We need urgent human help with an AWS Aurora PostgreSQL database provisioned through Vercel Marketplace. The database connection is failing, and the account access provided through the integration also blocks diagnosis and recovery.
Our existing data is effectively trapped: we cannot connect to the database, complete a recovery operation, or export the data to migrate elsewhere.
1. The current AWS connection failure
Both our production and staging Aurora PostgreSQL clusters in ap-northeast-2 are affected. They use the Express configuration with Internet Access Gateway and IAM authentication.
- First observed production failure: October 11, 2026 at 14:25 UTC.
- Latest independent check: October 11 at 19:15 UTC. Both environments still fail.
- TCP connects successfully, and PostgreSQL SSLRequest receives an S response.
- The TLS handshake then ends with an unexpected EOF after approximately 10 seconds, before database authentication or any query.
- We reproduced this with Python/OpenSSL/psql from a local machine, and independently from AWS CloudShell and the Vercel runtime. This is not confined to our application or ORM.
- Rebooting the staging writer did not resolve it. The console reports the databases as available.
The root cause is still unknown. We are not claiming that this is a region-wide AWS outage.
2. Diagnosis and recovery are also blocked
The Vercel-provided console role cannot perform the operations we need:
- Engine-log download is denied for rds:DownloadDBLogFilePortion.
- An attempted temporary point-in-time restore is denied for rds:AddTagsToResource. Both the console and CLI attempts failed; explicitly supplying no tags did not resolve it.
- No temporary restore resource was created, and we have not obtained a usable restored database from which to export our data.
We therefore cannot currently recover our data through the available restore path, and the original database connection failure prevents a normal dump/export. Moving to another provider does not solve that access problem: we still need to retrieve the existing data first.
3. The support dead end — what is and is not registered
A private Vercel support case has been submitted and is awaiting a human response. An AWS technical support case has NOT been opened.
The AWS account reached through the Marketplace integration has Basic Support, and its console blocks technical case creation. The integration's Provider Support link leads back to that same AWS support console.
The Vercel Marketplace case category initially blocked submission as well. We eventually submitted the private Vercel case under Account Settings, following the Vercel support assistant's instructions, while explicitly disclosing that this concerns the AWS Marketplace integration and its account permissions. The diagnostic evidence and resource details are already in that private case.
We understand that AWS owns Aurora troubleshooting. We need Vercel and the responsible AWS integration team to provide a working escalation and recovery path for the account provisioned through this integration. Advice to use permissions or support access that this account does not provide leaves us stuck.
4. The application is back, but our data has not been recovered
To get basic service operation running again, we switched production to another database provider and initialized a fresh database with seed data only.
None of the historical Aurora data has been recovered or migrated. The application running against a new database must not be mistaken for resolution of the original database-access and recovery incident.
5. Why this is especially concerning before launch
We are immediately before our service launch. This has been one of the worst infrastructure experiences we have had. If it had happened at launch with customers relying on us, the consequences would have been catastrophic.
We previously had to move away from another Marketplace-provisioned datastore after a Prisma Postgres SSO integration issue, documented here: https://community.vercel.com/t/unexpected-server-error-during-vercel-sso-oauth-integration-for-prisma-postgres/48214#post_3
That is background only; this post is about the current AWS Aurora incident. The technical causes may differ, but we are again changing our production datastore after a Marketplace integration problem. This time we cannot even access the old database to migrate the data.
What we need now
- A human owner who can coordinate with AWS and investigate the Aurora connection failure.
- A supported way to access and export the existing data, or a provider-assisted restore and export if the managed account cannot perform the necessary recovery operations.
- Preservation of the original databases and available backups/snapshots while recovery is investigated.
- A concrete next action, the responsible team, and a timeline for the next update.
- A clear explanation of how Marketplace customers can obtain technical support and recover their data when both database access and the supplied recovery/support routes are blocked.
We can provide account identifiers, endpoints, and detailed evidence privately. The immediate priority is retrieving our existing data safely. Please help us reach the team that can actually unblock this.