AWS api calls are throttled when there are too many parallel builds about aws-codebuild-jenkins-plugin

[oliver-schoenherr]

The AWS Java SDK includes a Backoff Strategy:
com.amazonaws.retry.PredefinedRetryPolicies.SDKDefaultBackoffStrategy

imho a retry parameter makes no sense from the User Perspective. Maybe this is a major drawback of the solution - because of, normal Plugin User don't understands the dependency between ThrottlingExceptions and Retry Counts.

You should start with something hardcoded like max 10 Retry.

envato/stack_master@12448d8#diff-1f8a98128798cdf6364b9ebc6358d44fR29

[uttamk]

Proposed solution

Add a pollingInterval build parameter to the CodeBuildStep. Keep the default to 5 seconds. Let the users decide if they want to increase it to suit their parallelism requirement.

I have a PR for this. See #32

[oliver-schoenherr]

The Java AWS SDK includes some Retry Logic concerning ThrottlingExceptions, maybe you should review the default Client Configuration and increase the maxErrorRetry Value.

com.amazonaws.retry.RetryUtils#isThrottlingException

[uttamk]

@oliver-schoenherr .. nice suggestion. Do we retry with increasing sleep times after every retry? We would still need a limit for the number of retries. Do we hard code this in the code or let the plugin have a retries parameter ?

[uttamk]

sounds good @oliver-schoenherr. We can start off by hard coding the retry limit/sleep time and see if it works for most use cases. If users need a higher retry limit, we can think about parameterizing the retry limit.
cc. @leobaran-aws @jvusich any thoughts ?

[leoherran-aws]

Sounds good to me! Implementing backoff for build status polling is a great idea.

[leoherran-aws]

Merged #35.