oscerd opened a new pull request, #25089:
URL: https://github.com/apache/camel/pull/25089
## Problem
Every AWS producer health check catches `AwsServiceException` and records two
details — the HTTP status code and the AWS error code. The status code is
guarded for null; `awsErrorDetails()` is dereferenced directly:
```java
if (ObjectHelper.isNotEmpty(e.statusCode())) { // guarded
builder.detail(SERVICE_STATUS_CODE, e.statusCode());
}
if (ObjectHelper.isNotEmpty(e.awsErrorDetails().errorCode())) { //
dereferenced unguarded
builder.detail(SERVICE_ERROR_CODE, e.awsErrorDetails().errorCode());
}
```
`AwsServiceException.awsErrorDetails()` is nullable — it is a builder field
that
defaults to null and is only populated when the SDK successfully unmarshals a
structured error response.
**This is reproduced, not theoretical.** Driving `Athena2ProducerHealthCheck`
with a client that throws
`AwsServiceException.builder().message("boom").statusCode(500).build()`
fails on
`main` with:
```
java.lang.NullPointerException: Cannot invoke
"software.amazon.awssdk.awscore.exception.AwsErrorDetails.errorCode()"
because
the return value of
"software.amazon.awssdk.awscore.exception.AwsServiceException.awsErrorDetails()"
is null
```
So instead of reporting the endpoint `DOWN` with the message it had already
prepared, the health check propagates an NPE — a genuine "service
unreachable"
signal is turned into a health-check failure.
The inconsistency is the tell: the author guarded `statusCode()` one line
above.
## Fix
Guarded `awsErrorDetails()` consistently with the neighbouring `statusCode()`
check, across all **21** producer health checks (athena, comprehend, cw, ddb,
ec2, ecs, eks, eventbridge, lambda, mq, polly, rekognition, step-functions,
textract, timestream query + write, translate, config, parameter-store,
secrets-manager, security-hub):
```java
if (ObjectHelper.isNotEmpty(e.awsErrorDetails()) &&
ObjectHelper.isNotEmpty(e.awsErrorDetails().errorCode())) {
```
`Textract2ProducerHealthCheck` additionally dereferenced it in an
`equals(...)`
test that classifies certain errors as healthy — two unguarded sites — so it
now
reads the error code once into a local:
```java
String errorCode = e.awsErrorDetails() != null ?
e.awsErrorDetails().errorCode() : null;
if ("InvalidParameterException".equals(errorCode) ||
"UnsupportedDocumentException".equals(errorCode)) {
```
## Tests
`Athena2ProducerHealthCheckErrorDetailsTest` drives the health check with a
mocked client and covers both paths:
- **null details** → `DOWN`, `service.status.code` present,
`service.error.code`
absent. Verified to fail on `main` with the NPE quoted above.
- **populated details** → `DOWN` and the error code still reported, so the
fix
does not silently drop the detail it was meant to record.
Full `camel-aws2-athena` suite: 49/49 green. Full reactor build clean.
## Backport
- `camel-4.18.x` — all 21 health checks present → backport opened.
- `camel-4.14.x` — 15 of the 21 present (parameter-store, security-hub,
comprehend, polly, rekognition and textract do not exist on that branch) →
backport covers those 15.
---
_Claude Code on behalf of oscerd_
--
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]
