Skip to content

Verified cloud accounts across 14 providers

What transfers when you change cloud accounts, and what does not

Moving between accounts inside the same provider is simpler than changing provider, but several things do not come along. Knowing which saves a surprise at cutover.

Changing accounts within one provider feels like it should be a copy operation. The platform is the same, the services are the same, the regions are the same. In practice the account is the boundary for more things than most teams expect, and the items that do not cross it are the ones that cause the incident.

Things that generally do move, with effort

  • Virtual machine images and snapshots. Most providers allow an image or snapshot to be shared with, or copied to, another account.
  • Object storage contents. A server side copy between accounts is usually supported and avoids pulling data out and back in.
  • Database contents. Through a snapshot share, a dump and restore, or replication, depending on the engine and how much downtime is acceptable.
  • Infrastructure definitions. If the environment is described in code, recreating it in a new account is largely a matter of changing credentials and a few identifiers.

Things that do not cross the boundary

These are account scoped by design, and each one needs an explicit plan.

  • Identities, roles and policies. Users, service accounts, keys and the policies attached to them are local to the account. They must be recreated, and anything holding an old credential must be updated.
  • Service quotas. A new account starts with default limits, which are often lower than the limits the old account accumulated. Quota increases usually require a request and are not instant, so they belong early in the plan rather than on cutover day.
  • Private network addressing. Networks, subnets and their peerings are account scoped. If address ranges were chosen without a plan, this is the moment that overlapping ranges become a problem.
  • Resource identifiers. Anything that embeds the account in its name or address changes. Hardcoded identifiers in application configuration are a frequent source of post migration failures.
  • Billing history and commitments. Past spend stays with the old account, and any committed use or reserved capacity discount is usually tied to the account that purchased it.

Credentials deserve their own checklist

Every credential issued by the old account stops being useful, and every system holding one needs updating. The list is longer than it first appears: continuous integration pipelines, deployment tooling, monitoring agents, backup jobs, third party integrations and anything a developer configured locally. Enumerating these before the move is considerably easier than discovering them from failure alerts afterwards.

A cutover order that limits surprises

  1. Recreate the network and identity layer in the new account, and request quota increases.
  2. Rebuild the environment from code, pointing at the new account.
  3. Copy data, then verify it.
  4. Re-issue every credential and update each consumer.
  5. Move traffic gradually, with the old account still able to serve.
  6. Keep the old account open, with access restricted, until at least one full billing and backup cycle has passed in the new one.

Why teams do it anyway

Despite the friction, account moves are common, and usually for one of a few reasons: separating environments that were sharing an account, splitting client work that had been combined, consolidating accounts after a reorganisation, or starting fresh in an account whose history and limits are better suited to the work. All of those are easier than they sound once the account is treated as the boundary it actually is.

Chat on Telegram
Scroll to Top