A course · driven from your phone

Message Queues, SQS-first

Built around one goal: understand the SQS in your ZenRows project well enough to read it and know, not guess. Every lesson grounds in scraping jobs. Other brokers show up only to sharpen why SQS chose what it did.

Lessons

• LESSON 1 The SQS mental model & visibility timeout A message is borrowed, never just read. The lifecycle, the borrow clock, and the two failure modes that explain everything else.
• LESSON 2 At-least-once & idempotency: making duplicates boring Guard vs converge, the ack contract, and a tour of your real task.go guards. Why conveyor skips FIFO and pays for correctness instead.
• LESSON 3 Long vs short polling: the receive loop The empty-receive storm and why both of conveyor's loops long-poll. The many-queues wrinkle in the dispatcher pool.
• LESSON 4 DLQs, redrive & maxReceiveCount Receive count vs failure count, and the COR-314 story: why your per-run queues use maxReceiveCount 20, not 5.
• LESSON 5 · the payload Visibility timeout as a control plane The advanced patterns you came for: backpressure by bouncing, and an 8-hour retry timer built from a 15-minute primitive.
• LESSON 6 · a real incident How conveyor "froze" — with a live simulator The capstone of 3–5. Run the receive-amplification sim, watch the freeze, then flip the pre-poll gate and watch it stop.
• LESSON 7 · NEXT The dynamic queue-per-run architecture Coming next — why conveyor creates a whole SQS queue per run, how workers discover them, and how they're cleaned up.

Reference

Living documents you'll come back to: Glossary · Mission · Trusted resources

I'm your teacher here — ask me anything that's unclear, and tell me when you want the next lesson.