🛡️ Sentinel: Fix unbounded retries on client errors#191
Conversation
🚨 Severity: MEDIUM 💡 Vulnerability: The retry logic blindly retried all `httpx.HTTPError` exceptions, including 400 (Bad Request) and 401/403 (Auth failures). This causes API spamming, potential account lockouts, and delays in error reporting. 🎯 Impact: Inefficient API usage, potential triggering of rate limits or account blocks, and delayed error feedback to users. 🔧 Fix: Modified `_retry_request` to check for `httpx.HTTPStatusError` and re-raise immediately if the status code is a client error (400-499) and not 429 (Too Many Requests). ✅ Verification: Verified with a reproduction script that call count reduced from 3 (retries) to 1 (fail fast) for 400 errors. Ran regression tests. Co-authored-by: abhimehro <84992105+abhimehro@users.noreply.github.com>
|
👋 Jules, reporting for duty! I'm here to lend a hand with this pull request. When you start a review, I'll add a 👀 emoji to each comment to let you know I've read it. I'll focus on feedback directed at me and will do my best to stay out of conversations between you and other bots or reviewers to keep the noise down. I'll push a commit with your requested changes shortly after. Please note there might be a delay between these steps, but rest assured I'm on the job! For more direct control, you can switch me to Reactive Mode. When this mode is on, I will only act on comments where you specifically mention me with New to Jules? Learn more at jules.google/docs. For security, I will only act on instructions from the user who triggered this task. |
|
Merging to
|
|
👋 Development Partner is reviewing this PR. Will provide feedback shortly. |
Summary of ChangesHello @abhimehro, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! This pull request addresses a potential denial-of-service (DoS) risk by refining the application's retry mechanism. It ensures that client-side errors (4xx HTTP status codes, excluding 429) are no longer subject to unbounded retries, thereby preventing unnecessary API spamming, potential account lockouts, and improving the efficiency of error handling. The change also includes documentation of this security enhancement in the project's journal. Highlights
Changelog
Activity
Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request prevents unbounded retries on 4xx client errors, enhancing security and stability, and adds a Sentinel Journal entry. However, the security review could not be completed as the content of main.py was inaccessible, which prevented a full SAST Reconnaissance Pass. Please ensure main.py is accessible for a complete security assessment. Additionally, consider refactoring the new error handling logic in main.py to improve maintainability by removing code duplication.
There was a problem hiding this comment.
Pull request overview
This PR hardens the HTTP retry behavior in main.py by failing fast on non-retriable client errors (4xx) while still allowing retries for rate limiting (429), and documents the lesson learned in the Sentinel journal.
Changes:
- Update
_retry_request()to immediately re-raisehttpx.HTTPStatusErrorfor non-429 4xx responses. - Add a
.jules/sentinel.mdjournal entry documenting the retry pitfall and prevention guidance.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.
| File | Description |
|---|---|
| main.py | Prevents retries on non-429 4xx HTTP status errors inside _retry_request() to avoid unnecessary repeated requests. |
| .jules/sentinel.md | Adds a Sentinel journal entry capturing the retry/HTTPStatusError learning and prevention guidance. |
| code = e.response.status_code | ||
| if 400 <= code < 500 and code != 429: | ||
| if hasattr(e, "response") and e.response is not None: | ||
| log.debug( | ||
| f"Response content: {sanitize_for_log(e.response.text)}" | ||
| ) |
There was a problem hiding this comment.
