Bulkhead Pattern

1. Introduction

In microservices architecture, a service usually communicates with multiple downstream services. If one dependency becomes slow or unavailable, it can consume all available threads and block the system.

The Bulkhead Pattern prevents this by isolating resources for different dependencies so that failure in one part does not affect the entire system.

2. What Problem Bulkhead Solves

Example architecture:

Order Service
   ├── Payment Service
   ├── Inventory Service
   └── Notification Service

If the Payment Service becomes slow, without Bulkhead:

Threads get blocked waiting for Payment Service
Inventory service calls fail
Notification service calls fail
Entire system becomes slow or unavailable

This is called resource exhaustion.

3. What Bulkhead Does

Bulkhead isolates resources for each dependency.

Example configuration:

Payment Service → 20 threads
Inventory Service → 10 threads
Notification Service → 10 threads

If Payment service becomes slow:

Only its 20 threads are affected
Other services continue to function normally

4. Why It Is Called Bulkhead

The concept comes from ship design.

Ships are divided into compartments called bulkheads.

If one compartment floods:

Other compartments remain dry
The ship continues to float

Similarly in microservices:

Each dependency is isolated with its own resource allocation.

5. Types of Bulkhead in Resilience4j

Resilience4j supports two types of Bulkhead.

Type	Description
Semaphore Bulkhead	Limits concurrent calls
ThreadPool Bulkhead	Uses separate thread pools

6. Semaphore Bulkhead

Semaphore Bulkhead limits the number of concurrent requests.

Example:

maxConcurrentCalls = 20

If 50 requests arrive:

20 requests allowed
30 requests rejected

This prevents system overload.

7. ThreadPool Bulkhead

ThreadPool Bulkhead creates separate thread pools for each dependency.

Example:

Payment Service → ThreadPool (20 threads)
Inventory Service → ThreadPool (10 threads)

If Payment service becomes slow:

Only Payment thread pool is blocked
Inventory and other services continue normally

8. Real Production Example

Example:

Transaction Service → Payment Gateway

If Payment gateway becomes slow:

Without Bulkhead:

All transaction threads blocked
System becomes unavailable

With Bulkhead:

Only payment calls are blocked
Other operations continue working

9. Example Configuration (application.yml)

resilience4j:
  bulkhead:
    instances:
      paymentService:
        maxConcurrentCalls: 20
        maxWaitDuration: 0

Explanation:

Property	Meaning
maxConcurrentCalls	Maximum concurrent calls allowed
maxWaitDuration	How long a request waits if limit reached

10. Using Bulkhead with Other Resilience Patterns

Bulkhead is often combined with other patterns.

Typical flow:

Client Request -> Bulkhead -> Retry -> TimeLimiter -> Circuit Breaker -> Service

Each layer protects the system from a different type of failure.

11. When Bulkhead Is Important

Bulkhead is especially important when:

A service calls multiple downstream services
External services may become slow
Thread pools are limited
High traffic systems (banking, fintech, e-commerce)

12. Summary

Bulkhead is a resilience pattern that isolates system resources such as threads or concurrent calls so that failure or slowness in one dependency does not affect the entire system.

Key benefit:

Prevent resource exhaustion and cascading failures.