使用KSQL将表从Kafka转储到MariaDB

Question

我正在用KSQL做聚合，需要在MariaDB中持久化输出表。我已经设置了MariaDB和JdbcSinkConnector。不幸的是，水槽就是不适合我。

这是KSQL中的表结构，我想在MariaDB中转储:

Field     | Type                           
--------------------------------------------
a        | VARCHAR(STRING)  (primary key) 
b        | VARCHAR(STRING)  (primary key) 
c        | VARCHAR(STRING)  (primary key) 
d        | INTEGER                        
--------------------------------------------

我按列a, b, c分组，并做一些聚合，即列d。

这是连接器:

create sink connector test with (
'tasks.max' = 1,
'key.converter.schema.registry.url' = 'http://schema-registry:8081', 
'value.converter.schema.registry.url' = 'http://schema-registry:8081',
'connector.class'          = 'io.confluent.connect.jdbc.JdbcSinkConnector',
'key.converter'='org.apache.kafka.connect.storage.StringConverter',
'value.converter'='org.apache.kafka.connect.storage.StringConverter',
'key.converter.schemas.enable' = 'false',
'value.converter.schemas.enable' = 'true',
'config.action.reload' = 'restart',
'errors.log.enable' = 'true',
'errors.log.include.messages' = 'true',
'print.key' = 'true',
'errors.tolerance' = 'all',
'topics'                   = 'my-topic',
'connection.url'           = 'jdbc:mysql://mariadb-docker-container:3306/ksql?autoReconnect=true&useSSL=false',
'connection.user'          = 'root', 
'connection.password'      = 'strongest-password-you-have-ever-seen',
'pk.fields' = 'a, b, c',
'pk.mode' = 'record_value',
'delete.enabled' = 'false');

运行此连接器会出现以下错误:

kafka-connect               | [2021-03-08 12:13:51,339] ERROR [TEST|task-0] WorkerSinkTask{id=TEST-0} Task threw an uncaught and unrecoverable exception (org.apache.kafka.connect.runtime.WorkerTask:187)
kafka-connect               | org.apache.kafka.connect.errors.ConnectException: Exiting WorkerSinkTask due to unrecoverable exception.
kafka-connect               |   at org.apache.kafka.connect.runtime.WorkerSinkTask.deliverMessages(WorkerSinkTask.java:591)
kafka-connect               |   at org.apache.kafka.connect.runtime.WorkerSinkTask.poll(WorkerSinkTask.java:326)
kafka-connect               |   at org.apache.kafka.connect.runtime.WorkerSinkTask.iteration(WorkerSinkTask.java:229)
kafka-connect               |   at org.apache.kafka.connect.runtime.WorkerSinkTask.execute(WorkerSinkTask.java:201)
kafka-connect               |   at org.apache.kafka.connect.runtime.WorkerTask.doRun(WorkerTask.java:185)
kafka-connect               |   at org.apache.kafka.connect.runtime.WorkerTask.run(WorkerTask.java:235)
kafka-connect               |   at java.base/java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:515)
kafka-connect               |   at java.base/java.util.concurrent.FutureTask.run(FutureTask.java:264)
kafka-connect               |   at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128)
kafka-connect               |   at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628)
kafka-connect               |   at java.base/java.lang.Thread.run(Thread.java:834)
kafka-connect               | Caused by: org.apache.kafka.connect.errors.ConnectException: Sink connector 'TEST' is configured with 'delete.enabled=false' and 'pk.mode=record_value' and therefore requires records with a non-null Struct value and non-null Struct schema, but found record at (topic='my-topic',partition=1,offset=30,timestamp=1615205630513) with a String value and string value schema.
kafka-connect               |   at io.confluent.connect.jdbc.sink.RecordValidator.lambda$requiresValue$2(RecordValidator.java:86)
kafka-connect               |   at io.confluent.connect.jdbc.sink.BufferedRecords.add(BufferedRecords.java:82)
kafka-connect               |   at io.confluent.connect.jdbc.sink.JdbcDbWriter.write(JdbcDbWriter.java:73)
kafka-connect               |   at io.confluent.connect.jdbc.sink.JdbcSinkTask.put(JdbcSinkTask.java:75)
kafka-connect               |   at org.apache.kafka.connect.runtime.WorkerSinkTask.deliverMessages(WorkerSinkTask.java:563)
kafka-connect               |   ... 10 more
kafka-connect               | [2021-03-08 12:13:51,339] ERROR [TEST|task-0] WorkerSinkTask{id=TEST-0} Task is being killed and will not recover until manually restarted (org.apache.kafka.connect.runtime.WorkerTask:188)
kafka-connect               | [2021-03-08 12:13:51,339] INFO [TEST|task-0] Stopping task (io.confluent.connect.jdbc.sink.JdbcSinkTask:119)
kafka-connect               | [2021-03-08 12:13:51,339] INFO [TEST|task-0] Closing connection #1 to MySql (io.confluent.connect.jdbc.util.CachedConnectionProvider:108)
kafka-connect               | [2021-03-08 12:13:51,340] INFO [TEST|task-0] [Consumer clientId=connector-consumer-TEST-0, groupId=connect-TEST] Revoke previously assigned partitions my-topic-1, my-topic-0 (org.apache.kafka.clients.consumer.internals.ConsumerCoordinator:307)
kafka-connect               | [2021-03-08 12:13:51,340] INFO [TEST|task-0] [Consumer clientId=connector-consumer-TEST-0, groupId=connect-TEST] Member connector-consumer-TEST-0 sending LeaveGroup request to coordinator broker:29092 (id: 2147483646 rack: null) due to the consumer is being closed (org.apache.kafka.clients.consumer.internals.AbstractCoordinator:1010)

