Cryptographically erase an AshVault scope.
This is irreversible
Every value ever encrypted under the scope becomes permanently unrecoverable. There is no undo, no backup of the key, and no support ticket that brings it back. The provider writes a tombstone, so the scope can never be re-minted either — silently re-minting would turn crypto-erasure into silent data loss.
Usage
mix ash_vault.destroy_keys MyApp.Accounts.User --tenant acmeThe task prints the scope, every resource and field that is about to become
undecryptable, and the current key version. It then asks you to type the scope name
back. --yes does not skip that prompt, and there is deliberately no --force: a
deploy script cannot destroy a tenant's data by passing one more flag.
In MIX_ENV=prod it additionally refuses to run unless MIX_ENV was passed explicitly
or --i-know-what-this-does is given.
Options
--tenant TENANT required when the key scope is `:tenant`
--domain MODULE Ash domain (inferred from `:ash_domains` when omitted)
--i-know-what-this-does required in production
--yes skips *other* confirmations; never the scope promptIdempotence
Destroying an already-destroyed scope prints status=already_destroyed and exits 0. It
does not prompt, because there is nothing left to lose.
Output
ash_vault.destroy_keys=target resource=MyApp.Accounts.User tenant=acme scope=acme \
vault=MyApp.Vault provider=MyApp.KeyProvider status=active version=3 fields=4
undecryptable=MyApp.Accounts.User.email
undecryptable=MyApp.Accounts.User.ssn
ash_vault.destroy_keys=done scope=acme versions_destroyed=3 tombstone=confirmed
DESTROYED MyApp.Vault scope acme: 3 key version(s), 4 field(s) across 2 resource(s).The tombstone is re-read from the provider before success is reported: if the provider
does not answer {:error, :destroyed} afterwards, the task exits non-zero rather than
claiming an erasure that did not happen.