Skip to content

Add the data connector health endpoint to the Preview spec - #631

Open
zilleeizad-inter wants to merge 4 commits into
mainfrom
zille.eizad/dc-health-openapi
Open

zilleeizad-inter wants to merge 4 commits into
mainfrom
zille.eizad/dc-health-openapi

Conversation

@zilleeizad-inter

Copy link
Copy Markdown
Contributor

Why?

Developers using data connectors can fetch execution results one at a time, but nothing tells them how a connector is doing overall — how often it succeeds, how slow it is, and what its failures actually are. This adds the Preview spec for the endpoint that answers that.

How?

Adds the path and its response schemas to the Preview description, keeping the failure-reason values aligned with the vocabulary the execution results endpoint already filters on.

Notes for reviewers

Draft until the endpoint it describes is available, so the spec does not land ahead of the API.

Generated with Claude Code

zilleeizad-inter and others added 3 commits August 21, 2026 12:27
GET /data_connectors/{data_connector_id}/health returns aggregated execution
health for a data connector over a requested time range: success rate, latency
and payload size percentiles, breakdowns of HTTP statuses and failure reasons,
and a time series of intervals.
Six details a developer would otherwise have to discover by calling the
endpoint:

- The default `bucket_size_minutes` divides the range into 300 intervals
  or fewer, rather than "a width chosen to suit the range".
- Interval boundaries align to multiples of the width from the Unix epoch,
  not to `start_ts`, so the first interval may begin before `start_ts`.
- A time-series latency field reports 0 when nothing in that interval
  carried the measurement, which is indistinguishable from a true zero.
- An execution counts as successful only when it completed AND returned a
  2xx status; one with no recorded status counts as a failure and appears
  under `unknown`.
- `http_status_distribution` omits executions with no recorded status, so
  its counts may add up to less than `total_executions`.
- Responses are cached for up to a minute.

The 200 example's `bucket_start` moves to an epoch-aligned value so it
illustrates the alignment rule rather than contradicting it.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@zilleeizad-inter
zilleeizad-inter marked this pull request as ready for review November 3, 2026 15:04
…-openapi

# Conflicts:
#	descriptions/0/api.intercom.io.yaml

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant