GorlosHao opened a new pull request, #419:
URL: https://github.com/apache/doris-thirdparty/pull/419
## Summary
Clamp `latencyMillis` before converting it to `int` in
`LatencyPercentile.add()`. This PR changes one line in one file.
## Observed failure
While a newly started FE was catching up, the existing master FE encountered
this exception in the replication feeder. Its BDB environment became invalid
and the FE service was interrupted until restart/recovery.
```text
Caused by: java.lang.ArrayIndexOutOfBoundsException: Index -1871525518 out
of bounds for length 1001
at
java.base/java.util.concurrent.atomic.AtomicLongArray.incrementAndGet(AtomicLongArray.java:234)
at
com.sleepycat.je.utilint.LatencyPercentile.add(LatencyPercentile.java:288)
at
com.sleepycat.je.rep.impl.node.Feeder$InputThread.processHeartbeatResponse(Feeder.java:981)
at
com.sleepycat.je.rep.impl.node.Feeder$InputThread.runResponseLoop(Feeder.java:856)
at com.sleepycat.je.rep.impl.node.Feeder$InputThread.run(Feeder.java:755)
```
The deployment used Doris 4.1.3-rc02 and Java 17.0.2; the JE runtime
reported version 18.3.12. The same narrowing-before-clamping code is present in
the current `bdbje` branch. The exact latency input was not logged.
## Root cause and fix
`Math.min((int) latencyMillis, maxTrackedLatencyMillis)` narrows a positive
`long` before applying the upper bound. Large values can produce a negative
histogram index or wrap into an incorrect lower bucket. For example,
`2423441778L` narrows to the negative index shown above.
Using `(int) Math.min(latencyMillis, maxTrackedLatencyMillis)` applies the
cap first and preserves the intended overflow-bucket behavior. Negative latency
handling is unchanged.
## Validation
Local isolated JUnit 4.12 tests compiled the actual upstream class and its
supporting classes with JDK 26 (`--release 8`). Four tests covered large
positive inputs, ignored negative inputs, normal in-range inputs, and a zero
maximum.
- Before: 4 tests run, 2 failed with `ArrayIndexOutOfBoundsException`.
- After: all 4 tests passed.
The test file is not included in this PR. The full Maven build and a live
Doris cluster reproduction were not run.
## Related fix
StarRocks addressed the same overflow failure in
[StarRocks/starrocks-bdb-je#12](https://github.com/StarRocks/starrocks-bdb-je/pull/12).
This change caps the original `long` value before conversion so oversized
samples remain in the maximum bucket.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
