DigitalOcean Outage History

A widespread external issue affected DNS resolution for .co top-level domains, causing query errors for users regardless of which DNS resolver they used. DigitalOcean's engineering team deployed temporary backend mitigations to minimize impact on customers using their DNS resolvers. The issue was resolved after 16.9 hours with DNS resolution for .co domains returning to normal operation.

DigitalOcean's Serverless Inference service experienced elevated error rates and service instability for 1.8 hours, causing intermittent HTTP 500 errors and terminated connections that resulted in some failed requests. The engineering team implemented a fix and confirmed full resolution, with all systems returning to normal operation.

DigitalOcean experienced a 2.5-hour outage affecting Managed Database resize operations globally. Users encountered errors when attempting to resize databases through both the Cloud Control Panel and API across all regions. The engineering team implemented a fix and restored normal resize functionality.

DigitalOcean experienced an issue preventing users from creating new Droplets across all regions, affecting the Cloud Control Panel and API for 35 minutes. The engineering team identified the root cause, implemented a fix, and confirmed full resolution of the Droplet creation functionality.

DigitalOcean's Serverless Inference service experienced high error rates and elevated latency for the Qwen 3 32B model in the tor1 region for 3 hours starting at 10:46 UTC. The issue was caused by higher-than-expected request volume without sufficient resources to scale, resulting in capacity constraints and multiple workers in a pending state. The service was restored by expanding the node pool size to improve available capacity, along with implementing stability improvements to prevent similar issues.

DigitalOcean's Serverless inference service experienced an issue causing high error rates specifically for the llama 3.3 70b open source model. The incident lasted 5.6 hours, during which users encountered failures when attempting to use this particular model. The engineering team implemented a fix and monitored the results before fully resolving the issue.

DigitalOcean's App Platform experienced deployment delays and failures in the Frankfurt (FRA1) region for approximately 11.5 hours, specifically affecting new app deployments and updates using Dedicated Egress IP. Users encountered delays when creating new apps or deploying existing apps, though running applications continued to function normally. The engineering team deployed a fix and confirmed resolution, advising users to redeploy any failed deployments.

DigitalOcean's Gradient AI Platform experienced a major outage lasting 5.4 hours, causing agents and services to become inaccessible with users experiencing failures and unresponsiveness when attempting to use these features. The engineering team identified the root cause and implemented a fix, fully restoring service functionality. All agents are now operational and the system is being monitored for stability.

DigitalOcean's Gradient AI service experienced availability issues affecting multiple AI models including Llama3.1-8b, Qwen3-32b, several embedding models, and Guardrails functionality, while the Llama3.3-70b model suffered degraded performance. The incident lasted 4.8 hours and impacted users' ability to access models and run inference operations. Engineering teams implemented a fix that fully restored service and returned all affected models to healthy status.

DigitalOcean experienced degraded performance with BYOK (Bring Your Own Key) Anthropic models affecting all Gradient AI agents and serverless inference for 39 minutes. Users encountered reduced performance when using these AI services during the incident. The engineering team resolved the issue and restored normal functionality to all Anthropic BYOK models in Gradient AI.