mongo-ai/posthog-pipeline-diagnosis

Data pipeline / CDP function diagnosis workflow

Before you diagnose: fetch live docs

1. Run docs-search("CDP pipeline troubleshooting {specific symptom}")
2. If potential known bug: run gh search issues --repo PostHog/posthog "CDP {symptom}"
3. Use the fetched docs as your primary reference — NOT the steps below

The steps below are a diagnostic FRAMEWORK (what questions to ask, in what order). The docs responses are the CONTENT (current answers, version requirements, config details).

Goal

Determine whether the pipeline problem is caused by:

• function disabled or misconfigured
• filter/trigger conditions not matching events
• transformation logic error
• destination authentication failure
• destination rate limiting or payload rejection
• execution timeout or resource limits
• known product bug

Step 1: Identify the function

Use CDP function tools to find the relevant function:

• list functions to find by name or destination type
• retrieve function details including config, filters, and status

Step 2: Check function status and recent invocations

Inspect:

• enabled/disabled state
• recent execution logs
• error rate and error messages
• last successful execution timestamp

Use: cdp-functions-list, cdp-functions-retrieve
Use: logs-query (filter for the function ID)

Step 3: Check trigger and filter conditions

If events are reaching PostHog but not the destination:

• verify the function trigger matches the event name
• verify filter conditions match event properties
• verify the function is attached to the correct project

Step 4: Check execution logs for errors

Query logs for the function:

• look for authentication errors (401, 403)
• look for payload validation errors (400, 422)
• look for rate limiting (429)
• look for timeout errors
• look for transformation/code errors

Step 5: Check destination-side issues

If the function executes but the destination doesn't receive data:

• verify destination credentials are current
• verify the destination endpoint URL is correct
• verify payload format matches destination expectations
• check whether the destination has its own ingestion delays

Step 6: Check for ingestion-side issues

If the problem is events not reaching PostHog at all:

• defer to posthog-event-diagnosis skill
• the pipeline can only process what PostHog receives

Step 7: Search known issues

Search GitHub with at least 2 variants:

• gh search issues --repo PostHog/posthog "CDP {destination name} {symptom}"
• gh search issues --repo PostHog/posthog "pipeline {symptom}"

Common resolution patterns

• Function not firing -> disabled, filter mismatch, or trigger condition wrong
• Function firing but destination empty -> auth failure, payload format, or destination-side issue
• Intermittent failures -> rate limiting or transient destination errors
• Transformation errors -> code bug in the Hog function logic
• Events missing entirely -> ingestion issue, not pipeline issue (use event diagnosis)