Persist relational state in a Durable Object using the SQLite storage backend (ctx.storage.sql) and schedule follow-up work with the Alarms API

Documented steps

1. In wrangler.jsonc, define the Durable Object binding and enable the SQLite backend via a migration: {"migrations": [{"tag": "v1", "new_sqlite_classes": ["MyDurableObject"]}]}. In wrangler.toml this is a [[migrations]] block with tag and new_sqlite_classes. This must be set when the class is first created.
2. In the Durable Object constructor, run schema setup once: ctx.storage.sql.exec("CREATE TABLE IF NOT EXISTS items (id INTEGER PRIMARY KEY, name TEXT, created_at INTEGER)").
3. Perform parameterized reads and writes with ? placeholders: let cursor = ctx.storage.sql.exec("INSERT INTO items (name, created_at) VALUES (?, ?)", name, Date.now()). Consume results via iteration, .toArray(), .one(), or .raw(), and inspect cursor.rowsRead / cursor.rowsWritten for billing.
4. Wrap multi-statement writes atomically with ctx.storage.transactionSync(() => { ... }) for synchronous SQL, or ctx.storage.transaction(async (txn) => { ... }) for async workflows. Any run of writes with no intervening await is already atomic.
5. Schedule follow-up work: await ctx.storage.setAlarm(Date.now() + 60_000). Only one alarm can be pending per Durable Object, so record multiple pending events as rows in SQLite rather than as multiple alarms.
6. Implement async alarm(alarmInfo) { ... } on the class: query the table for due work, process it, and call ctx.storage.setAlarm(...) again inside the handler if more work remains, since each alarm fires once.
7. Use alarmInfo.retryCount and alarmInfo.isRetry to detect retried executions and guard side effects with an idempotency key stored in the table before reprocessing.
8. Use await ctx.storage.getAlarm() to read the next scheduled time (ms since epoch or null) and await ctx.storage.deleteAlarm() to cancel a pending alarm.
9. Deploy with npx wrangler deploy. Because the new_sqlite_classes migration is irreversible, confirm the class name and backend before the first deploy.
10. Official documentation verified for this route: https://developers.cloudflare.com/durable-objects/api/storage-api/ | https://developers.cloudflare.com/durable-objects/api/alarms/ | https://developers.cloudflare.com/durable-objects/platform/limits/ | https://developers.cloudflare.com/durable-objects/reference/durable-objects-migrations/