升级到 Confluent 4.1 后，Schema Registry 不会启动

Question

我最近将 Confluent 升级到 4.1，但架构注册表似乎存在一些问题。在 confluent start schema-registry（因此 ksql-server）上无法启动。

这是我在 schema-registry 的日志中得到的错误：

[2018-04-20 11:27:38,426] ERROR Error starting the schema registry (io.confluent.kafka.schemaregistry.rest.SchemaRegistryRestApplication:65)
io.confluent.kafka.schemaregistry.exceptions.SchemaRegistryInitializationException: Error initializing kafka store while initializing schema registry
        at io.confluent.kafka.schemaregistry.storage.KafkaSchemaRegistry.init(KafkaSchemaRegistry.java:203)
        at io.confluent.kafka.schemaregistry.rest.SchemaRegistryRestApplication.setupResources(SchemaRegistryRestApplication.java:63)
        at io.confluent.kafka.schemaregistry.rest.SchemaRegistryRestApplication.setupResources(SchemaRegistryRestApplication.java:41)
        at io.confluent.rest.Application.createServer(Application.java:165)
        at io.confluent.kafka.schemaregistry.rest.SchemaRegistryMain.main(SchemaRegistryMain.java:43)
Caused by: io.confluent.kafka.schemaregistry.storage.exceptions.StoreInitializationException: io.confluent.kafka.schemaregistry.storage.exceptions.StoreException: Failed to write Noop record to kafka store.
        at io.confluent.kafka.schemaregistry.storage.KafkaStore.init(KafkaStore.java:139)
        at io.confluent.kafka.schemaregistry.storage.KafkaSchemaRegistry.init(KafkaSchemaRegistry.java:201)
        ... 4 more
Caused by: io.confluent.kafka.schemaregistry.storage.exceptions.StoreException: Failed to write Noop record to kafka store.
        at io.confluent.kafka.schemaregistry.storage.KafkaStore.getLatestOffset(KafkaStore.java:423)
        at io.confluent.kafka.schemaregistry.storage.KafkaStore.waitUntilKafkaReaderReachesLastOffset(KafkaStore.java:276)
        at io.confluent.kafka.schemaregistry.storage.KafkaStore.init(KafkaStore.java:137)
        ... 5 more
Caused by: java.util.concurrent.ExecutionException: org.apache.kafka.common.errors.NotLeaderForPartitionException: This server is not the leader for that topic-partition.
        at org.apache.kafka.clients.producer.internals.FutureRecordMetadata.valueOrError(FutureRecordMetadata.java:94)
        at org.apache.kafka.clients.producer.internals.FutureRecordMetadata.get(FutureRecordMetadata.java:77)
        at org.apache.kafka.clients.producer.internals.FutureRecordMetadata.get(FutureRecordMetadata.java:29)
        at io.confluent.kafka.schemaregistry.storage.KafkaStore.getLatestOffset(KafkaStore.java:418)
        ... 7 more
Caused by: org.apache.kafka.common.errors.NotLeaderForPartitionException: This server is not the leader for that topic-partition.
[2018-04-20 11:27:38,430] INFO Shutting down schema registry (io.confluent.kafka.schemaregistry.storage.KafkaSchemaRegistry:726)
[2018-04-20 11:27:38,430] INFO [kafka-store-reader-thread-_schemas]: Shutting down (io.confluent.kafka.schemaregistry.storage.KafkaStoreReaderThread:66)
[2018-04-20 11:27:38,431] INFO [kafka-store-reader-thread-_schemas]: Stopped (io.confluent.kafka.schemaregistry.storage.KafkaStoreReaderThread:66)
[2018-04-20 11:27:38,440] INFO [kafka-store-reader-thread-_schemas]: Shutdown completed (io.confluent.kafka.schemaregistry.storage.KafkaStoreReaderThread:66)
[2018-04-20 11:27:38,446] INFO KafkaStoreReaderThread shutdown complete. (io.confluent.kafka.schemaregistry.storage.KafkaStoreReaderThread:227)

我不知道为什么会报告此错误，并且错误消息对我来说没有意义。

失败后，confluent start schema-registry 和 confluent start ksql-server 启动了这两个服务，但是在启动 KSQL 时我收到以下警告：

**************** WARNING ******************
Remote server address may not be valid:
Error issuing GET to KSQL server
Caused by: java.net.ConnectException: Connection refused (Connection refused)
Caused by: Could not connect to the server.
*******************************************

尝试运行命令时（例如show tables;）报以下错误：

ksql> show tables;
Error issuing POST to KSQL server
Caused by: java.net.ConnectException: Connection refused (Connection refused)
Caused by: Could not connect to the server.

编辑：我已经通过破坏当前运行解决了这个问题（confluent destroy)，但如果有人能解释这个问题会很有趣。

Answer 1

根据您发布的信息，感觉您可能在某处有一些僵尸进程或错误数据，但我不能确定。

Schema Registry 抱怨它无法向 Kafka 写入消息，因为 Kafka 代理抱怨它不拥有 Schema Registry 正在写入的主题分区。这可能是由以前的 Kafka 代理（从旧安装）引起的，仍在运行。

升级前你confluent stop了吗？

使用confluent destroy，就像你做的那样，扁平化/重置安装总是一个不错的选择，只要你不珍惜你的数据。检查虚假进程（或使用旧的“重启机器”技巧）也可以是在事情未按预期运行时开始的好地方。

很高兴现在一切都整理好了：D

安迪