Bulk Writes: How to Insert 1 Million Docs Fast

Welcome to Day 2! 📦

A common mistake new developers make is running database operations in a loop.

Bad Code:

// 10,000 requests to the DB! 🐢
for (const user of users) {
  await db.collection('users').insertOne(user);
}

Even with Promise.all(), you are flooding the network with 10,000 separate TCP packets.

1. Enter bulkWrite()

MongoDB allows you to send one massive batch of operations in a single network request.

const bulkOps = users.map(user => ({
  insertOne: {
    document: user
  }
}));

// 1 request! 🚀
await db.collection('users').bulkWrite(bulkOps);

2. It’s not just for Inserts

You can mix and match operations!

db.collection('products').bulkWrite([
  { insertOne: { document: { name: "New Product" } } },
  { updateOne: { filter: { id: 1 }, update: { $set: { price: 99 } } } },
  { deleteOne: { filter: { status: "defective" } } }
]);

3. Ordered vs Unordered

By default, bulkWrite is Ordered (ordered: true).

• Operations execute in sequence (1, then 2, then 3).
• If Op #5 fails, it stops. Operations 6-10 are NOT executed.

If you don’t care about order (e.g., importing independent log entries), use Unordered.

• ordered: false
• MongoDB creates parallel threads to execute them.
• If Op #5 fails, it continues with the rest. Faster!

await db.collection('logs').bulkWrite(ops, { ordered: false });

🧠 Daily Challenge

1. Create a script that generates 10,000 dummy products.
2. Measure the time it takes to insert them using a for...of loop with await.
3. Measure the time it takes using bulkWrite.
4. Be amazed. (Usually 10x-50x faster).

See you on Day 3 for Change Streams! (Real-time magic) ✨

Next Step

Next: Change Streams →