In the HTTPStatusError branch, e.response.status_code is accessed before the hasattr(e, "response")/e.response is not None guard. If you want this guard, it needs to wrap the status_code access too; otherwise the guard is misleading. Consider either removing the guard entirely (and rely on HTTPStatusError always having a response) or moving it before reading status_code.
| code = e.response.status_code | |
| if 400 <= code < 500 and code != 429: | |
| if hasattr(e, "response") and e.response is not None: | |
| log.debug( | |
| f"Response content: {sanitize_for_log(e.response.text)}" | |
| ) | |
| # Ensure the response exists before accessing status_code or text. | |
| # While httpx.HTTPStatusError is expected to always have a response, | |
| # this guard keeps the code robust against library or usage changes. | |
| if not hasattr(e, "response") or e.response is None: | |
| raise | |
| code = e.response.status_code | |
| if 400 <= code < 500 and code != 429: | |
| log.debug( | |
| f"Response content: {sanitize_for_log(e.response.text)}" | |
| ) |
| # Security Enhancement: Do not retry client errors (4xx) except 429 (Too Many Requests). | ||
| # Retrying 4xx errors is inefficient and can trigger security alerts or rate limits. | ||
| if isinstance(e, httpx.HTTPStatusError): | ||
| code = e.response.status_code | ||
| if 400 <= code < 500 and code != 429: | ||
| if hasattr(e, "response") and e.response is not None: |
There was a problem hiding this comment.
This change introduces new behavior (fail-fast on non-429 4xx). There are existing tests for _retry_request logging/redaction, but none that assert we don’t retry on 4xx (e.g., mock_func.call_count == 1 with max_retries>1 for a 401/400) and that we still retry on 429. Adding a focused test would prevent regressions in retry semantics.
| ## 2026-10-24 - Unbounded Retries on Client Errors (DoS Risk) | ||
|
|
||
| **Vulnerability:** The retry logic blindly retried all `httpx.HTTPError` exceptions, including 400 (Bad Request) and 401/403 (Auth failures). This causes API spamming, potential account lockouts, and delays in error reporting. | ||
|
|
||
| **Learning:** `httpx.HTTPStatusError` (raised by `raise_for_status()`) inherits from `httpx.HTTPError`. Generic `except httpx.HTTPError:` blocks will catch it and retry client errors unless explicitly handled. | ||
|
|
||
| **Prevention:** Inside retry loops, catch `httpx.HTTPStatusError` first. Check `response.status_code`. If `400 <= code < 500` (and not `429`), re-raise immediately. |
There was a problem hiding this comment.
The heading/description calls this "Unbounded Retries", but _retry_request is bounded by MAX_RETRIES (currently 10). To avoid misleading future readers, consider rewording to "Retries on client errors (4xx)" (or similar) and keep the DoS discussion focused on unnecessary repeated requests rather than "unbounded" behavior.
| ## 2026-10-24 - Unbounded Retries on Client Errors (DoS Risk) | |
| **Vulnerability:** The retry logic blindly retried all `httpx.HTTPError` exceptions, including 400 (Bad Request) and 401/403 (Auth failures). This causes API spamming, potential account lockouts, and delays in error reporting. | |
| **Learning:** `httpx.HTTPStatusError` (raised by `raise_for_status()`) inherits from `httpx.HTTPError`. Generic `except httpx.HTTPError:` blocks will catch it and retry client errors unless explicitly handled. | |
| **Prevention:** Inside retry loops, catch `httpx.HTTPStatusError` first. Check `response.status_code`. If `400 <= code < 500` (and not `429`), re-raise immediately. | |
| ## 2026-10-24 - Retries on Client Errors (4xx) Causing DoS Risk | |
| **Vulnerability:** The retry logic blindly retried all `httpx.HTTPError` exceptions, including 400 (Bad Request) and 401/403 (Auth failures). Even though retries are bounded by `MAX_RETRIES`, this still causes unnecessary repeated requests, API spamming, potential account lockouts, and delays in surfacing real errors. | |
| **Learning:** `httpx.HTTPStatusError` (raised by `raise_for_status()`) inherits from `httpx.HTTPError`. Generic `except httpx.HTTPError:` blocks will catch it and retry client errors unless explicitly handled. | |
| **Prevention:** Inside retry loops, catch `httpx.HTTPStatusError` first. Check `response.status_code`. If `400 <= code < 500` (and not `429`), re-raise immediately so client errors fail fast instead of being retried. |
main.pyto prevent retrying on 4xx errors (except 429).PR created automatically by Jules for task 7567691608950293777 started by @abhimehro