Neo4j Aura incident
AuraDB issue: Errors when running CREATE VECTOR INDEX
Resolved after 2 hours 59 minutes, affecting AuraDB Professional on AWS (*.databases.neo4j.io) and AuraDS Enterprise on AWS (*.databases.neo4j.io) and 1 more, on June 30, 2026.
## What Happened An issue was identified following the deployment of version 2026.06 on 2026-06-30 at 08:00 UTC, where running `CREATE VECTOR INDEX` on upgraded Aura instances triggered a system exception. This failure in index generation occurred because the underlying store and kernel required additional updates to support the operation This problem only impacted the creation of new indexes; existing vector indexes are not affected. Any vector index created before 09:00 on 2026-06-30 was unaffected and continued to work normally. ## How the service was affected Following the rollout of version 2026.06, execution of the CREATE VECTOR INDEX statement failed, impacting applications that depended on adding new vector indexes. Affected clients received the error: _Creating a vector index with provided settings is not supported in V2026\_02. The required version for operation is V2026\_06. Please upgrade DBMS_. Existing indexes \(created prior to this incident\) and other queries remained entirely unaffected and fully functional. ## What are we doing now Neo4j teams identified the root cause as a bug in the Cypher planner and resolved the issue with a code fix. This fix has been deployed to all affected Aura instances, fully resolving the error To prevent future occurrences, we are enhancing our monitoring and alerting systems to detect similar issues earlier, particularly within staging environments, before rollouts occur. Simultaneously, we are refining our deployment processes to minimize customer impact and recovery times Additionally, we are investigating options to implement automated, post-upgrade validation tests specifically for critical operations such as vector index creation
Affected components
Update timeline
- identified Jun 30, 2026, 02:45 PM UTC
Running CREATE VECTOR INDEX fails with error 51N31 error: system configuration or operation exception - not supported. Creating a vector index with provided settings is not supported in V2026_02. Required version for operation is V2026_06. Please upgrade DBMS.. This only affects new index, existing VECTOR INDEX are unaffected. Any vector index created prior to 2026-06-30 at 09:00 continue to operate normally.
- identified Jun 30, 2026, 03:41 PM UTC
We are starting now to deploy a fix to address this issue and will provide an update soon on progress.
- monitoring Jun 30, 2026, 04:05 PM UTC
A fix is rolling out and progressing well. We will update with progress.
- resolved Jun 30, 2026, 05:45 PM UTC
The fix has been fully deployed to all impacted instances. This incident is now resolved.
- postmortem Jul 20, 2026, 08:40 AM UTC
## What Happened An issue was identified following the deployment of version 2026.06 on 2026-06-30 at 08:00 UTC, where running `CREATE VECTOR INDEX` on upgraded Aura instances triggered a system exception. This failure in index generation occurred because the underlying store and kernel required additional updates to support the operation This problem only impacted the creation of new indexes; existing vector indexes are not affected. Any vector index created before 09:00 on 2026-06-30 was unaffected and continued to work normally. ## How the service was affected Following the rollout of version 2026.06, execution of the CREATE VECTOR INDEX statement failed, impacting applications that depended on adding new vector indexes. Affected clients received the error: _Creating a vector index with provided settings is not supported in V2026\_02. The required version for operation is V2026\_06. Please upgrade DBMS_. Existing indexes \(created prior to this incident\) and other queries remained entirely unaffected and fully functional. ## What are we doing now Neo4j teams identified the root cause as a bug in the Cypher planner and resolved the issue with a code fix. This fix has been deployed to all affected Aura instances, fully resolving the error To prevent future occurrences, we are enhancing our monitoring and alerting systems to detect similar issues earlier, particularly within staging environments, before rollouts occur. Simultaneously, we are refining our deployment processes to minimize customer impact and recovery times Additionally, we are investigating options to implement automated, post-upgrade validation tests specifically for critical operations such as vector index creation
Your product runs on more than Neo4j Aura
Add the vendors you depend on and get one message when any of them reports trouble. Free for 3 vendors, no card.
Amazon Web Services