opschain admin database backup-info

Included from Basic tier

Verifies the system is QUIESCED and emits the backup manifest — a GET of /api/admin/database/backup_info. The PostgreSQL backup lane is DBA-first: OpsChain never streams database bytes; it makes the window safe and records what a complete capture must contain. The lane is opschain admin platform quiesce → backup-info → the DBA's pg_dump (and the blobstore copy) → opschain admin platform resume.

The manifest records:

• the server version and commit, and the goose schema migration level;
• exact per-table row counts of the durable tables (stable because the write barrier is closed), under tables; the volatile operational tables — replica membership and heartbeats, the named and job leases, runner liveness tokens and challenges, attempt-log ownership, the run inbox — are named under tables_excluded and never counted, because a restored server legitimately holds a different set of them;
• the blobstore root path that must be captured alongside the dump, and every blob digest the database references;
• the quiesce posture (id, reason, who, when);
• the blessed pg_dump recipe (custom format), with the CloudNativePG note for HA estates;
• a SHA-256 content digest over the manifest body (recomputed canonically, so re-serialising the file does not break it).

opschain admin database verify-restore later reconciles all of it against the restored store.

Flags

Flag	Meaning
--server <url>	Server base URL (default: the active context / OPSCHAIN_SERVER).
--auth-token <token>	Admin bearer token override (default: your opschain login session, then $OPSCHAIN_ADMIN_TOKEN, then the token stored by opschain admin login).
--out <file>	Write the manifest to this file and print a summary; without it the manifest JSON prints to stdout.

Exit codes

Code	Meaning
0	The manifest was emitted.
1	The system is not quiesced (409), the store is not PostgreSQL (422), request failed, or the token was rejected.
2	Usage error — no server or no credential resolvable.

Named errors

• the system is not quiesced — the backup manifest describes a database that must not be changing: run 'opschain admin platform quiesce ...' — HTTP 409 (exit 1; there is no warn-and-run lane).
• the configured store is not PostgreSQL — 'opschain admin database backup --out <file>' and 'restore <file>' are the embedded SQLite lane — HTTP 422 (exit 1).

Examples

opschain admin database backup-info --out manifest.json
opschain admin database backup-info > manifest.json

The full window, in order: quiesce (opschain admin platform quiesce, with a reason), fetch this manifest, run the pg_dump recipe the manifest prints, copy the blobstore root it names, then opschain admin platform resume.

See also

• opschain docs cli/admin-database-verify-restore — the post-pg_restore reconciliation of this manifest.
• opschain docs cli/admin-platform-quiesce — opening (and closing) the backup window.
• opschain docs cli/admin-database-backup — the embedded SQLite lane, where the built-in IS the right tool.