我也尝试过使用AvroConverter os JsonConverter，但这没有帮助，他们只是抛出了其他错误(解析错误)，这并没有多大帮助。根据一些研究，我知道这与数据的结构有关。但是，如果表是由分组列和聚合产生的，那么我应该如何提供结构或将表转换为"结构"呢?

任何想法?我还在考虑干脆不使用连接器，而是编写一个小程序，将表/主题拉出来并将它们写入MariaDB，以防连接器不起作用。

Answer 1

如果你正在使用JDBC Sink，你需要为你的数据使用包含模式的序列化格式，例如使用Avro, Protobuf或JSON模式。

在ksqlDB中，您可以在创建对象时指定:

CREATE TABLE MY_TABLE WITH (FORMAT='AVRO') AS 
SELECT A,B,C,COUNT(*) AS D 
FROM STREAM_FOO 
GROUP BY A,B,C;

注意，在ksqlDB 0.15中添加了Avro键支持。

现在您的数据在Avro中，您可以使用适当的转换器创建您的接收器连接器

create sink connector test with (
'connector.class'                     = 'io.confluent.connect.jdbc.JdbcSinkConnector',
'tasks.max'                           = 1,
'key.converter.schema.registry.url'   = 'http://schema-registry:8081',
'value.converter.schema.registry.url' = 'http://schema-registry:8081',
'key.converter'                       = 'io.confluent.connect.avro.AvroConverter',
'value.converter'                     = 'io.confluent.connect.avro.AvroConverter',
'key.converter.schemas.enable'        = 'false',
'value.converter.schemas.enable'      = 'true',
'config.action.reload'                = 'restart',
'errors.log.enable'                   = 'true',
'errors.log.include.messages'         = 'true',
'print.key'                           = 'true',
'errors.tolerance'                    = 'all',
'topics'                              = 'my-topic',
'connection.url'                      = 'jdbc:mysql://mariadb-docker-container:3306/ksql?autoReconnect=true&useSSL=false',
'connection.user'                     = 'root',
'connection.password'                 = 'strongest-password-you-have-ever-seen',
'pk.fields'                           = 'a, b, c',
'pk.mode'                             = 'record_key',
'delete.enabled'                      = 'false');

你遇到的两个问题:

1. 使用StringConverter表示没有模式存在，因此错误消息中的报告为String value and string value schema.。
2. 表的键(GROUP BY)被写入Kafka消息的键，所以从这里(record_key)你应该得到pk.fields。

裁判:

• Kafka Connect Deep Dive - converter and Serialization Explained
• 🎥ksqlDB和Kafka连接JDBC Sink(✍️notes)