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:

Property	Type	Description	Required
`strategy`	`string`	The retry strategy to use. Only `backoff` is currently supported.	Yes
`backoff`	`object`	The configuration of the backoff strategy, if chosen.	No
`backoff.initialInterval`	`integer`	The initial interval to use for the backoff strategy (in milliseconds).	No
`backoff.maxInterval`	`integer`	The maximum interval between retries (in milliseconds).	No
`backoff.maxElapsedTime`	`integer`	The maximum elapsed time to retry for (in milliseconds).	No
`backoff.exponent`	`number`	The exponent used to increase the interval between retries.	No
`statusCodes`	`array`	The HTTP status codes to retry on.	Yes
`retryConnectionErrors`	`boolean`	Whether 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"}});

s = SDK(retry_config=RetryConfig(
  strategy='none',
  backoff=None,
  retry_connection_errors=False
))

s := sdk.New(sdk.WithRetryConfig(utils.RetryConfig{Strategy: "none"}))

SDK s = SDK.builder()
    .retryConfig(RetryConfig.noRetries())
    .build();

var sdk = new SDK(
    retryConfig: new RetryConfig(
        strategy: RetryConfig.RetryStrategy.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,
});

s = sdk()
res = s.find_pets_by_status(request, RetryConfig(strategy='backoff',
  backoff=BackoffStrategy(
    initial_interval=100,
    max_interval=10000,
    exponent=1.1,
    max_elapsed_time=60000),
    retry_connection_errors=False
))

s := sdk.New()
s.FindPetsByStatus(&operations.FindPetsByStatusRequest{},
  operations.WithRetries(utils.RetryConfig{
    Strategy: "backoff",
    Backoff: &utils.BackoffStrategy{
      InitialInterval: 100,
      MaxInterval: 10000,
      MaxElapsedTime: 60000,
      Exponent: 1.1,
    },
    RetryConnectionErrors: false,
  },
)

import <namepace>.SDK;
import <namepace>.models.operations.FindPetsByStatusResponse

SDK s = SDK.builder().build();
FindPetsByStatusResponse res = sdk.findPetsByStatus()
    .retryConfig(RetryConfig.builder()
                    .backoff(BackoffStrategy.builder()
                                .initialInterval(100L, TimeUnit.MILLISECONDS)
                                .maxInterval(10000L, TimeUnit.MILLISECONDS)
                                .maxElapsedTime(60000L, TimeUnit.MILLISECONDS)
                                .exponent(1.1f)
                                .retryConnectError(false)
                                .build())
                    .build())
    .call();

var sdk = new SDK();

var res = await sdk.FindPetsByStatusAsync(retryConfig: new RetryConfig(
    strategy: RetryConfig.RetryStrategy.BACKOFF,
    backoff: new BackoffStrategy(
        initialIntervalMs: 100L,
        maxIntervalMs: 10000L,
        maxElapsedTimeMs: 60000L,
        exponent: 1.1
    ),
    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