It deleted the theme it had stored offset on its own,

Then configure auto.offset.reset as the earliest task. If the offset cannot be read, the data will be consumed from the oldest, causing duplicate consumption and backlog. The entire __consumer_offsets will be deleted and rebuilt. The time I read the oldest data is the time after deletion

The specific investigation logs only show deletion operations, and no logs with exceptions were found

Here are the logs

[2025-03-14 10:38:27,293] DEBUG [Controller id=3] Topics not in preferred replica for broker 3 Map() (kafka.controller.KafkaController)

[2025-03-14 10:38:27,293] TRACE [Controller id=3] Leader imbalance ratio for broker 3 is 0.0 (kafka.controller.KafkaController)

[2025-03-14 10:38:27,293] DEBUG [Controller id=3] Topics not in preferred replica for broker 4 Map() (kafka.controller.KafkaController)

[2025-03-14 10:38:27,293] TRACE [Controller id=3] Leader imbalance ratio for broker 4 is 0.0 (kafka.controller.KafkaController)

[2025-03-14 10:41:19,863] DEBUG [Controller id=3] Delete topics listener fired for topics __consumer_offsets to be deleted (kafka.controller.KafkaController)

[2025-03-14 10:41:19,863] INFO [Controller id=3] Starting topic deletion for topics __consumer_offsets (kafka.controller.KafkaController)

[2025-03-14 10:41:19,932] INFO [Topic Deletion Manager 3] Handling deletion for topics __consumer_offsets (kafka.controller.TopicDeletionManager)

[2025-03-14 10:41:19,932] INFO [Topic Deletion Manager 3] Deletion of topic __consumer_offsets (re)started (kafka.controller.TopicDeletionManager)

[2025-03-14 10:41:20,259] INFO [RequestSendThread controllerId=3] Controller 3 connected to 172.21.216.57:9092 (id: 5 rack: null) for sending state change requests (kafka.controller.RequestSendThread)

[2025-03-14 10:41:20,259] INFO [RequestSendThread controllerId=3] Controller 3 connected to 172.21.216.55:9092 (id: 3 rack: null) for sending state change requests (kafka.controller.RequestSendThread)

14 Mar 2025 15:08:59,786 INFO [SinkRunner-PollingRunner-DefaultSinkProcessor] (.apache.hadoop.hbase.client.AsyncRequestFutureImpl.waitUntilDone:1197) - #6, waiting for 55 actions to finish on table: ads_alarm_detail_808