19-Nov-2026 — track S3 moves, show the next migration step and fix S3 copy, key rotation and Validate #231
What's new
- AWS S3 sits next to MongoDB Atlas under Environment Management, with Provisioning, Buckets and Environments tabs in the same shape as the Atlas page.
- The Environments tab lists each site's current bucket against its expected
…-eu-west-2 bucket in the Ramblers NGX account, with sites still on the old account sorted first. Each row has an S3 console link, and remaining sites have a "Move to Ramblers storage" icon that opens Environment Migration for that site.
- Plan & Restore is now a single panel. It shows where the database and the photos and files move from and to, then one button for the next step: create database login, check, copy, then switch the site over. There are no choices to make. The page works out what is left from the site's migration history, so a site whose database has already moved is only offered its photos and files.
- The Ramblers storage bucket and keys are filled in automatically from the platform setup keys and never shown.
- Copying cannot start until the check has passed.
- While a step runs, the panel shows its stage in plain words (for example "Backing up the current database", "Restoring into the new cluster") and the time of the last update. When everything has moved it says so.
- Switch over works after a page reload. The new database login's password is saved, encrypted, with the migration record, so the switch no longer depends on the page still holding it. The saved copy is removed once the site has switched.
- Reloading the page while a copy is running picks its progress back up, and after a switch over the page offers whatever is still left to move.
- Copy S3 objects no longer fails with "Access Denied" for sites still on the old account. The copy now reads the source bucket with the site's own key, as the nightly backup already does, instead of the archive key, which now belongs to the London archive account.
- Rotate Credentials waits up to a minute for a newly created bucket-scoped key to start working before reporting a failure. Before, it tested the key within a second of creating it and failed with "UnknownError" while AWS was still activating it.
- Retrying a rotation keeps the site's current live key. When the IAM user already has two keys, the oldest key other than the live one is removed, so a failed retry cannot leave the site without S3 access.
- The migration history records the site's own key as the source key.
- Validate Plan no longer fails straight after a destination database user is created. It waits for MongoDB Atlas to activate the user, the same way new site registration does, and keeps the migration's heartbeat fresh while it waits.
- The database user timeout message now says "Retry in a few minutes", since it is no longer only shown during registration.