Hi ,we have use mysql-connector in production env more then one year,

it works well, Thank you very much!

Now ,I am work on the task to get the cdc data from oracle realtime.

I am testing by the logminer mode.

But when I insert massive data to Oracle for the test,

the data sent to the kafka by logminer were lost 

As there is no error, I can only show my env, setting and config

---------my test env--------------

env : a standby server

        core 32, more then 500g memory

ps: only one node ,not cluster

---------------Oracle-------------------

Oracle 11g  EE

redo log file size  12g 

6 groups, but I didn't set multiplexing

I didn't set  supplemental log data  on the database level

I set supplemental log on the specail tables like this

ALTER TABLE CUSER.t_order_detail_log ADD SUPPLEMENTAL LOG DATA (ALL) COLUMNS

------------some config I set manually-----------------

kafka: jvm  max 6g

kafka-connect: jvm max 6g

----kafka.properties

log.retention.bytes=536870912000

log.retention.hours=168

----kafka-conenct.properties

key.converter=io.confluent.connect.avro.AvroConverter
value.converter=io.confluent.connect.avro.AvroConverter

producer.buffer.memory=33554432

producer.batch.size=327680

producer.compression.type=lz4

consumer.heartbeat.interval.ms=6000
consumer.max.poll.records=10000
consumer.max.partition.fetch.bytes=52428800
consumer.partition.assignment.strategy=org.apache.kafka.clients.consumer.RoundRobinAssignor

— connector json

"connector.class" : "io.debezium.connector.oracle.OracleConnector",
"tasks.max" : "1",
"database.server.name" : "**",
"database.hostname" : "**",
"database.port" : "1539",
"database.user" : "logminer",
"database.password" : "logminer",
"database.dbname" : "**",
"database.tablename.case.insensitive": "true",
"database.oracle.version": "11",
"database.history.kafka.bootstrap.servers" : "****",
"database.history.kafka.topic": "schema-changes.cuser",
"database.connection.adapter": "logminer",
"schema.include.list":"FOPTRADE,CUSER",
"table.include.list":"CUSER.T_CLOSING_ACCOUNT_POS,CUSER.T_ORDER....."

----- my Data insertion simulation---

------table basic insert num---------
table.CUSER.T_CLOSING_ACCOUNT_POS=100000 
table.CUSER.T_ORDER_DETAIL_LOG=17000
table.CUSER.T_CLOSING_ACCOUNT_GREEKS=10000
table.CUSER.T_CLOSING_ACCOUNT_SELL_OP=10000 
table.CUSER.B_CLOSING_ACCOUNT=6500 
table.CUSER.T_GUARANTEE_HIST=6500 
table.CUSER.T_GUARANTEE_LOG=4000 

1. 1. data insert run time, unit minute 
      generator.runtime=1 
   2. data insert interval ,unit second
      generator.interval=20 
   3. data insert threads
      generator.thread_num=3 
      ##base_row_factor is Supported for decimals 
      generator.base_row_factor=5  # table insert num multiply this param

-------------------------------------------------------

In my test ,the data generator will insert 3830000 rows into oracle in 1 minute

I have test for 1 ,3 ,4, 7 minutes

so 3830000~ about  2000w rows  insert in a short time.

I have test more then 10 times 

80% or more probability will lose data ,

the lost num is not too much, like this

when I debugge the dbz-oracle-connector,

Small amount of data , it works well without loss,

mass amount of data ,hard to trace.

----------------

Is there any parameter or config I need to adjust?

Shall anyone can give me some guide to research?