Retry Using Polly — Complete Guide

Introduction

Retry Using Polly — Complete Guide is essential for .NET architects building ShopNest Cloud-Native Enterprise Platform — Toolliyo's 120-article microservices master path covering RabbitMQ, Saga, Kubernetes, API Gateway, observability, ASP.NET Core integration, and senior interview preparation. Every article includes minimum 2 detailed production real-world examples (Flipkart, banking, Swiggy, SaaS) in different business domains.

In Indian delivery projects (TCS, Infosys, Wipro), interviewers expect retry using polly with real Flipkart-scale e-commerce, HDFC-style banking, Swiggy delivery, or SaaS multi-tenant examples — not toy animal demos. This article delivers two mandatory enterprise examples on Identity Service.

After this article you will

• Explain Retry Using Polly in plain English and in distributed systems and cloud-native terms
• Implement retry using polly in ShopNest Cloud-Native Enterprise Platform (Identity Service)
• Compare the wrong approach vs the production-ready enterprise approach
• Answer fresher, mid-level, and senior microservices and distributed systems interview questions confidently
• Connect this lesson to Article 106 and the 120-article Microservices roadmap

Prerequisites

• Software: .NET 8 SDK, VS 2022 or VS Code, SQL Server Express / LocalDB
• Knowledge: C# basics
• Previous: Article 104 — API Versioning — Complete Guide
• Time: 28 min reading + 30–45 min hands-on

Concept deep-dive

Level 1 — Analogy

Circuit breaker is a fuse on downstream calls — when Payment API fails repeatedly, stop hammering it and return graceful fallback.

Level 2 — Technical

Retry Using Polly integrates with the LINQ query layer: write queries against IEnumerable or IQueryable, understand deferred execution, project to DTOs for ShopNest Cloud-Native reports. On ShopNest Cloud-Native this powers Identity Service without coupling UI to database internals.

Level 3 — Architecture

[Browser] → [HTTPS/Kestrel] → [Middleware Pipeline]
  → [Routing] → [Controller Action] → [Service Layer]
  → [EF Core / Identity] → [Razor View Engine] → [HTML Response]

Project structure

ShopNest Cloud-Native/
├── ShopNest.Cloud/
├── src/
│   ├── Gateway/              ← YARP API Gateway (JWT, rate limit)
│   ├── Services/
│   │   ├── Identity.Api/
│   │   ├── User.Api/
│   │   ├── Product.Api/
│   │   ├── Order.Api/
│   │   ├── Payment.Api/
│   │   ├── Inventory.Api/
│   │   ├── Notification.Api/
│   │   └── Analytics.Api/
│   ├── BuildingBlocks/       ← EventBus, Outbox, Polly policies
│   └── docker-compose.yml
├── k8s/                      ← Helm charts per service
└── .github/workflows/        ← CI/CD per service

Step-by-Step Implementation — ShopNest (Identity Service)

Follow the prompt template: create project → core classes → interfaces → pattern implementation → client code → run → enterprise refactor.

Step 1 — The wrong way

// ❌ BAD — fat controller, no ViewModel, sync DB call
public IActionResult Index()
{
    return _context.Products.Find(id); // sync, exposes entity, no auth
}

Step 2 — The right way

// ✅ CORRECT — Retry Using Polly on ShopNest (Identity Service)
var results = await _context.Products
    .Where(p => p.IsPublished && p.CategoryId == categoryId)
    .OrderBy(p => p.Name)
    .Select(p => new ProductReportDto { Id = p.Id, Name = p.Name, Revenue = p.Orders.Sum(o => o.Total) })
    .ToListAsync(ct);

Step 3 — Apply Retry Using Polly

// Retry Using Polly — ShopNest Cloud-Native (Identity Service)
builder.Services.AddScoped<IRetryUsingPollyService, RetryUsingPollyService>();

docker compose up --build
# Verify Retry Using Polly — check RabbitMQ management UI and kubectl get pods and integration tests pass

Distributed system challenges — Retry Using Polly

Production microservices fail in predictable ways. ShopNest engineers plan for these explicitly:

• Network failures — Payment service timeout must not hang Order API thread pool; use Polly timeout + async messaging
• Eventual consistency — Inventory may lag 200ms after order; UI shows "confirming stock" not silent wrong state
• Duplicate messages — RabbitMQ redelivery requires idempotent consumers (Idempotency-Key, unique constraints)
• Retry storms — Exponential backoff + jitter; never retry 503s infinitely without circuit breaker
• Cascade failures — Bulkhead isolates Notification failures from blocking Payment path

Real-World Example 1 — Uber-Style Ride Booking Platform

MANDATORY production scenario (Uber / Ola (India)): how Retry Using Polly applies in ShopNest Cloud-Native Identity Service.

