Revision 1 — the current text. The evidence on this page is this revision’s own — it has not been carried forward from, or to, any other revision.
Covers genuine duplicate-key inserts and the check-then-insert race that produces the same error under concurrency. Does not cover foreign-key constraint failures (a different SQLITE_CONSTRAINT variant) or CHECK-constraint failures — those need a different diagnosis of the data itself, not of insert timing.
Unverifiedno reproductions yetWhy this confidence?8 steps, exactly as this revision published them.
A write to a table with a UNIQUE column or index (often the primary business key, like userid or email) fails with D1ERROR: UNIQUE constraint failed.
What happens next
Query the table directly for the value that was rejected.
SELECT * FROM subscriptions WHERE user_id = ?;Expected result
A row is returned, showing the value is already present.What happens next
If a duplicate for this key is a legitimate, expected case (e.g. re-running a signup flow), use SQLite's upsert syntax instead of a bare INSERT so the constraint stops being an error path at all.
Changes system or service state. Review before running.
Changes the insert into an upsert: a new user_id is inserted as before, but an existing user_id now has its plan overwritten instead of throwing. Review whether overwrite is the behaviour you actually want before shipping this.
INSERT INTO subscriptions (user_id, plan) VALUES (?, ?)
ON CONFLICT(user_id) DO UPDATE SET plan = excluded.plan;What happens next
If a SELECT to check for an existing row ran immediately before the INSERT and found nothing, but the INSERT still failed, another request for the same key completed its own insert in between the two calls. D1 gives you no cross-request lock to make a check-then-insert pattern safe from that race.
npx wrangler tail --format prettyExpected result
Two near-simultaneous invocations for the same user_id, both passing their SELECT check, one losing the subsequent INSERT.What happens next
Send the same insert twice (or trigger the flow that previously raced) and confirm it no longer throws an uncaught error to the caller.
curl -s -X POST https://your-worker.example.workers.dev/subscribe -d '{"userId":"u_1"}' & curl -s -X POST https://your-worker.example.workers.dev/subscribe -d '{"userId":"u_1"}' & waitExpected result
Both requests return a normal response; neither surfaces an unhandled SQLITE_CONSTRAINT error.What happens next
Remove the SELECT-then-INSERT pattern and attempt the INSERT directly, catching SQLITE_CONSTRAINT as your "already exists" signal. This is race-proof because SQLite enforces the constraint atomically at insert time, which a separate SELECT beforehand cannot be.
Changes system or service state. Review before running.
Changes error handling only: a duplicate insert now resolves gracefully instead of throwing an uncaught error to the caller. It does not change what gets written for a genuinely new row.
try {
await env.DB.prepare("INSERT INTO subscriptions (user_id, plan) VALUES (?, ?)").bind(userId, plan).run();
} catch (err) {
if (String(err).includes("SQLITE_CONSTRAINT")) {
// already exists — treat as success, or fetch and return the existing row
} else {
throw err;
}
}What happens next
If no existing row matches and this isn't a check-then-insert race, look at the schema for an unintentionally narrow UNIQUE index — for example a compound key that's more restrictive than the application logic assumes.
Duplicate inserts for the same key are now handled deliberately instead of throwing an unhandled error.
What would strengthen it: 6 more independent reproductions. Reproductions from 3 more distinct environments.
This counts only what was recorded against revision 1 itself. Nothing reported against another revision is included here — see the revision history for why.