HDDS-16379. Recon returns 500 NPE instead of proxying error when metrics endpoint responds with a non-enum status code

[henry3260]

What changes were proposed in this pull request?

Recon's metrics proxy and metrics service providers checked the upstream HTTP status with
Response.Status.fromStatusCode(code).getFamily(). Response.Status is a JAX-RS enum that
only lists common status codes, and fromStatusCode() returns null for any code not in the
enum (e.g. 422, which the Prometheus HTTP API returns when a query expression cannot be
executed). Calling .getFamily() on that null throws a NullPointerException, so Recon
responds 500 with a stack trace instead of setting 502 and proxying the Prometheus error body
back as intended.

This PR:

• Replaces the check with Response.Status.Family.familyOf(code), which classifies any status
  code by its first digit and never returns null, in MetricsProxyEndpoint,
  PrometheusServiceProviderImpl, and JmxServiceProviderImpl.
• Guards against HttpURLConnection#getErrorStream() returning null (per its javadoc) in the
  error path of MetricsProxyEndpoint, which previously caused a second NPE in
  Channels.newChannel(null) and in the finally close.

What is the link to the Apache JIRA

https://issues.apache.org/jira/browse/HDDS-16379

How was this patch tested?

Added two unit tests to the existing TestEndpoints suite, following the mock pattern of
testGetMetricsResponse:

• testGetMetricsResponseWithNonEnumStatusCode: upstream returns 422 with an error body;
  verifies Recon sets 502 and proxies the body back (the path that previously threw NPE).
• testGetMetricsResponseWithNullErrorStream: upstream returns an error status with no error
  body; verifies no exception and a 502 response.

[Copilot]

🟢 Approval recommended

The change is narrowly scoped to robust HTTP status/error-stream handling and is covered by targeted regression unit tests for the reported failure modes.

Pull request overview

This PR fixes an error-handling bug in Recon’s metrics proxy/service providers where certain upstream HTTP status codes (eg, 422 from the Prometheus API) could trigger a NullPointerException, causing Recon to return a 500 instead of a 502 while proxying the upstream error body.

Changes:

• Replace Response.Status.fromStatusCode(code).getFamily() with Response.Status.Family.familyOf(code) to correctly classify non-enum HTTP status codes without returning null.
• Add a null-guard for HttpURLConnection#getErrorStream() in MetricsProxyEndpoint to avoid NPEs when the upstream returns an error with no body.
• Add unit tests in TestEndpoints covering both non-enum status codes with an error body and error responses with a null error stream.

File summaries

File	Description
hadoop-ozone/recon/src/main/java/org/apache/hadoop/ozone/recon/api/MetricsProxyEndpoint.java	Makes status-family checks null-safe for non-enum codes and avoids NPEs when error bodies are missing.
hadoop-ozone/recon/src/main/java/org/apache/hadoop/ozone/recon/spi/impl/PrometheusServiceProviderImpl.java	Uses status-family classification that works for any valid HTTP status code.
hadoop-ozone/recon/src/main/java/org/apache/hadoop/ozone/recon/spi/impl/JmxServiceProviderImpl.java	Uses status-family classification that works for any valid HTTP status code.
hadoop-ozone/recon/src/test/java/org/apache/hadoop/ozone/recon/api/TestEndpoints.java	Adds regression tests for the prior NPE path and for null error bodies.

Review details

• Files reviewed: 4/4 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[chihsuan]

Thanks for the patch! +1 LGTM