Summary

Following a Cloud deployment, some charger boot requests began failing. The issue affected requests that reused data cached before the deployment; requests using newly generated data were unaffected.

Root cause

The application cached a structured charger-model record for up to one hour. A code change altered how that record was represented, but the cache persisted across the deployment. When the new application version encountered an older cached record, it could not restore it correctly and returned an HTTP 500 response.

Why we did not detect it earlier

Our automated tests start with a clean cache and verified both model creation and reuse within a single application version. They did not simulate cached data surviving an application upgrade, which was the condition required to trigger this failure.

Resolution and prevention

We removed the unnecessary cache and now use the existing indexed database lookup directly. We also added a regression test that reproduces the cross-deployment cache condition.

We are reviewing other persistent caches for version-dependent records and will prefer stable, version-independent cache formats where caching remains necessary.