DigitalOcean Outage History

DigitalOcean's App Platform experienced deployment delays starting at 20:40 UTC, preventing users from deploying new apps for 4.3 hours globally. The engineering team worked with their upstream provider to implement a fix, and the issue was fully resolved by 23:00 UTC with all stalled deployments completing successfully.

DigitalOcean experienced a 5.2-hour incident where newly provisioned Managed Kubernetes nodes encountered DNS timeouts, specifically affecting customers running NAT Gateways within their VPC and VPC-native clusters. The DNS timeouts temporarily disrupted cluster services for affected configurations. The engineering team implemented a fix and confirmed resolution, with all cluster services returning to normal operation.

DigitalOcean's App Platform experienced deployment issues for 1.8 hours, causing delays and failures when users attempted to deploy new and existing applications globally. The engineering team implemented a fix and confirmed full resolution, with deployments returning to normal functionality.

DigitalOcean experienced connectivity issues between Internal Load Balancers and their associated target droplets from 19:57 UTC to 01:03 UTC, potentially causing service disruptions or traffic routing failures for customers globally. The engineering team implemented mitigation measures and confirmed full resolution of the issue. Internal Load Balancers are now functioning normally after approximately 5 hours of impact.

DigitalOcean's App Platform experienced build failures for applications using older versions of the Node.js buildpack, preventing users from successfully deploying their applications. The issue affected customers globally for approximately 11.7 hours between 14:30 UTC on February 26th and 00:01 UTC on February 27th. Engineering teams implemented a fix and restored normal build and deployment operations, with all App Platform builds now succeeding as expected.

DigitalOcean's Llama 3.3-70B model experienced intermittent errors for 6.4 hours, affecting users making serverless inference requests via APIs and Agents. The issue was caused by a few problematic requests to the model and was resolved after deploying a fix and monitoring the resources.