Course contentsShow
← System Design
Lesson 1080 of 1,91927. Circuit Breakers and BulkheadsFree lesson

Queue Saturation and Backpressure Loss

When message queues fill up during cascading failures, dropping messages or blocking upstream services until full system paralysis.

Queue Saturation and Backpressure Loss

What you'll learn: How overflowing message queues during cascading failures can force message drops or freeze upstream services, potentially paralyzing the entire system.


When a downstream service slows or fails, messages accumulate in queues faster than they can be processed. Once those queues fill to capacity, you face two equally painful choices: drop incoming messages (losing data) or block upstream services (causing congestion to propagate backward).

How Queue Saturation Happens

Imagine a restaurant kitchen with a finite ticket rail. If the chef falls behind, new orders pile up. Once the rail is full, you either refuse new diners (dropping requests) or make the waitstaff stand frozen at the window, unable to serve other tables (blocking upstream).

In distributed systems, this scenario unfolds when:

  1. A dependency becomes slow or unresponsive
  2. Clients keep sending requests, filling the queue
  3. The queue reaches maximum capacity
  4. Backpressure kicks in: the system must either reject new work or halt upstream producers

The Paralysis Dilemma

Dropping messages means data loss—unacceptable for critical operations like payments or orders.

Blocking producers creates a domino effect: if Service A blocks waiting for queue space, its own queues fill, forcing its callers to block. This congestion ripples upstream through the entire call chain, eventually freezing the system at the entry point.

Neither option is ideal during a cascading failure—which is why you need bulkheads to isolate failure domains and circuit breakers to stop flooding failed dependencies before queues saturate.


Key Takeaway: Queue saturation during failures forces a choice between data loss (dropping messages) and system paralysis (blocking upstream), making early failure detection and isolation critical.