Fink SQL not reading Kafka data from a few hours ago

[dz902 <dz...@gmail.com>]

Hi,

I'm using Kinesis Analytics Studio which has the runtime "Apache Flink
1.13, Apache Zeppelin 0.9".

I created a table with Debezium connector:

CREATE TABLE tryme_0 (
  `id` INT PRIMARY KEY,
  `value` STRING
) WITH (
 'connector' = 'kafka',
 'topic' = 'test.dbo.tryme2',
 'properties.bootstrap.servers' = 'kafka:9092',
 'scan.startup.mode' = 'earliest-offset',
 'properties.group.id' = 'testGroup2',
 -- using 'debezium-json' as the format to interpret Debezium JSON messages
 -- please use 'debezium-avro-confluent' if Debezium encodes messages in
Avro format
 'format' = 'debezium-json'
);

And ran "SELECT * FROM tryme". It works well for new data. I aborted the
job. The second day, when I try the same query again (data timestamp would
be 8-10 hours before now), the result table is empty.

The same query worked in Flink SQL client. I can see all the data in the
topic.

Then I created another table in Zeppelin and changed the "group.id" just in
case it was some internal group id offset cache. However this still did not
work and no data coming in.

Then I tried to insert a new row. Now all the data started showing.

Is there some kind of implicit windowing in Zeppelin? How do I debug this?

Thanks,
Dai