prince286 opened a new pull request, #21349:
URL: https://github.com/apache/kafka/pull/21349

   This Pull Request implements two key fault tolerance features within 
MirrorMaker's `MirrorSourceTask` to enhance data integrity in mission-critical 
deployments.
   
   The changes address the risk of silent data loss (truncation) and service 
disruption (topic reset).
   
   ### 1. Log Truncation Detection (Fail-Fast)
   
   **Problem:** Kafka's retention policies can delete log segments, causing 
MM2's consumer to lose its expected offset and silently skip data.
   **Solution (in MirrorSourceTask.java):**
   An in-memory map tracks the last processed offset for each partition. If the 
newly polled offset is greater than the last processed offset + 1, a data gap 
is confirmed.
   - **Action:** Throws a `ConnectException` with a **CRITICAL** log message, 
forcing a Fail-Fast stop and preventing further silent data loss.
   
   ### 2. Graceful Topic Reset Handling (Auto-Recovery)
   
   **Problem:** Deleting and recreating the source topic resets its offset to 
0. MM2's stored offset is now invalid.
   **Solution (in MirrorSourceTask.java):**
   The logic detects if the newly polled offset is less than the last processed 
offset (indicating a log reset/topic recreation).
   - **Action:** Logs an **AUTO-RECOVERED** event but does **not** throw an 
exception. The consumer automatically resets to the earliest offset (0) based 
on Connect's default behavior, allowing replication to resume immediately.
   
   ### 3. Environment Fix (MirrorUtils.java)
   
   **Problem:** Single-node Connect deployments fail due to Kafka 4.0.0's 
strict default replication factor of 3 for internal topics.
   **Solution (in MirrorUtils.java):**
   The `createCompactedTopic` method is modified to explicitly set 
`replicationFactor = 1`, ensuring compatibility with single-broker testing 
environments.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to