Coinbase said a July 14 service outage was caused by a Kubernetes resource naming conflict introduced during a routine configuration update, which knocked out a critical network component and disrupted transfers, card transactions and onchain services. The company said customer funds were never at risk.
How the incident started
In its post-incident review, Coinbase said it deployed a standard configuration change at 12:34 p.m. Eastern Time on July 14 to a shared production Kubernetes cluster that supports core infrastructure services. The update was part of a migration to a new service deployment model designed to improve system performance and reliability. Coinbase said the change went through its normal review process and had been classified as low risk.
The problem emerged because a pre-production check failed to catch a resource naming conflict. Coinbase said the configuration unexpectedly modified Kubernetes resources related to the Istio Ingress Gateway inside the cluster, making the ingress gateway unavailable. By 12:37 p.m. ET, all traffic entering that cluster had stopped, preventing internal services from reaching several infrastructure components.
Services that were affected
Coinbase said the disrupted asynchronous workflows covered nearly its entire funds movement system, including trade settlement, asset transfers and card transaction authorization. Once infrastructure services became unreachable, those workflows were paused across the platform.
- Retail users could not complete offchain trades, deposits or withdrawals, and some in-flight transactions appeared stuck before completing automatically after service was restored.
- Coinbase Card debit transactions temporarily failed. Credit card payments remained available, though some card management features were unavailable.
- Coinbase DEX swap services on Base and Solana were paused.
- Institutional clients on Coinbase Exchange and Prime saw transfer and settlement failures or delays.
- Customers using Coinbase Developer Platform could not complete onboarding, fund transfers or fiat funding services.
Recovery timeline
During the recovery effort, Coinbase said it paused affected workflows and kept users updated through its Retail, Exchange and Prime status pages. The ingress gateway was restored at 1:20 p.m. ET, and the incident was largely mitigated by 1:23 p.m. ET.
The system then began processing queued work. Coinbase said most services recovered quickly, and the remaining backlog was cleared within hours.
Steps Coinbase plans to take
Coinbase said it will make three changes to improve reliability after the outage:
- expand deployment-stage safeguards to automatically detect and block Kubernetes resource naming conflicts so updates do not affect unrelated components;
- increase redundancy between deployment tools and the underlying infrastructure so a failure does not hit recovery tooling at the same time;
- regularly review and rehearse emergency access mechanisms to speed recovery during major incidents.
Coinbase said its goal remains zero-downtime operations and that it will keep reworking infrastructure so similar failures can be rolled back and restored faster in the future. The company also apologized for the impact on users and said anyone still seeing related issues should contact official support.