DEFAULT_REQUEST_RETRY_BACKOFFLIMIT value

[manusa]

Description

Will be applicable after merging #4825

The current value for Config.DEFAULT_REQUEST_RETRY_BACKOFFLIMIT is currently set to 0.

kubernetes-client/kubernetes-client-api/src/main/java/io/fabric8/kubernetes/client/Config.java

Line 147 in 7ff51aa

public static final Integer DEFAULT_REQUEST_RETRY_BACKOFFLIMIT = 0;

This means that the ExponentialBackoffIntervalCalculator will never actually do a retry by default.

kubernetes-client/kubernetes-client-api/src/main/java/io/fabric8/kubernetes/client/utils/ExponentialBackoffIntervalCalculator.java

Lines 69 to 71 in 7ff51aa

	public boolean shouldRetry() {
	return maxRetries < 0 || currentReconnectAttempt.get() < maxRetries;
	}

Should this value changed to something else (for example 5) in order to take advantage of the HttpClient retry functionality?

[shawkins]

Should this value changed to something else (for example 5) in order to take advantage of the HttpClient retry functionality?

I'm in favor of that. My guess at a the default retries (10) from the go client: https://github.com/kubernetes/client-go/blob/58ff029093df37cad9fa28778a37f11fa495d9cf/rest/request.go#L157

This will help mitigate or remove the need for the timeout handling introduced in #4637 - as it should be 500s that we're seeing prior to being able to perform pod operations.