refresh

Read the live schema and write it to the state store. Use this to seed or repair state.

nschema refresh

The refresh command captures the whole schema, and so takes no project or --scope options. This makes migrations a lot safer, because it allow NSchema to warn about changes outside its managed set. For example, if you have an unmanaged orders table that depends on a managed customers table, without the full schema, dropping the customers table would produce a successful plan.

Refresh is how you record drift that happened between applies, or seed the state store with the current schema before you start managing a database. Since plan always diffs the recorded state against your project, a refresh is also how you tell NSchema about a change someone made to the database out of band.

A capture is an observation, not an adoption: it records what the database contains without adding anything to the managed set, so refreshing an unmanaged object never puts it at risk of being dropped.

Options

• --no-lock — skip taking the state-store lock for this run. Use it only when you’ve coordinated access by other means (for example you already hold the lock via nschema lock acquire).
• --force — replace an existing state payload that cannot be read, resetting the script ledger.

Recovery

If your state becomes corrupted somehow, the refresh operation will fail rather than overwriting it. The payload can still be pulled for repair, or you can pass --force to replace it, which resets therun-once script ledger. Restore the ledger afterwards with script untaint.