Start with five questions
Use the same questions during normal operation and after an incident. Record IDs and destination references make the answer easier to verify than a dashboard count alone.
- Did the expected run start and finish?
- Was the source record read, or is it outside the import boundary?
- Was it deliberately skipped by a filter?
- Was a destination write confirmed?
- Is there an unresolved issue, uncertain outcome or retry still to review?
Distinguish the outcomes before acting
A skipped record may be behaving exactly as configured. A missing required field needs a data correction. An expired connection needs restored access. A timeout after sending a write needs a destination check before another write is attempted.
| Observed outcome | Investigate | Next step |
|---|---|---|
| Not read | Import scope, permissions, source updates and schedule | Confirm eligibility before replaying. |
| Skipped | Filter rule and the actual input value | Change the rule only if the record should qualify. |
| Mapping or validation issue | Missing field, value conversion or destination requirement | Correct the cause, then test the affected record. |
| Authentication failure | Expired or revoked access and required scopes | Reconnect and validate the connection. |
| Uncertain write | Whether a matching destination record already exists | Reconcile the outcome before another create attempt. |
How Integrator represents the work
The implemented engine records read, mapping, skip, write and error events for source records. It stores run counts and checkpoints, and keeps source-to-destination identities for recovery. Dry runs show intended writes without creating destination records.
Recognized problems can include a cause and suggested fix from the approved error catalogue. Unknown errors remain unexplained until reviewed. This avoids presenting an invented diagnosis as a fact.
These mechanisms support investigation. They are not an uptime commitment, a guarantee that an issue will fix itself or proof that a beta recipe is ready for every production account.
Close the loop after a retry
First correct the cause and determine whether the destination already received the write. Then use the appropriate retry or replay and check the resulting record. Keep the evidence needed to explain why it is safe to resume.
Document who owns the integration, how often they review unresolved records and when to pause a flow. Use the business impact to set review priorities, rather than treating every failure as equally urgent.
- Keep the source ID, run ID and destination reference together.
- Verify the destination outcome, including duplicate checks.
- Confirm the affected issue is resolved and expected records are accounted for.
- Review the next normal run before considering the incident closed.
