Configuring AWS SDK Retry Policies via application.properties

[Shanzeee]

Hi everyone,

I’ve been working on configuring the retry behavior for AWS SDK clients in Quarkus applications, and I noticed there’s no built-in way to define retry policies (e.g., RetryMode or maximum retries) via the application.properties file.

Currently, configuring AWS SDK retry policies requires manually creating a custom bean with ClientOverrideConfiguration.

Example for DynamoDbAsyncClient:

DynamoDbAsyncClient.builder()
    .overrideConfiguration(
        ClientOverrideConfiguration.builder()
            .retryPolicy(RetryMode.STANDARD)
            .build()
    )
    .build()

This approach works but feels inconsistent with Quarkus's configuration-driven approach.

Questions:

• Is there an easier way to configure retry policies that I might have missed?
• Are there plans to support properties like this in the future?

Thanks in advance for your help!

[scrocquesel]

Hi,

Does something very simple like this would be enough.

quarkus.dynamodb.advanced.retry-strategy.retry-mode=STANDARD # one of RetryMode enum
quarkus.dynamodb.advanced.retry-strategy.max-attempts=3

[scrocquesel]

@Shanzeee Why did you close the issue ?

[Shanzeee]

Hi @scrocquesel,

I’m really sorry for accidentally closing the issue earlier – it was unintentional.
I’ve tried the proposed solution (quarkus.dynamodb.advanced.retry-strategy.retry-mode=STANDARD), but it doesn’t seem to have any effect.

WARN  [io.qua.config] (Quarkus Main Thread) Unrecognized configuration key "quarkus.dynamodb.advanced.retry-strategy.retry-mode" was provided; it will be ignored;

I can provide a minimal reproducer project

Thank you for your patience, and apologies again for the confusion!

Quarkus version: 3.17.4

[scrocquesel]

It is a proposal. It is only a pull request for now. I was challenging the requirement as configuring the retry strategy by code allow a lot more possibility and I'm not sure it is really needed. When the base config is not enough, I guess users can still do what you mentioned

[Shanzeee]

Hi @scrocquesel,

After further investigation, I realized that my current approach:

class DynamoDbClientConfig {
    @ApplicationScoped
    fun dynamoDbAsyncClient(): DynamoDbAsyncClient {
        return DynamoDbAsyncClient.builder()
            .overrideConfiguration(
                ClientOverrideConfiguration.builder()
                    .retryStrategy(STANDARD)
                    .build()
            )
            .build()
    }

    @ApplicationScoped
    fun dynamoDbEnhancedAsyncClient(): DynamoDbEnhancedAsyncClient {
        return DynamoDbEnhancedAsyncClient.builder()
            .dynamoDbClient(dynamoDbAsyncClient())
            .build()
    }
}

does not work as expected. Instead of modifying the Quarkus-managed DynamoDbAsyncClient, it creates a completely new instance. This leads to inconsistencies, especially when using other Quarkus-provided integrations like quarkus.dynamodbenhanced.client-extensions.

Could you confirm if there’s a way to modify the existing Quarkus-managed DynamoDbAsyncClient rather than creating a new one? If not, would it make sense to extend the configuration options to support retry policies natively?

Thanks again for your help!