Retries

With Speakeasy, you can generate SDKs that will automatically retry requests that fail due to network errors or any configured HTTP status code. You can enable retries globally for all requests or on a per-request basis by using the x-speakeasy-retries extension in your OpenAPI document. The SDK end user can then override the default configuration by passing in a RetryConfig object to the operation at runtime.

Global Retries


openapi: 3.0.3
info:
title: Swagger Petstore - OpenAPI 3.0
version: 1.0.11
servers:
- url: https://petstore3.swagger.io/api/v3
x-speakeasy-retries:
strategy: backoff
backoff:
initialInterval: 500 # 500 milliseconds
maxInterval: 60000 # 60 seconds
maxElapsedTime: 3600000 # 5 minutes
exponent: 1.5
statusCodes:
- 5XX
retryConnectionErrors: true

If you add the x-speakeasy-retries extension to the root of the OpenAPI document, the SDK Generator will generate a global retry configuration that will be used for all requests made by the SDK. The x-speakeasy-retries extension supports the following properties:

PropertyTypeDescriptionRequired
strategystringThe retry strategy to use. Only backoff is currently supported.Yes
backoffobjectThe configuration of the backoff strategy, if chosen.No
backoff.initialIntervalintegerThe initial interval to use for the backoff strategy (in milliseconds).No
backoff.maxIntervalintegerThe maximum interval between retries (in milliseconds).No
backoff.maxElapsedTimeintegerThe maximum elapsed time to retry for (in milliseconds).No
backoff.exponentnumberThe exponent used to increase the interval between retries.No
statusCodesarrayThe HTTP status codes to retry on.Yes
retryConnectionErrorsbooleanWhether to retry any connection errors.No

The statusCodes property supports passing a list of distinct HTTP status codes or a wildcard range to retry requests on. For example, 5XX will retry requests on all status codes between 500 (inclusive) and 600 (exclusive).

Set the retryConnectionErrors property to true to configure retrying requests that fail due to network errors like DNS resolution errors or connection refused errors.

The defaults for the backoff strategy are:

  • initialInterval: 500
  • maxInterval: 60000
  • maxElapsedTime: 3600000
  • exponent: 1.5

Per-request Retries

Add the x-speakeasy-retries extension to any operation to override the global retry configuration for that operation, or to configure retries for the operation if retries aren't defined globally. For example, you might want to configure retries for an operation on a different set of HTTP status codes or specify different intervals between retries.


openapi: 3.0.3
info:
title: Swagger Petstore - OpenAPI 3.0
version: 1.0.11
servers:
- url: https://petstore3.swagger.io/api/v3
paths:
/pet/findByStatus:
get:
operationId: findPetsByStatus
x-speakeasy-retries:
strategy: backoff
backoff:
initialInterval: 500 # 500 milliseconds
maxInterval: 60000 # 60 seconds
maxElapsedTime: 3600000 # 5 minutes
exponent: 1.5
statusCodes:
- 408
- 500
- 502
- 503
retryConnectionErrors: true
parameters:
- name: status
in: query
description: Status values that need to be considered for filter
required: false
explode: true
schema:
type: string
default: available
enum:
- available
- pending
- sold
responses:
'200':
description: successful operation
content:
application/json:
schema:
type: array
items:
$ref: '#/components/schemas/Pet'
application/xml:
schema:
type: array
items:
$ref: '#/components/schemas/Pet'
'400':
description: Invalid status value

Per-request retries are configured the same way as global retries.

SDK Usage Override

Users of the SDK can override the retry strategy globally or at the request level by providing a utils.RetryConfig object. This object supports most of the same properties as the x-speakeasy-retries extension, except for the status codes to retry on.

Global

To override the retry strategy globally, initialize the SDK object with the appropriate options parameter. In all cases, if no override is provided, the retry configuration provided in the OpenAPI document will be used.

In this example, we use the RetryConfig object to disable retries globally:


const sdk = new SDK({retryConfig: {strategy: "none"}});

Request-Level Override

Any endpoints that support retries allow retry configuration to be overridden. In Go, override request-level retry configuration with operations.WithRetries. In Python and TypeScript, the optional retries is accepted.


const sdk = new SDK({});
await sdk.findPetsByStatus(request, {
strategy: "backoff",
backoff: {
initialInterval: 100,
maxInterval: 10000,
exponent: 1.1,
maxElapsedTime: 60000,
},
retryConnectionErrors: false,
});

Idempotency

If your endpoint has a defined idempotency header, the request will be retried with the exact idempotency key that was provided for the initial request.


paths:
/pet:
put:
operationId: putPet
x-speakeasy-retries:
strategy: backoff
backoff:
initialInterval: 500
maxInterval: 60000
maxElapsedTime: 3600000
exponent: 1.5
statusCodes:
- 5XX
retryConnectionErrors: false
parameters:
- name: Idempotency-Key
schema:
type: string
in: header
requestBody:
content:
application/json:
schema:
$ref: '#/components/schemas/Pet'
responses:
'200':
description: successful operation
content:
application/json:
schema:
$ref: '#/components/schemas/Pet'
default:
description: Error

Respecting Retry-After

If your API returns an HTTP standard retry-after header, the SDK will respect that value as the retry interval as long as the time is in the future and below the max elapsed retry time. You don't need to change any configurations to enable this; simply return a retry-after header from your API. We currently support this feature in TypeScript and will add support for additional languages in the future.


responses:
'429':
description: Too Many Requests
headers:
Retry-After:
description: The date and time after which the client can retry the request.
schema:
type: string
format: date-time
example: "Wed, 21 Oct 2023 07:28:00 GMT"


responses:
'429':
description: Too Many Requests
headers:
Retry-After:
description: The number of seconds to wait before retrying the request.
schema:
type: integer
example: 120