Resilience4j causes thread leaks

[tsingheng]

spring-cloud-circuitbreaker-resilience4j-2.0.1

1. I found that resilience4j creates a thread pool for every feign method. There are hundreds of feign method in my project. So resilience4j will create thousands of bulkhead thread. org.springframework.cloud.circuitbreaker.resilience4j.Resilience4jBulkheadProvider(line 115)

   private <T> Supplier<CompletionStage<T>> decorateBulkhead(final String id, final Supplier<T> supplier) {
       // I would like id to be the name of @FeignClient
       Resilience4jBulkheadConfigurationBuilder.BulkheadConfiguration configuration = configurations
               .computeIfAbsent(id, defaultConfiguration);
   
       if (bulkheadRegistry.find(id).isPresent() && !threadPoolBulkheadRegistry.find(id).isPresent()) {
           Bulkhead bulkhead = bulkheadRegistry.bulkhead(id, configuration.getBulkheadConfig());
           CompletableFuture<T> asyncCall = CompletableFuture.supplyAsync(supplier);
           return Bulkhead.decorateCompletionStage(bulkhead, () -> asyncCall);
       }
       else {
           ThreadPoolBulkhead threadPoolBulkhead = threadPoolBulkheadRegistry.bulkhead(id,
                   configuration.getThreadPoolBulkheadConfig());
           return threadPoolBulkhead.decorateSupplier(supplier);
       }
   }

2. TimeLimiter creates newSingleThreadScheduledExecutor for every feign invoke. It may lead thread leaks. org.springframework.cloud.circuitbreaker.resilience4j.Resilience4jBulkheadProvider(line 104)

   public <T> T run(String id, Supplier<T> toRun, Function<Throwable, T> fallback, CircuitBreaker circuitBreaker,
           TimeLimiter timeLimiter) {
       Supplier<CompletionStage<T>> bulkheadCall = decorateBulkhead(id, toRun);
       Supplier<CompletionStage<T>> timeLimiterCall = timeLimiter
               .decorateCompletionStage(Executors.newSingleThreadScheduledExecutor(), bulkheadCall);
       Supplier<CompletionStage<T>> circuitBreakerCall = circuitBreaker.decorateCompletionStage(timeLimiterCall);
       try {
           return circuitBreakerCall.get().toCompletableFuture().get();
       }
       catch (Exception e) {
           System.out.println("exception " + e.getMessage());
           return fallback.apply(e);
       }
   }

[spencergibb]

Maybe bulkhead isn't the proper strategy for hundreds of feign methods and it should be moved to a different layer. Also see #113 for the option to default to semaphore.