Allocate order numbers atomically instead of read-then-increment

Order ids were duplicating in production: 160 distinct (tenant, orderid) pairs
are shared by more than one order, worst of them "1135-1" on 108 orders, and
every order tenant 1147 has ever placed is numbered "1147-1".

getSequenceno read MAX(seqno)+1 and updateSeqno incremented, both against
r.db rather than the order's transaction and separated by the whole order
insert. Two concurrent orders therefore read the same number before either
wrote, and an order that rolled back still consumed one. Three further
defects made it worse:

  - A NULL orderseqno made COALESCE(MAX(orderseqno) + 1, 1) evaluate
    NULL + 1 = NULL and fall through to a hardcoded "<tenantid>-1". The
    increment then computed NULL + 1 = NULL too, so the counter could never
    leave NULL and every subsequent order reused that same id.

  - Tenants with several ordersequences rows (tenant 1135 has ~25) hit a
    GROUP BY returning multiple rows, of which Scan kept the first
    arbitrarily, while the increment updated all of them.

  - A tenant with no row at all fell back to "<tenantid>-1" indefinitely,
    because nothing ever created one.

nextSequenceNo replaces both functions with a single UPDATE ... RETURNING run
inside the caller's transaction, so the counter row stays locked until the
order commits and concurrent orders queue rather than collide. A NULL seeds
from the tenant's existing order count — at least as high as any number
already issued, so recovery cannot reissue a used id — the counter is pinned
to the tenant's lowest sequenceid so reads and writes address one row, and a
missing row is created on first use.

Verified against production data in rolled-back transactions: tenant 1147
(NULL) now yields 1147-9, 1147-10, ...; tenant 1135 (NULL plus duplicate rows)
1135-356 onward; tenant 916 keeps its 916-2024115209 subprefix format; an
unknown tenant creates its row and starts at 1. Eight concurrent allocations
produced eight distinct ids. Two real orders through the API returned 1147-9
and 1147-10, then were cancelled with stock restoring to its baseline.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-07-29 13:13:43 +05:30
parent c94ddd34c7
commit a11c4843ca

