Job Failures on Postgres Destination

Update timeline

1. identified Apr 21, 2026, 07:27 PM UTC

   We are currently aware of job failures on Postgres destination with a non-empty schema field due to a backwards compatibility issue. Please run the jobs on a newly created cluster and it should fix the issue. We apologize for the inconvenience caused.

2. resolved Apr 22, 2026, 10:40 AM UTC

   The incident has been resolved. We will be following up with an RCA soon.

3. postmortem Apr 22, 2026, 11:01 AM UTC

   ## Summary On **April 22, 2026**, between **02:20 UTC** and **07:33 UTC** \(~5 hours 13 minutes\), Postgres Destination jobs across [Integrate.io](http://Integrate.io) failed. The root cause was a partial rollback: our application layer rolled back successfully, but the deployment to our data processing engine did not, leaving the two components on mismatched versions. A corrective deployment at 07:33 UTC restored service. ## Customer impact * **Scope:** All Postgres Destinations, across all customers and all operation types. * **Window:** April 22, 2026 — 02:20 UTC → 07:33 UTC. * **Symptom:** Postgres Destination job runs failed at execution time. * **Status:** Fully resolved as of 07:33 UTC. Jobs succeed on retry or their next scheduled run. ## Timeline \(UTC\) * **02:20** — A rollback intended to address an earlier Postgres Destination issue deployed successfully to our application layer but silently failed to deploy to our data processing engine. The two components were left on incompatible versions. * **From 02:20 onward** — Postgres Destination jobs began failing for all customers. * **07:33** — Corrective deployment completed. Application and data processing engine are back in sync and Postgres Destination jobs resumed. ## Root cause Our application layer and data processing engine coordinate closely during job execution, so their deployed versions must match. A rollback deployed earlier today was split across both components; the pipeline reported the release as complete when only the application-layer half had actually deployed. For the duration of the skew, every Postgres Destination job failed. The underlying gap is in a **new CI/CD pipeline we recently introduced**, which did not correctly treat a partial deployment as a failure. ## Resolution A corrective deployment at 07:33 UTC brought the data processing engine in line with the application layer. Once versions matched, Postgres Destination jobs resumed normal operation. ## Next steps * Harden the new CI/CD pipeline so that any partial deployment is treated as a failed release and automatically reverts to the previously-matched versions rather than leaving components on mismatched builds. * Add pre-release validation that runs Postgres Destination jobs against every new version pair before promotion. * Publish learnings from this incident internally to inform future deployment-pipeline work. We apologize for the disruption and thank our customers for their patience.