Supabase Outage History

Supabase experienced issues with customer projects getting stuck during PostgreSQL version upgrades, requiring customers to avoid upgrade attempts during the incident. The engineering team identified the root cause and applied a fix, with most affected projects automatically recovering to their previous healthy state. The incident was resolved after 2.9 hours with upgrade operations returning to normal functionality.

A network connectivity issue in Supabase's us-east-1-az4 availability zone caused increased connection times and prevented users from accessing their databases or running authentication-dependent services for a limited number of projects. The incident was caused by an upstream infrastructure provider outage, which led Supabase to temporarily disable project creation, pausing, and compute resizes in the region during recovery efforts. The infrastructure provider resolved the underlying issue after 28 hours, allowing affected projects to recover automatically or through manual restoration, with all services fully restored and normal operations resumed.

Supabase experienced a major incident where PostgREST requests were returning 403 errors across multiple regions, affecting the API Gateway component. The issue was identified and resolved with an exceptions fix that was rolled out and monitored for stability, with the incident lasting 2.1 hours total.

Supabase experienced a major outage lasting 6.4 hours where projects became unavailable across multiple regions due to an upstream EC2 instance issue, initially detected in eu-west-3 but later found to affect other regions as well. The incident impacted the Dashboard, Auth, and Database components, preventing users from accessing their projects. The team resolved the issue by fixing and restarting affected projects, with users also able to perform manual project restarts through the dashboard as a workaround during the incident.

A recent change to Supabase's Postgres systemd service created a dependency on a prestart script that was missing from older projects, causing Postgres to fail to start and leaving projects offline when restarted. The incident affected projects running Postgres versions older than 15.1.1.57 and impacted the Dashboard component. Supabase resolved the issue by manually patching affected projects and rolling out a permanent fix across their fleet over a 9.3-hour period.

Supabase experienced capacity issues in US-East-1, US-East-2, and AP-Northeast-1 regions that caused failures in project creation, resize requests, and project restarts. The service temporarily disabled these operations in the affected regions while working with their infrastructure provider to obtain additional capacity. The incident was resolved after 1.5 hours when capacity was restored and all regions were re-enabled with normal functionality.

Supabase experienced a 3.3-hour outage where newly created projects were unreachable due to DNS resolution failures caused by issues propagating new DNS records in their zone. The problem affected the Dashboard component and was resolved by working with upstream network providers to implement a fix. DNS resolution returned to normal operation after the fix was implemented and monitored for stability.

Supabase experienced a 13-hour networking incident caused by their upstream provider that affected users primarily in North and South America, with symptoms including DNS lookup failures and HTTP 530/5XX errors when accessing Supabase projects. The issue impacted multiple components including Auth, Database, Edge Functions, Realtime, Storage, and API Gateway, preventing users from connecting to their projects and experiencing login issues. The incident was resolved after the upstream networking provider identified and implemented fixes across all affected regions, with service gradually improving as mitigations were rolled out region by region.

Supabase experienced delays in transactional email delivery for 22 hours, affecting welcome emails, OAuth approvals, project shutdown notifications, and payment failure alerts. Customers received notifications after their issues were already resolved, such as payment failure alerts arriving after payments were successfully processed. The email queues returned to normal after the team cleared the backlog and addressed the underlying causes of the email spike.

Supabase experienced a major incident where project creation failed in some APAC regions for 1.3 hours, while existing projects remained unaffected. The issue was caused by capacity problems in the impacted regions. The service was fully restored after the team recovered capacity and confirmed project creation was operating normally.