Idempotency keys for safe retries in Rest API - Step-by-Step Execution

Concept Flow - Idempotency keys for safe retries

Client sends request with Idempotency Key

↓

Server checks if Key seen before?

Yes→Return stored response

No↓

Process request normally

↓

Store response with Idempotency Key

↓

Send response to client

The client sends a request with a unique key. The server checks if it already processed this key. If yes, it returns the saved response. If no, it processes and saves the response for future retries.

Execution Sample

Rest API

POST /payment
Headers: Idempotency-Key: abc123

Server:
if key in store:
  return stored_response
else:
  process payment
  store response
  return response

This code shows how a server handles a payment request with an idempotency key to avoid duplicate processing.

Execution Table

Step	Action	Idempotency Key	Key Seen?	Server Action	Response Sent
1	Receive request	abc123	No	Process payment, store response	Payment processed
2	Receive retry request	abc123	Yes	Return stored response	Payment processed
3	Receive new request	xyz789	No	Process payment, store response	Payment processed
4	Receive retry request	xyz789	Yes	Return stored response	Payment processed
5	Receive request	abc123	Yes	Return stored response	Payment processed

💡 Requests with seen keys return stored response to avoid duplicate processing.

Variable Tracker

Variable	Start	After Step 1	After Step 2	After Step 3	After Step 4	After Step 5
Idempotency Store	{}	{"abc123": "Payment processed"}	{"abc123": "Payment processed"}	{"abc123": "Payment processed", "xyz789": "Payment processed"}	{"abc123": "Payment processed", "xyz789": "Payment processed"}	{"abc123": "Payment processed", "xyz789": "Payment processed"}

Key Moments - 3 Insights

Why does the server return the stored response instead of processing the request again?

Because the idempotency key was seen before (see Step 2 in execution_table), the server avoids duplicate processing by returning the saved response.

What happens if the client sends a request with a new idempotency key?

The server processes the request normally and stores the response with the new key (see Step 3 in execution_table).

Can the client safely retry the same request multiple times?

Yes, retries with the same idempotency key return the stored response without reprocessing (see Steps 2, 4, and 5).

Visual Quiz - 3 Questions

Look at the execution_table, what does the server do at Step 2 when it receives the retry request with key 'abc123'?

A Ignore the request

B Process payment again and store new response

C Return stored response without processing

D Generate a new idempotency key

💡 Hint

Check the 'Server Action' column at Step 2 in the execution_table.

At which step does the server first store the response for key 'xyz789'?

A Step 3

B Step 2

C Step 1

D Step 4

💡 Hint

Look at the 'Idempotency Key' and 'Key Seen?' columns in execution_table.

If the client sends a request with a new key 'newkey', what will the server do according to the flow?

A Return an error because key is unknown

B Process the request and store the response

C Return the last stored response

D Ignore the idempotency key and process normally

💡 Hint

Refer to the concept_flow where new keys lead to processing and storing the response.

Concept Snapshot

Idempotency keys are unique IDs sent with requests.
Server checks if key was seen before.
If yes, returns stored response to avoid duplicates.
If no, processes request and saves response.
This ensures safe retries without side effects.

Full Transcript

Idempotency keys help servers handle repeated requests safely. When a client sends a request with a unique key, the server checks if it has processed that key before. If it has, the server returns the saved response instead of processing again. If not, it processes the request, stores the response with the key, and returns it. This way, clients can retry requests without causing duplicate actions. The execution table shows steps where requests with keys are processed or returned from storage. The variable tracker shows how the server's store of keys and responses grows. Key moments clarify why stored responses are returned and how new keys are handled. The quiz tests understanding of these steps and behaviors.

Practice

1. What is the main purpose of using an idempotency key in REST APIs?

easy

A. To speed up the server response time

B. To prevent duplicate processing of the same request

C. To encrypt the request data

D. To authenticate the user

Solution

1. Step 1: Understand idempotency key concept

   An idempotency key is a unique identifier sent with a request to ensure the server processes it only once.

2. Step 2: Identify the main purpose

   This prevents duplicate actions like double payments or repeated orders if the client retries the request.

