[GitHub] [kafka] antonio-tomac opened a new pull request #10933: Fix NPE on concurrent RecordHeader.key() access

[GitBox <gi...@apache.org>]

antonio-tomac opened a new pull request #10933:
URL: https://github.com/apache/kafka/pull/10933


   Jira issue (detailed description): https://issues.apache.org/jira/browse/KAFKA-12999
   
   ## Summary
   After upgrading clients to `2.8.0`, reading `ConsumerRecord`'s header keys started resulting in occasional `java.lang.NullPointerException` in case of concurrent access from multiple(2) threads.
   
   ### NPE location
   NPE happens here RecordHeader.java:45:
   ```java
   public String key() {
       if (key == null) {
           key = Utils.utf8(keyBuffer, keyBuffer.remaining()); // NPE here 
           keyBuffer = null;
       }
       return key;
   }
   ```
   
   ### What introduced issue
   Cause of issue is introduced by changes of apache#9223
   
   ### Consequences
   Current implementation renders RecordHeader not thread-safe for read-only access.
   
   ### Reproducibility
   Since issue s concurrent race condition, reproducibility is non-deterministic but easy to reproduce with enough re-attempts
   This PR has test which almost always reproduces issue (wiithout `synchronized`)
   
   ### Fix strategy
   This PR avoids race-condition by having `synchronized` on `key()` and `value()` methods
   
   ### Benchmark
   Comparison between with and without `synchronized` on `RecordHeader.key()` method
   ```
   Benchmark                              Mode  Cnt   Score   Error  Units
   RecordHeaderBenchmark.key              avgt    15  31.308 ± 7.862  ns/op
   RecordHeaderBenchmark.synchronizedKey  avgt    15  31.853 ± 7.096  ns/op
   ````
   
   ### Committer Checklist (excluded from commit message)
   - [ ] Verify design and implementation 
   - [ ] Verify test coverage and CI build status
   - [ ] Verify documentation (including upgrade notes)
   


-- 
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: jira-unsubscribe@kafka.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org