(2026-04-07 17:22(JST))
We sincerely apologize for the long-term login issues that occurred on our MyPage system, which also affected the ability of some new users to start using the service after registration. We deeply regret the inconvenience and trouble this has caused our users.
We will report on the overview, cause, and future measures to prevent recurrence of this incident.

1. Overview of the Disability

• Occurrence time: 2026-04-03 20:53(JST) to 2026-04-06 18:23(JST)
  • Login service restored for existing users: Around 13:41(JST) on 2026-04-06
  • Restoration of new user registration related services: Around 18:23(JST) on 2026-04-06
• Target service: MyPage
• Impact: All users experienced an issue where they were unable to complete the login process for their MyPage. Additionally, some users who registered during the outage period were unable to use the API even after the service was restored due to failures in related processes.

2. Causes and Circumstances

On April 3, 2026, at 20:53(JST), the communication path used to retrieve the settings information required for logging into MyPage was unintentionally changed, resulting in failures in some parts of the login process and new user registration process.

Details

On the MyPage, we used reCAPTCHA during the login process, and therefore obtained authentication keys from AWS Systems Manager Parameter Store.

The direct cause of this incident was that on 2026-04-03 at 20:53(JST), a VPC endpoint for AWS Systems Manager was created and Private DNS was enabled, which resulted in the access destination from the application server to ssm.ap-northeast-1.amazonaws.com switching from the conventional public endpoint to the new VPC endpoint. In addition, the security group (connection permission settings) configured for this VPC endpoint was not allowing access from the MyPage application server, which resulted in a failure to access Parameter Store during the login process.

Furthermore, because the new user registration process also retrieved information necessary for Stripe integration from AWS Systems Manager Parameter Store, some users who registered during the outage period did not have the necessary processes for starting to use the AmiVoice API completed successfully.

This issue came to light through an inquiry from an API user, and investigation and recovery work began around 09:02(JST) on 2026-04-06. Around 13:41(JST), the VPC endpoint settings were reviewed, and the login issue for MyPage was resolved. Subsequently, around 18:23(JST), failed Stripe-related jobs were re-executed, and the process that had stalled due to new user registration was completed, making the service available again. Finally, the scope of the impact was confirmed at 18:42(JST).

The root cause of the unintended changes was that a development and testing environment for some systems existed within the production environment, and network configuration changes that affected the production environment were made during the construction of this environment. While operational rules (review of the scope of impact, IaC management, and a general prohibition of changes before holidays) are applied to changes to the production environment, in this case, the construction of the development environment violated these rules.

Background:

Times of Day	Description
2026-04-03 (Fri) 20:53(JST)	An issue occurred. Create a VPC endpoint for Systems Manager and enable Private DNS.
2026-04-06 (Mon) 09:02(JST)	The development team has begun an investigation.
2026-04-06 (Mon) 10:40(JST)	Publication of the outage report on MyPage
2026-04-06 (Mon) 13:41(JST)	Login issue resolved.
2026-04-06 (Mon) 14:04(JST)	We will inform you of the recovery in the outage report.
2026-04-06 (Mon) 18:16(JST)	We confirmed that the Stripe integration process also failed in relation to the new user registration.
2026-04-06 (Mon) 18:23(JST)	Rerun Stripe-related failed jobs
2026-04-06 (Mon) 18:42(JST)	Confirmation of the scope of impact is complete.

3. Countermeasures

The following measures were implemented.

• Configuration Restoration: The production environment settings have been restored to their normal state.
• Rerunning failed Stripe-related jobs: Related processes that failed during the outage period have been rerun.

4. Measures to prevent recurrence

Separation of development and staging environments

Because some development and staging systems reside within the production environment, we have been gradually separating them. We will now completely separate the development and staging environments for MyPage from the production environment.

Reducing runtime dependencies on AWS Systems Manager Parameter Store

AWS Systems Manager Parameter Store allows you to inject values ​​at container startup using task definition secrets in ECS/Fargate. For fixed configuration values, instead of retrieving them each time a login or external integration process occurs, you can change the configuration to read them when the application server starts.

Strengthening detection and operations

Since the application server itself was functioning normally, the existing health monitoring system failed to detect the anomaly, resulting in a delay in discovering the problem until we received inquiries from API users. We will now add external API call failures to the monitoring scope to strengthen early problem detection.

(April 6, 2026 14:04(JST))
The service was restored around 13:50(JST). We will provide a further report on the cause and future countermeasures at a later date. This is just a quick update.

(April 6, 2026 10:38(JST))
We are currently experiencing an issue where users are unable to log in to their MyPage. We are investigating the cause and will provide further information as soon as it becomes available.

We sincerely apologize for any inconvenience this may have caused.