Api::V1::VideosController#create (avg 1563ms, max 4144ms)

Now I'll invoke the RCA skill to perform the investigation.

This is a latency cluster for `Api::V1::VideosController#create` with avg 1268ms and max 4144ms acro

Now I know the repository is `tesla` at `$REPOS_DIR/tesla`. Let me search Datadog logs and explore t

I can see the key data. The logs show `duration` field values. Let me look at more details and under

Interesting pattern: `duration` ranges from 518ms to 1513ms but `db` time is only 21-103ms, and `vie

Now I can see the `create` method in `ApiController` just renders the model. The key flow is: `Video

Now let me look at the key model concerns — especially `Resourcable::Video` since the logs show `upl

Now let me look at the `Resourcable` concern for Video (the module included in the model), and the `

Now I see two key things:

Now I have the key insight. Let me check the `upload_url` method that's likely called during seriali

Now let me see the full `presigned_upload_url` and `Cupix::StorageService.client` to understand S3 c

Now let me look at the VideoSerializer to see what fields trigger the `upload_url` generation, since

Now I see the full picture. The `upload_url` attribute in the serializer calls `video.resource_uploa

Now let me look at the `set_capture` before_action and the `CaptureRepository.show` call to understa

Now let me look at what the `factory_instance` method does, and check if there's any S3 presigned UR

Now I have a clear picture of the execution flow. Let me also check if there's a `Resourcable` base

Now let me check Datadog APM traces for more detail on what's consuming the time in these requests.

Now let me understand the full timing breakdown. The key observation is:

Let me check the `Eventable::Video` concern and the `Parameter::Video` set_parameters call — these c

Now I have enough evidence. Let me check one more thing - how much time the STS `get_federation_toke

Clear evidence: there's a **970-2335ms gap** between the DB time and total duration across all 4 reg

The ES query duration values are ~0.07-0.12 seconds (70-120ms). That's significant but doesn't fully

Now I'll invoke the RCA skill to perform the analysis.

This is a latency cluster for `Api::V1::VideosController#create` with avg 1268ms, max 4144ms across

The key finding is that DB time is only ~48ms but total duration is 1268-4144ms, meaning 95-99% of t

Now I can see the critical path. The `presigned_upload_url` method at line 111-131 creates an S3 pre

Now I have a clear picture. The key finding is:

Now I have all the evidence needed. Let me create the incident directory and write the RCA report.

Now I need to update the cluster frontmatter to mark the RCA as completed. The user explicitly said

RCA report written to `./content/docs/incidents/bfeae7b6-2a07-4498-8ed8-e427f853c6c8/rca.mdx`.

RCA complete. The report is at `./content/docs/incidents/bfeae7b6-2a07-4498-8ed8-e427f853c6c8/rca.md