View File

@@ -1091,40 +1091,35 @@ func (r *orderRepository) UpdateOrder(order *models.Orders) error {
return tx.Commit().Error return tx.Commit().Error
} }
func (r *orderRepository) getSequenceno(tid int, prefix string) string { // nextSequenceNo claims the next order (or invoice) number for a tenant and
type SeqResult struct { // returns it formatted as tenantid-subprefix+seqno (e.g. 916-2024115209).
Orderseqno string //
} // It runs inside the caller's transaction and both reads and increments the
// counter in a single UPDATE ... RETURNING, which is what makes the number
var q1 string // unique. The previous implementation split this into getSequenceno (read
// Formats the ID as tenantid-subprefix+seqno (e.g., 908-20245189) // MAX+1) and updateSeqno (increment), both on r.db rather than the order's
switch prefix { // transaction, so two concurrent orders read the same value before either
case "ORD": // wrote, and a rolled-back order still consumed a number.
q1 = `SELECT CONCAT(tenantid, '-', //
CASE WHEN subprefix IS NULL OR CAST(subprefix AS TEXT) IN ('0', '0.0', '') THEN '' ELSE CAST(subprefix AS TEXT) END, // Three further defects that produced duplicate ids in production:
COALESCE(MAX(orderseqno) + 1, 1)) AS orderseqno //
FROM ordersequences WHERE tenantid = ? // - A NULL orderseqno made COALESCE(MAX(orderseqno) + 1, 1) evaluate
GROUP BY tenantid, subprefix` // NULL + 1 = NULL, falling through to a hardcoded "<tenantid>-1"; the
case "INV": // increment then computed NULL + 1 = NULL as well, so the counter could
q1 = `SELECT CONCAT(tenantid, '-', // never leave NULL. Every order such a tenant ever placed was numbered
CASE WHEN subprefix IS NULL OR CAST(subprefix AS TEXT) IN ('0', '0.0', '') THEN '' ELSE CAST(subprefix AS TEXT) END, // "<tenantid>-1" — 108 orders share "1135-1" today. A NULL is now seeded
COALESCE(MAX(invoiceseqno) + 1, 1)) AS orderseqno // from the tenant's existing order count, which is at least as high as any
FROM ordersequences WHERE tenantid = ? // number already handed out, so recovery never reissues a used id.
GROUP BY tenantid, subprefix` //
} // - Tenants with more than one ordersequences row (tenant 1135 has ~25) hit
// a GROUP BY that returned several rows, of which Scan silently kept the
var result SeqResult // first, while the increment updated every row. The counter is now pinned
r.db.Raw(q1, tid).Scan(&result) // to the tenant's lowest sequenceid, so reads and writes address the same
// row whatever duplicates exist.
// Fallback if no row exists in the database //
if result.Orderseqno == "" { // - A tenant with no row at all fell back to "<tenantid>-1" forever, since
return fmt.Sprintf("%d-1", tid) // nothing created one. The row is now created on first use.
} func nextSequenceNo(tx *gorm.DB, tid int, prefix string) (string, error) {
return result.Orderseqno
}
func (r *orderRepository) updateSeqno(tid int, prefix string) error {
var field string var field string
switch prefix { switch prefix {
case "ORD": case "ORD":
@@ -1132,26 +1127,42 @@ func (r *orderRepository) updateSeqno(tid int, prefix string) error {
case "INV": case "INV":
field = "invoiceseqno" field = "invoiceseqno"
default: default:
return fmt.Errorf("invalid prefix: %s", prefix) return "", fmt.Errorf("invalid prefix: %s", prefix)
} }
// 🛠️ Improved: Check if row exists, if not, create it // field is not user input — it comes from the switch above.
var count int64 formatted := fmt.Sprintf(`CONCAT(tenantid, '-',
r.db.Table("ordersequences").Where("tenantid = ?", tid).Count(&count) CASE WHEN subprefix IS NULL OR CAST(subprefix AS TEXT) IN ('0', '0.0', '')
THEN '' ELSE CAST(subprefix AS TEXT) END,
%s)`, field)
if count == 0 { var seq string
newSeq := map[string]interface{}{ err := tx.Raw(fmt.Sprintf(`
"tenantid": tid, UPDATE ordersequences
field: 1, SET %s = COALESCE(%s, (SELECT COUNT(*) FROM orders WHERE tenantid = ?)) + 1,
"subprefix": nil, // Use NULL so CONCAT ignores it or set default here updated = NOW()
} WHERE sequenceid = (SELECT MIN(sequenceid) FROM ordersequences WHERE tenantid = ?)
return r.db.Table("ordersequences").Create(&newSeq).Error RETURNING %s`, field, field, formatted), tid, tid).Scan(&seq).Error
if err != nil {
return "", err
}
if seq != "" {
return seq, nil
} }
// If it exists, perform the update // No counter row for this tenant yet — create one, seeded past whatever
return r.db.Table("ordersequences"). // their existing orders already used.
Where("tenantid = ?", tid). err = tx.Raw(fmt.Sprintf(`
Update(field, gorm.Expr(fmt.Sprintf("%s + 1", field))).Error INSERT INTO ordersequences (tenantid, %s, created, updated)
VALUES (?, (SELECT COUNT(*) FROM orders WHERE tenantid = ?) + 1, NOW(), NOW())
RETURNING %s`, field, formatted), tid, tid).Scan(&seq).Error
if err != nil {
return "", err
}
if seq == "" {
return "", fmt.Errorf("could not allocate %s sequence for tenant %d", prefix, tid)
}
return seq, nil
} }
func (r *orderRepository) CreateOrder(data models.Orders) (models.Orders, error) { func (r *orderRepository) CreateOrder(data models.Orders) (models.Orders, error) {
@@ -1244,7 +1255,15 @@ func (r *orderRepository) CreateOrder(data models.Orders) (models.Orders, error)
} }
// 🛠️ Step 2: Create Order Header // 🛠️ Step 2: Create Order Header
data.Orderid = r.getSequenceno(data.Tenantid, "ORD") // Claimed inside tx so the row lock on the counter holds until commit:
// concurrent orders queue for it instead of reading the same number, and a
// rollback releases it rather than burning it.
orderid, err := nextSequenceNo(tx, data.Tenantid, "ORD")
if err != nil {
tx.Rollback()
return models.Orders{}, fmt.Errorf("failed to allocate order number: %w", err)
}
data.Orderid = orderid
if err := tx.Create(&data).Error; err != nil { if err := tx.Create(&data).Error; err != nil {
tx.Rollback() tx.Rollback()
@@ -1290,10 +1309,6 @@ func (r *orderRepository) CreateOrder(data models.Orders) (models.Orders, error)
syncProductLocationStatus(tx, data.Tenantid, itemLocID, item.Productid) syncProductLocationStatus(tx, data.Tenantid, itemLocID, item.Productid)
} }
if err := r.updateSeqno(data.Tenantid, "ORD"); err != nil {
log.Println("updateSeqno error:", err)
}
if err := tx.Commit().Error; err != nil { if err := tx.Commit().Error; err != nil {
return models.Orders{}, err return models.Orders{}, err
} }