Business problem

Matching riders to drivers requires low-latency gRPC between Location, Pricing, and Dispatch services. Network partitions must not double-assign drivers — Saga orchestration coordinates RideRequested → DriverAssigned → PaymentCaptured with compensating CancelRide.

Why Retry Using Polly matters here

Indian enterprise teams at TCS, Infosys, Wipro, and product companies like Uber / Ola face this exact distributed systems challenge. Retry Using Polly is not academic — it directly affects uptime during peak load, deployment frequency, and incident recovery.

Architecture diagram

[Rider App] → [BFF] → [Ride.Api]
  → gRPC → [Dispatch.Service] (driver pool in Redis)
  → [Pricing.Service] surge multiplier
  → [Payment.Service] pre-auth hold
Choreography fallback if orchestrator unavailable; OpenTelemetry traces across all hops.

Production implementation

// gRPC — ShopNest.Dispatch.Grpc/Dispatch.proto
service DriverDispatch {
  rpc FindNearestDriver (FindDriverRequest) returns (DriverAssignment);
}

// Polly on HttpClientFactory — ShopNest.Ride.Api/Program.cs
builder.Services.AddHttpClient<IPricingClient, PricingClient>()
    .AddPolicyHandler(Policy.TimeoutAsync<HttpResponseMessage>(TimeSpan.FromSeconds(2)))
    .AddPolicyHandler(GetRetryPolicy())
    .AddPolicyHandler(GetCircuitBreakerPolicy());

Production metrics and outcome

Driver match p99 180ms in Bangalore pilot; circuit breaker prevented cascade when Pricing service degraded.

Distributed system lessons

• Design for failure — network partitions and partial outages are normal at scale
• Prefer async messaging for cross-service workflows; sync only when latency requires it
• Instrument with OpenTelemetry from day one — you cannot debug what you cannot trace
• Run load tests before Big Billion Day / salary-day / lunch-rush peaks

Real-World Example 2 — HDFC-Style Core Banking Transfers

MANDATORY production scenario (Indian Banking (NEFT/IMPS)): how Retry Using Polly applies in ShopNest Cloud-Native Identity Service.

Business problem

Fund transfers must be auditable, idempotent, and eventually consistent across Account, Ledger, Fraud, and Notification services. A shared database caused lock contention — 200ms p99 became 4s under salary-day load.

Why Retry Using Polly matters here

Indian enterprise teams at TCS, Infosys, Wipro, and product companies like Indian Banking face this exact distributed systems challenge. Retry Using Polly is not academic — it directly affects uptime during peak load, deployment frequency, and incident recovery.

Architecture diagram

[Mobile Banking] → [API Gateway + mTLS]
  → [Transfer.Api] → Outbox table → [Ledger.Worker]
  → [Fraud.Api] (sync gRPC, 200ms timeout)
  → [Notification.Api] via Kafka topic transfer.completed
Each service owns its DB; Saga compensates if fraud blocks after debit.

Production implementation

// ShopNest.Payment.Api — Idempotent transfer endpoint
[HttpPost("transfers")]
public async Task<IActionResult> Transfer([FromBody] TransferRequest req,
    [FromHeader(Name = "Idempotency-Key")] string idempotencyKey)
{
    var existing = await _cache.GetAsync<TransferResult>(idempotencyKey);
    if (existing != null) return Ok(existing);

    var cmd = new InitiateTransferCommand(req.FromAccount, req.ToAccount, req.Amount, idempotencyKey);
    var result = await _mediator.Send(cmd);
    await _cache.SetAsync(idempotencyKey, result, TimeSpan.FromHours(24));
    return Accepted(result);
}

// Saga compensation on fraud failure
public async Task CompensateAsync(Guid transferId) =>
    await _bus.Publish(new ReverseTransferCommand(transferId));

Production metrics and outcome

Salary-day throughput: 12,000 TPS with 99.99% success; zero duplicate debits after idempotency keys + outbox.

