|
|
| 13:54 |
| Thursday 7 September 2023. Reconstructed to the hour from the company's post-incident report. |
| At 1:54 PM Eastern, Square products and services became unavailable. Sellers mid-transaction found readers that would not complete a sale. The devices were powered and connected; what they could not reach was the authorization layer behind them. |
| What had changed was not in the payment code. A host-based firewall rule deployed that morning at 11:04 had increased on-node firewall rule sizes, and the internal DNS caches that had been absorbing the resulting load for nearly three hours expired at once. Every service that depended on DNS to find its counterpart went down together. |
|
|
| 7 Sep 2023, 11:04 AM ET | A host-based firewall rule change is deployed to enable communication between regions. The change increases the size of firewall rules stored on each node. Internal DNS caches continue resolving service requests and no payment surface registers a problem. |
|
| 7 Sep 2023, 1:54 PM ET | Square products and services become unavailable. The DNS caches have expired under the enlarged firewall rule set, and services can no longer locate one another. Two minutes later, at 1:56 PM, a DNS zone change is applied. |
|
| 7 Sep 2023, 2:02 PM ET | Engineers are notified of infrastructure issues. Incident response begins, focused on DNS investigation. |
|
| 7 Sep 2023, 2:52 PM ET | Work begins to recover internal access and tooling. The internal tools and services used for troubleshooting and recovery resolve through the same DNS layer and are also down. Engineers must rebuild their own access before they can begin restoring payment services. |
|
| 7 Sep 2023, 3:56 PM ET | Networking traffic is shed to DNS servers. Manual work begins to bring up new DNS servers, because the Square deployment pipeline is unavailable. |
|
| 7 Sep 2023, 6:29 PM ET | Internal access services recover, more than four hours after the outage began. Engineers can now work in parallel on authorization and the control plane. |
|
| 7 Sep 2023, 8:25 PM ET | The status page confirms Balance and Transfers as separately affected, a distinct item from the payment processing outage acknowledged hours earlier. Sellers whose terminals were down and whose funds were frozen encountered both faults at once. |
|
| 7 Sep 2023, 8:36 PM ET | The deployment pipeline recovers. Automated code and configuration changes are possible again, after nearly five hours of manual work. |
|
| 8 Sep 2023, 12:55 AM ET | After hours of manual rebuilding and a reverted configuration push, all DNS servers report healthy. |
|
| 8 Sep 2023, 2:40 AM ET | Payment traffic fully recovered. The incident is marked resolved on the status page at 5:19 AM. |
|
| 8 Sep 2023, 9:52 AM to 1:57 PM ET | Background processing of offline payments begins at 9:52 AM. Completed at 1:57 PM for sellers on supported configurations. Transactions queued on devices since the previous afternoon clear through the system over the course of four hours. |
|