Troubleshoot data imports
Diagnose authorization failures, missing records, incomplete pagination and incremental filtering.
Use these checks when the connector fails during extraction.
Connection passes verification but extraction is unauthorized
The verification endpoint may not exercise the scopes needed by later API calls. Test discovery, metadata, records and attachment endpoints using the same connection account. Replace unauthenticated placeholder verification URLs. Inspect insufficient scopes, revoked access and differences between organization and project permissions. See authentication.
Fixture completes but records are wrong or missing
Inspect the current artifacts under code/extracted_files/. Verify the selected fixture, external sync unit, extraction scope and resolved time window. Fixture execution can use a real external client and different account permissions from production. Check artifact counts, normalized IDs and metadata types rather than treating the runner's completion message as semantic validation.
Only the first API page is imported
Implement pagination in the connector's own external client/worker. Persist the next cursor after durable progress and resume when the runtime requests continuation. Do not return success while pages remain. The template's in-memory Todo pagination is illustrative. See data extraction.
Incremental sync misses records or reimports everything
Read the SDK-resolved extract_from and extract_to from event context. Verify external timestamp semantics, timezone handling and inclusive/exclusive boundaries. Preserve connector-owned cursors separately from SDK bookkeeping. Ensure an empty window does not mean an unconditional full import. See runtime and SDK migration.
Related: troubleshooting index.
Last updated on