feat(preprod): Reduce status check API calls #107886
Conversation
Without rules configured, both calls compute NEUTRAL status, so the skip logic deduplicates the second call. Configure rules so the first call posts IN_PROGRESS and the second posts SUCCESS (different statuses).
github-actions
bot
added
the
Scope: Backend
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
…enum Adds StatusCheckPostPolicy enum (TRY_TO_DEDUPLICATE, ALWAYS_POST) to control dedup behavior instead of checking caller == "rerun_endpoint". Defaults to TRY_TO_DEDUPLICATE so existing callers need no changes.
| ).values_list("id", "extras") | ||
| ) | ||
| if _should_skip_status_check( | ||
| all_artifacts, extras_by_artifact_id, size_metrics_map, status |
There was a problem hiding this comment.
I'm confused why we have all_artifacts already but still need to do another query for extras_by_artifact_id
| return False | ||
|
|
||
|
|
||
| def _write_posted_status_claim(preprod_artifact: PreprodArtifact, posted_status: str) -> None: |
There was a problem hiding this comment.
Don't we already have another method that updates the posted_status_checks field? Shouldn't that be consolidated together? Otherwise it's a bit confusing we have multiple ways to update that.
| artifact.save(update_fields=["extras"]) | ||
|
|
||
|
|
||
| def _clear_posted_status_claim(preprod_artifact: PreprodArtifact) -> None: |
There was a problem hiding this comment.
It might be fine to keep this duplication but perhaps _clear_posted_status_claim and _write_posted_status_claim can be combined?
There was a problem hiding this comment.
I think some of the message formatting needs to be updated to work with this properly. For example, now that we will no longer render separate in-progress/success statuses per app, we should probably have a single generic "Your apps are being analyzed" type of messaging.
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.
| include_approve_action=bool(triggered_rules), | ||
| ) | ||
| except Exception as e: | ||
| _clear_posted_status_claim(preprod_artifact) |
There was a problem hiding this comment.
Claim cleared unconditionally regardless of post policy
Low Severity
_clear_posted_status_claim is called unconditionally on API failure, but _write_posted_status_claim is only called inside the TRY_TO_DEDUPLICATE block. When post_policy is ALWAYS_POST (rerun endpoint), no claim was written, yet the clear still runs — removing a valid posted_status set by a previous successful run. This performs an unnecessary DB transaction and briefly leaves extras in an inconsistent state between the clear and the subsequent _update_posted_status_check call, during which concurrent sibling tasks could read stale dedup state.
Additional Locations (1)
1 participant
Summary
Key changes
_should_skip_status_check: pure function deciding skip/post based on sibling extras_write_posted_status_claim/_clear_posted_status_claim: claim management in extrascreate_preprod_status_check_taskwith two-phase approach_update_posted_status_checknow persistsposted_statuson successTest plan