issue filed as a bug against 4.0.0, please respond to this message.

Thanks all,

Joseph Hammerman

[mag...@confluent.io]

The Null Pointer Exception is just a  side effect of Schema Registry not initializing successfully. Unfortunately, its also swallowing the actual exception that caused SR to not initialize. The NPE is already fixed by the community in master. Having said that, I don't think the version here really matters.

[jhamm...@squarespace.com]

Good morning Mag,

I'm confused by your response. If the version is irrelevant, why is the behavior reproducible in 4.0.0, but does not manifest in 3.1?

I agree with your analysis regarding the real error message getting swallowed. It seems that you are indicating that the root cause is a known issue that has been addressed? When is the patch slated for release? Do you by any chance have a link to the commit that contains the fix?

Thanks in advance for any further guidance you can provide.

Joseph Hammemran