banger/internal/store/migrations.go
Thales Maciel ec6fc9d185
store,bangerd: add --check-migrations flag for pre-swap schema check
Prerequisite for `banger update`. Before swapping a staged binary
into place, the updater needs to confirm the new bangerd recognises
the running install's DB schema. Without this, an operator could end
up with a service that won't open its store after the binary swap +
restart.

  * store.InspectSchemaState(path): opens the DB read-only (reusing
    OpenReadOnly's mode=ro DSN), reads the schema_migrations table,
    and classifies the relationship between applied and known IDs:
    SchemaCompatible (lockstep), SchemaMigrationsNeeded (binary
    newer, will auto-migrate on first Open), or SchemaIncompatible
    (DB has applied IDs the binary doesn't know about).
    Missing schema_migrations table is treated as "all migrations
    pending" rather than an error — matches the fresh-install case.
  * bangerd --check-migrations: opens the configured DB read-only,
    prints a one-line classification, and exits 0/1/2. The exit
    code is the contract:
        0 — compatible
        1 — migrations needed (binary newer; safe to swap)
        2 — incompatible (binary older than DB; abort the swap)
    Honours --system to pick between system StateDir and user mode.
  * bangerdExit indirection so future tests can capture the exit
    code without terminating the test process. Production points
    at os.Exit.

Tests cover the four classifications: compatible (fully migrated
DB), migrations-needed (only baseline applied), incompatible
(synthetic id=99 inserted), and missing-table (fresh DB). Live
exercise on this dev host returned `migrations needed: pending [3]
(binary will apply on first Open)` and exit 1, matching the
contract.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 18:41:31 -03:00

