Coinbase Goes Down for 50 Minutes Over a Name Collision

Coinbase's trading and transfer services were degraded for roughly 50 minutes on Tuesday afternoon after a routine configuration update disabled the gateway that routes inbound traffic into the exchange's production systems. The company said the update was deployed at 12:34 p.m. ET and that inbound cluster traffic stopped three minutes later, at 12:37 p.m. ET. The gateway was restored at 1:20 p.m. ET and the incident was marked mitigated at 1:23 p.m. ET, with degraded service running to about 1:25 p.m. ET and residual processing delays after that. Withdrawals, card payments and on-chain swaps were among the functions affected. No customer funds were at risk, according to the company's postmortem.

The Failure: A Collision in Resource Names

The postmortem attributed the outage to the update inadvertently modifying Kubernetes resources tied to the Istio ingress gateway. Kubernetes is the open-source system that schedules and runs the many small containers a service of Coinbase's size is assembled from, and an ingress gateway is the component that accepts connections from outside and hands them to the services behind it. Names are how Kubernetes identifies the objects inside a cluster, so when a new definition claims a name that an existing object already holds, the cluster can apply one configuration where the other was expected. Here the gateway's own resources were altered, it became unavailable, and requests arriving from the public internet had nothing to route them onward — the exchange was running, but nothing could reach it. Coinbase described the trigger as "a collision in resource names that was not detected by our pre-production checks."

The Blast Radius: Withdrawals, Card Payments and Settlement

Because the failure sat at the entry point rather than inside any one product, it cut across almost every surface Coinbase operates at once. The company listed the affected functions as:

Transfers and withdrawals across the retail, institutional and developer platforms
Coinbase Card debit transactions
On-chain swaps on Base and Solana
Settlement on Exchange and on Prime, the institutional venue
Onboarding on the Developer Platform

The pattern is familiar from other shared-infrastructure failures: one component that everything depends on fails, and products with no functional relationship to each other stop working in the same minute. Coinbase said the disruption was to availability rather than to balances or custody. Processing delays persisted after the gateway came back, as queued work cleared.

The Context: A Second Coinbase-Operated Outage in Three Weeks

It was the second interruption in three weeks to infrastructure Coinbase runs. In late June, Base — the Ethereum layer-2 network the company operates — halted block production twice in two days, first for close to two hours and then for about 20 minutes during the restart, after a bug in its sequencer's block-building logic produced an invalid state transition. That episode delayed Base's Beryl hardfork and revived criticism of the network's reliance on a single centralized sequencer with no automatic failover. Tuesday's outage had an unrelated cause, but it landed on the same operator and in the same public ledger of downtime.

The timing was awkward in another respect. Inbound traffic stopped at 12:37 p.m. ET, the same minute the June consumer price index was released, and bitcoin ran roughly 4% higher over the afternoon to above $64,500 while parts of the exchange were unreachable. Market data cited that day put June spot bitcoin ETF outflows at $4.5 billion, with daily trading volume down 78% to $1.25 billion from a $5.8 billion peak — a thin tape, but an actively moving one.

What Is Unresolved: The Checks That Did Not Catch It

Restoring the gateway took under an hour, but the harder question the postmortem raises is why a name collision reached production at all. Coinbase's own account states that its pre-production checks did not detect the conflict, which places the gap in validation rather than in the update itself. The company has not detailed what changes to those checks will follow, and no timeline for them was given alongside the postmortem. For institutional customers whose settlement paused for the better part of an hour, that is the piece worth watching.