Distributed system lessons

• Design for failure — network partitions and partial outages are normal at scale
• Prefer async messaging for cross-service workflows; sync only when latency requires it
• Instrument with OpenTelemetry from day one — you cannot debug what you cannot trace
• Run load tests before Big Billion Day / salary-day / lunch-rush peaks

Security checklist (every ShopNest service)

Even non-auth articles must follow: HTTPS only, no secrets in appsettings committed to git, validate JWT on gateway, least-privilege DB users per service, and structured audit logs for Payment/Identity operations.

ASP.NET Core microservices integration — Retry Using Polly

Register services in Program.cs, configure MassTransit/RabbitMQ, expose health endpoints for Kubernetes, and use IHttpClientFactory with Polly for sync calls between ShopNest services.

Microservices integration patterns & ASP.NET Core integration

Modern C# 12 implementations use primary constructors, records, and DI. Register pattern abstractions in Program.cs with appropriate lifetimes (Singleton for stateless, Scoped for request-bound, Transient for lightweight factories).

Microservices: Apply Retry Using Polly within bounded contexts — each ShopNest service (Orders, Payments, Inventory) owns its pattern implementation.

Architecture comparison & when NOT to use

Compare Retry Using Polly with alternative microservices approaches. Avoid overengineering — if a simple function or DI registration suffices, do not force a pattern. Senior architects value judgment over pattern count.

Common errors & fixes

🔴 Mistake 1: Fat controllers with EF Core queries inline
✅ Fix: Move data access to services/repositories; keep controllers thin.

🔴 Mistake 2: Calling .ToList() too early materializing millions of rows into memory
✅ Fix: Defer execution — build IQueryable pipeline, then ToListAsync() once at the end.

🔴 Mistake 3: Filtering in memory after .ToList() instead of in the database query
✅ Fix: Keep filters in IQueryable, use Select projection, paginate with Skip/Take before materialization.

🔴 Mistake 4: Hard-coding connection strings in controllers
✅ Fix: Use appsettings.json + User Secrets locally; Azure Key Vault in production.

Best practices

• 🟢 Use async/await end-to-end for database and I/O calls
• 🟢 Register DbContext as Scoped; avoid capturing it in singletons
• 🟡 Use IQueryable until the last moment; avoid multiple enumeration; project with Select before ToList
• 🟡 Prefer method syntax for complex chains; use query syntax for joins when readability wins
• 🔴 Log structured data with Serilog — include OrderId, UserId, not passwords
• 🔴 Use HTTPS, secure cookies, and authorization policies in production

Interview questions

Fresher level

Q1: What is Retry Using Polly in ASP.NET Core MVC?
A: Retry Using Polly is a core MVC capability used in ShopNest Cloud-Native for Identity Service. Explain in one sentence, then describe controller/view/service placement.

Q2: How would you implement Retry Using Polly on a TCS-style delivery project?
A: Deferred execution, IQueryable pipelines, Select projection, Skip/Take pagination, and SQL logging in development.

Q3: IEnumerable vs IQueryable — when to use which?
A: IEnumerable for in-memory collections; IQueryable for EF Core database queries that translate to SQL.

Mid / senior level

Q4: Explain LINQ deferred execution and query translation briefly.
A: LINQ → Expression Tree → IQueryProvider → SQL (EF) or Iterator (in-memory) → Results.

Q5: Common production mistake with this topic?
A: Skipping validation, exposing secrets in Git, or untested edge cases (null model, unauthorized user).

Q6: .NET LINQ vs SQL — when to push logic to database?
A: Core is cross-platform, faster, cloud-ready; Framework is maintenance mode on Windows/IIS.

Coding round

Implement Retry Using Polly for ShopNest Identity Service: show interface, concrete class, DI registration, and xUnit test with mock.

public class RetryUsingPollyPatternTests
{
    [Fact]
    public async Task ExecuteAsync_ReturnsSuccess()
    {
        var mock = new Mock();
        mock.Setup(s => s.ExecuteAsync(It.IsAny(), default))
            .ReturnsAsync(Result.Success("test-id"));
        var result = await mock.Object.ExecuteAsync(new Request("test-id"));
        Assert.True(result.IsSuccess);
    }
}

Summary & next steps

• Article 105: Retry Using Polly — Complete Guide
• Module: Module 11: Additional Advanced Topics · Level: ADVANCED
• Applied to ShopNest Cloud-Native — Identity Service

Previous: API Versioning — Complete Guide
Next: CORS in Microservices — Complete Guide

Practice: Add one small feature using today's pattern — commit with feat(microservices): article-105.

FAQ