303 lines
8.6 KiB
Go

package store
import (
"database/sql"
"fmt"
"sort"
"time"
)
// migration is one ordered, atomic schema step. id must be unique and
// strictly increasing across the slice. name is a human-readable label
// stored alongside the id for debugging, and up receives a *sql.Tx so
// DDL + data backfills land atomically — either the migration fully
// applies and a schema_migrations row is written, or the whole thing
// rolls back and gets retried on next Open().
type migration struct {
id int
name string
up func(*sql.Tx) error
}
// migrations is the canonical ordered history. Append new migrations
// at the bottom with the next id. Never edit or reorder existing
// entries — installed DBs key off the id column.
var migrations = []migration{
{id: 1, name: "baseline", up: migrateBaseline},
{id: 2, name: "drop_images_docker", up: migrateDropImagesDocker},
{id: 3, name: "add_vm_workspace", up: migrateAddVMWorkspace},
}
// runMigrations ensures schema_migrations exists, then applies every
// migration whose id hasn't been recorded yet, in id order.
func runMigrations(db *sql.DB) error {
if _, err := db.Exec(`CREATE TABLE IF NOT EXISTS schema_migrations (
id INTEGER PRIMARY KEY,
name TEXT NOT NULL,
applied_at TEXT NOT NULL
)`); err != nil {
return fmt.Errorf("create schema_migrations: %w", err)
}
applied, err := loadAppliedMigrations(db)
if err != nil {
return err
}
sorted := make([]migration, len(migrations))
copy(sorted, migrations)
sort.Slice(sorted, func(i, j int) bool { return sorted[i].id < sorted[j].id })
seen := map[int]bool{}
for _, m := range sorted {
if seen[m.id] {
return fmt.Errorf("duplicate migration id %d (%q)", m.id, m.name)
}
seen[m.id] = true
}
for _, m := range sorted {
if _, ok := applied[m.id]; ok {
continue
}
if err := applyMigration(db, m); err != nil {
return fmt.Errorf("migration %d (%s): %w", m.id, m.name, err)
}
}
return nil
}
// SchemaCompatibility classifies the relationship between this
// binary's known migrations and a (possibly stale) DB's applied set.
type SchemaCompatibility int
const (
// SchemaCompatible: every applied id is known to this binary AND
// every known id has been applied. Binary and DB are in lockstep.
SchemaCompatible SchemaCompatibility = iota
// SchemaMigrationsNeeded: binary knows ids the DB hasn't applied
// yet. Open() would auto-migrate; safe.
SchemaMigrationsNeeded
// SchemaIncompatible: DB has applied ids this binary doesn't
// know about. Binary is older than the running install. Refuse
// the swap.
SchemaIncompatible
)
// SchemaState describes the migration status of a DB relative to
// this binary's compiled-in `migrations` slice. Used by
// `bangerd --check-migrations` to gate `banger update`'s binary swap
// before service restart — a staged binary must not be allowed to
// take over a DB whose schema it doesn't know how to read.
type SchemaState struct {
Compatibility SchemaCompatibility
AppliedIDs []int
KnownMaxID int
Pending []int // known IDs not yet applied
Unknown []int // applied IDs the binary doesn't recognise
}
// InspectSchemaState opens path read-only and reports how the DB's
// applied-migration set compares to the binary's known set. Returns
// an error only on real I/O failures (file missing, permission
// denied, corrupt SQLite); a "DB ahead of binary" state is reported
// via Compatibility, not as an error.
func InspectSchemaState(path string) (SchemaState, error) {
dsn, err := sqliteReadOnlyDSN(path)
if err != nil {
return SchemaState{}, err
}
db, err := sql.Open("sqlite", dsn)
if err != nil {
return SchemaState{}, err
}
defer db.Close()
if err := db.Ping(); err != nil {
return SchemaState{}, err
}
// schema_migrations may not exist on a fresh install. Treat that
// as "applied = ∅" rather than an error — the equivalent of
// "the new binary will create the table on first Open".
rows, err := db.Query("SELECT id FROM schema_migrations")
if err != nil {
// modernc.org/sqlite doesn't expose a typed "no such table"
// error; sniff the message. Anything else bubbles.
if errMissingTable(err) {
return classifySchemaState(nil), nil
}
return SchemaState{}, err
}
defer rows.Close()
var applied []int
for rows.Next() {
var id int
if err := rows.Scan(&id); err != nil {
return SchemaState{}, err
}
applied = append(applied, id)
}
if err := rows.Err(); err != nil {
return SchemaState{}, err
}
return classifySchemaState(applied), nil
}
func classifySchemaState(applied []int) SchemaState {
known := map[int]struct{}{}
knownMax := 0
for _, m := range migrations {
known[m.id] = struct{}{}
if m.id > knownMax {
knownMax = m.id
}
}
appliedSet := map[int]struct{}{}
var unknown []int
for _, id := range applied {
appliedSet[id] = struct{}{}
if _, ok := known[id]; !ok {
unknown = append(unknown, id)
}
}
var pending []int
for _, m := range migrations {
if _, ok := appliedSet[m.id]; !ok {
pending = append(pending, m.id)
}
}
state := SchemaState{
AppliedIDs: append([]int(nil), applied...),
KnownMaxID: knownMax,
Pending: pending,
Unknown: unknown,
}
switch {
case len(unknown) > 0:
state.Compatibility = SchemaIncompatible
case len(pending) > 0:
state.Compatibility = SchemaMigrationsNeeded
default:
state.Compatibility = SchemaCompatible
}
return state
}
func errMissingTable(err error) bool {
if err == nil {
return false
}
msg := err.Error()
// modernc.org/sqlite wraps the underlying SQLITE_ERROR with this
// canonical sub-string for missing-table errors.
return contains(msg, "no such table: schema_migrations")
}
func contains(s, sub string) bool {
if len(sub) > len(s) {
return false
}
for i := 0; i+len(sub) <= len(s); i++ {
if s[i:i+len(sub)] == sub {
return true
}
}
return false
}
func loadAppliedMigrations(db *sql.DB) (map[int]struct{}, error) {
rows, err := db.Query("SELECT id FROM schema_migrations")
if err != nil {
return nil, fmt.Errorf("load schema_migrations: %w", err)
}
defer rows.Close()
applied := map[int]struct{}{}
for rows.Next() {
var id int
if err := rows.Scan(&id); err != nil {
return nil, err
}
applied[id] = struct{}{}
}
return applied, rows.Err()
}
func applyMigration(db *sql.DB, m migration) error {
tx, err := db.Begin()
if err != nil {
return err
}
if err := m.up(tx); err != nil {
_ = tx.Rollback()
return err
}
if _, err := tx.Exec(
"INSERT INTO schema_migrations (id, name, applied_at) VALUES (?, ?, ?)",
m.id, m.name, time.Now().UTC().Format(time.RFC3339),
); err != nil {
_ = tx.Rollback()
return fmt.Errorf("record migration: %w", err)
}
return tx.Commit()
}
// migrateBaseline creates the full current schema.
func migrateBaseline(tx *sql.Tx) error {
stmts := []string{
`CREATE TABLE IF NOT EXISTS images (
id TEXT PRIMARY KEY,
name TEXT NOT NULL UNIQUE,
managed INTEGER NOT NULL DEFAULT 0,
artifact_dir TEXT,
rootfs_path TEXT NOT NULL,
work_seed_path TEXT,
kernel_path TEXT NOT NULL,
initrd_path TEXT,
modules_dir TEXT,
build_size TEXT,
seeded_ssh_public_key_fingerprint TEXT,
docker INTEGER NOT NULL DEFAULT 0,
created_at TEXT NOT NULL,
updated_at TEXT NOT NULL
);`,
`CREATE TABLE IF NOT EXISTS vms (
id TEXT PRIMARY KEY,
name TEXT NOT NULL UNIQUE,
image_id TEXT NOT NULL,
guest_ip TEXT NOT NULL UNIQUE,
state TEXT NOT NULL,
created_at TEXT NOT NULL,
updated_at TEXT NOT NULL,
last_touched_at TEXT NOT NULL,
spec_json TEXT NOT NULL,
runtime_json TEXT NOT NULL,
stats_json TEXT NOT NULL DEFAULT '{}',
FOREIGN KEY(image_id) REFERENCES images(id) ON DELETE RESTRICT
);`,
}
for _, stmt := range stmts {
if _, err := tx.Exec(stmt); err != nil {
return err
}
}
return nil
}
// migrateDropImagesDocker removes the legacy images.docker column.
// SQLite supports ALTER TABLE ... DROP COLUMN since 3.35 (2021), and
// banger ships against modern SQLite, so a single statement is enough.
// Existing values are simply discarded — the field never affected
// runtime behaviour.
func migrateDropImagesDocker(tx *sql.Tx) error {
_, err := tx.Exec(`ALTER TABLE images DROP COLUMN docker;`)
return err
}
// migrateAddVMWorkspace adds the workspace_json column that records
// the last workspace.prepare result (guest path, host source path,
// HEAD commit, and timestamp) per VM. Default '{}' means no workspace
// has been prepared yet. The column is managed exclusively via
// Store.SetVMWorkspace; lifecycle UpsertVM calls never touch it so
// workspace state survives VM stop/start cycles.
func migrateAddVMWorkspace(tx *sql.Tx) error {
_, err := tx.Exec(`ALTER TABLE vms ADD COLUMN workspace_json TEXT NOT NULL DEFAULT '{}'`)
return err
}