Bug report

For bug reports, provide this information, please:

What Debezium connector do you use and what version?

2.6.1.Final

What is the connector configuration?

"quote.identifiers": "true",
"key.converter": "io.confluent.connect.protobuf.ProtobufConverter",
"delete.enabled": "true",
"key.converter.schema.registry.url": "http://apicurio-schema-registry:8080/apis/ccompat/v6",
"primary.key.mode": "record_key",
"connection.url": "jdbc:mysql://xxx/xxx?jdbcCompliantTruncation=false",
"value.converter.schema.registry.find-latest": "true",
"connection.password": "xxx",
"database.time_zone": "UTC",
"key.converter.schema.registry.find-latest": "true",
"batch.size": "2000",
"connection.username": "xxx",
"value.converter.schema.registry.url": "http://apicurio-schema-registry:8080/apis/ccompat/v6",
"schema.evolution": "none",
"database": "xxx",
"key.converter.schema.registry.auto-register": "true",
"value.converter": "io.confluent.connect.protobuf.ProtobufConverter",
"table.name.format": "${topic}",
"tasks.max": "1",
"insert.mode": "upsert",
"connector.class": "io.debezium.connector.jdbc.JdbcSinkConnector",
"value.converter.schema.registry.auto-register": "true"

What is the captured database version and mode of depoyment?

(E.g. on-premises, with a specific cloud provider, etc.)

cloudsql mysql 5.7

What behaviour do you expect?

All update buffer must flush before delete execution

What behaviour do you see?

when size of update buffer less than connector batch size and an event of delete come, delete buffer execute and not flush update buffer, make loss a delete for the same record after it update