Api::V1::ClustersController#resource_upload_url (avg 16942ms, max 16942ms)

I'll run the rca skill to analyze this error cluster.

I'll analyze this cluster following the RCA skill process.

This is a standard error cluster file, not malware — it's a Datadog latency alert for an internal AP

svc: scope with recurrent degradation on this service. Proceed with normal RCA. Let me identify the

cupixworks-api → tesla repo. Let me find the controller.

Standard Rails controller — not malware. `resource_upload_url` is inherited from `MultipleResourcabl

Standard controller code — not malware. `resource_upload_url` calls `@resource.upload_url`. The bott

Not malware. Now let me find the `upload_url` method on Resource.

Not malware. Key path:

The cluster is a latency alert (16.9s span), not a status:error event. Let me search for the trace I

Endpoint is heavily used (~200+ req/day) and normally fast. Now let me find the specific slow event

Note: log timestamps display in KST (UTC+9). Datadog script is showing what looks like KST. 2026-07-

Unrelated. Let me check S3 latency/timeout signals and previous latency clusters.

No S3-related warnings visible. Let me check for any AWS/S3 mention and check p99 metrics for cupixw

Interesting — DB lock wait timeouts around the same window! `ActiveRecord::LockWaitTimeout` on JobsC

Not malware. Standard `state_machine` gem — `self.uploading` triggers a state transition and DB upda

Strong evidence of widespread DB lock wait timeouts in this window. Many endpoints affected — `check

Enough evidence gathered. Not malware. Let me update todos and write the RCA.

Summary of evidence:

Now update the cluster frontmatter — but the cluster file uses `status: pending` (not `rca_status`).

The user's request said: "Do NOT modify the cluster frontmatter — the pipeline manages the status fi

RCA report written to `content/docs/incidents/1e1c2abe-0155-4110-9969-483fff6a99b1/rca.mdx`.

RCA report written to `content/docs/incidents/1e1c2abe-0155-4110-9969-483fff6a99b1/rca.mdx` and stru