DigitalOcean Outage History

DigitalOcean's Serverless Inference service experienced a 41-minute outage caused by an incident with their provider Anthropic. The outage affected all users attempting to access Sonnet 4.6 and Opus 4.7 AI models. The issue was resolved once Anthropic fixed their underlying service disruption.

DigitalOcean's DNS service experienced an issue where DNS resolution continued to work, but new DNS record changes failed to propagate to the edge, preventing Let's Encrypt certificate creation and delaying Managed MongoDB cluster provisioning and scaling operations. The engineering team implemented a fix after several hours and processed the backlog of delayed requests. The incident was fully resolved after 6.9 hours with all affected services returning to normal operation.

The GradientAI Agent Platform Playground experienced a major outage where users encountered "Something went wrong" errors for all agent interactions through the web interface. API endpoints continued to function normally throughout the incident. The issue was identified, fixed, and resolved within 1.3 hours with normal service restored.

DigitalOcean's Control Panel experienced an issue that prevented users from enabling Two-Factor Authentication (2FA) and Google/GitHub authentication, while also causing access problems for teams with secure sign-in enabled. The incident lasted 3.5 hours, from 5:35 to 14:25 UTC, affecting the Cloud Control Panel component. The engineering team implemented necessary changes and confirmed service was restored to normal functionality.

An upstream Let's Encrypt outage prevented DigitalOcean customers from creating new SSL certificates for Spaces, Load Balancers, and App Platform Custom Domains, while also causing delays in Managed Database operations including creates, forks, and restores for MongoDB, PostgreSQL, and MySQL. The incident lasted 1.2 hours and resolved automatically when Let's Encrypt restored service, with affected operations completing normally afterward.

DigitalOcean experienced a 2-hour service disruption affecting multiple services in their NYC2 region, including Droplet connectivity and API requests. Users with resources in NYC2 encountered connectivity issues and problems accessing various services during this period. The engineering team identified and implemented a fix, restoring normal operations across all affected services.

DigitalOcean's serverless inference service experienced elevated 5xx errors and "context canceled" responses for 5.9 hours, causing intermittent request failures for customers. The team identified the root cause and restored service by implementing tighter rate limits to prevent recurrence. All services returned to normal operation with continued monitoring to ensure ongoing reliability.

DigitalOcean's Serverless Inference service experienced intermittent rate limiting issues affecting customers using Anthropic models, causing request failures and HTTP 429 responses over approximately two hours. The engineering team identified the root cause and applied a mitigation to restore service functionality. The incident was fully resolved after 1.1 hours of active investigation and monitoring.

DigitalOcean experienced intermittent internal errors affecting specific AI models (Llama 3.3 70B, GPT OSS 120B, GPT OSS 20B, Qwen3 32B, and Deepseek R1 70B) in their ATL1 region's Serverless Inference service. Users with models hosted in ATL1 experienced intermittent errors when attempting to use Serverless Inference. The incident was resolved after 1.5 hours.

DigitalOcean's App Platform experienced build failures causing deployment errors, and Kubernetes (DOKS) nodes were being incorrectly marked as unhealthy by load balancers, impacting traffic routing. The engineering team implemented a fix after 2 hours, and all App Platform deployments returned to normal operation with the issue fully resolved at 09:22 UTC.