3. Final Answer:

   To prevent duplicate processing of the same request -> Option B

4. Quick Check:

   Idempotency key = prevent duplicates [OK]

Hint: Idempotency keys stop repeated actions safely [OK]

Common Mistakes:

• Confusing idempotency key with authentication token
• Thinking it speeds up response time
• Assuming it encrypts data

2. Which HTTP header is commonly used to send an idempotency key in a REST API request?

easy

A. Idempotency-Key

B. Cache-Control

C. Content-Type

D. Authorization

Solution

1. Step 1: Recall standard header for idempotency

   The standard practice is to send the unique idempotency key in the Idempotency-Key HTTP header.

2. Step 2: Differentiate from other headers

   Authorization is for credentials, Content-Type defines data format, and Cache-Control manages caching.

3. Final Answer:

   Idempotency-Key -> Option A

4. Quick Check:

   Idempotency key header = Idempotency-Key [OK]

Hint: Idempotency keys go in Idempotency-Key header [OK]

Common Mistakes:

• Using Authorization header for idempotency key
• Confusing Content-Type with idempotency key
• Sending key in request body instead of header

3. Consider this simplified server logic for handling POST requests with idempotency keys:

stored_keys = {}

def handle_request(idempotency_key, data):
    if idempotency_key in stored_keys:
        return stored_keys[idempotency_key]
    response = process_data(data)
    stored_keys[idempotency_key] = response
    return response

# Client sends:
print(handle_request('abc123', 'order1'))
print(handle_request('abc123', 'order1'))
print(handle_request('xyz789', 'order2'))

What will be the output?

medium

A. None, None, None

B. process_data output for order1, None, process_data output for order2

C. process_data output for order1, process_data output for order2, process_data output for order2

D. process_data output for order1, process_data output for order1, process_data output for order2

Solution

1. Step 1: Analyze first call with key 'abc123'

   Key not in stored_keys, so process_data('order1') runs, result saved and returned.

2. Step 2: Analyze second call with same key 'abc123'

   Key found, stored response returned without reprocessing.

3. Step 3: Analyze third call with new key 'xyz789'

   Key not found, process_data('order2') runs, result saved and returned.

4. Final Answer:

   process_data output for order1, process_data output for order1, process_data output for order2 -> Option D

5. Quick Check:

   Repeated key returns cached response [OK]

Hint: Repeated keys return cached response, new keys process data [OK]

Common Mistakes:

• Assuming second call reprocesses data
• Confusing keys and responses
• Thinking None is returned on repeats

4. A developer wrote this code snippet to handle idempotent POST requests:

cache = {}
def post_request(key, data):
    if key not in cache:
        process(data)
    return cache[key]

What is the bug in this code?

medium

A. It never stores the processed result in cache

B. It processes data twice for the same key

C. It raises an error if key is missing

D. It uses the wrong HTTP method

Solution

1. Step 1: Check caching logic

   The code checks if key not in cache, then calls process(data), but never saves the result in cache.

2. Step 2: Identify consequence

   Since cache[key] is never set, returning cache[key] will cause a KeyError or return None incorrectly.

3. Final Answer:

   It never stores the processed result in cache -> Option A

4. Quick Check:

   Missing cache storage causes errors [OK]

Hint: Always save processed result in cache before returning [OK]

Common Mistakes:

• Assuming process(data) stores result automatically
• Ignoring missing cache assignment
• Thinking HTTP method affects caching here

5. You are designing a payment API that uses idempotency keys to avoid duplicate charges. Which approach best ensures safe retries without charging twice if the client resends the same request due to a timeout?

hard

A. Ignore the idempotency key and process every request independently

B. Reject any request with an idempotency key that was seen before

C. Store the payment result linked to the idempotency key and return it on repeated requests

D. Process the payment again but log a warning for duplicate keys

Solution

1. Step 1: Understand safe retry requirements

   Safe retries mean the client can resend the same request without causing duplicate charges.

2. Step 2: Evaluate options

   Store the payment result linked to the idempotency key and return it on repeated requests stores the payment result linked to the key and returns it on repeats, preventing double charges. Ignore the idempotency key and process every request independently processes every request, risking duplicates. Reject any request with an idempotency key that was seen before rejects repeats, which may cause client errors. Process the payment again but log a warning for duplicate keys processes again, causing duplicates.

3. Final Answer:

   Store the payment result linked to the idempotency key and return it on repeated requests -> Option C

4. Quick Check:

   Store and reuse result for idempotency [OK]

Hint: Save and reuse payment result by key to avoid duplicates [OK]

Common Mistakes:

• Rejecting repeated keys causing client errors
• Processing duplicates causing double charges
• Logging warnings but still charging twice