[HTTPCLIENT-1800] Why is Retry around Redirect and not the other way round

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 4.5.2
• Fix Version/s: 4.5.3, 5.0 Alpha2
• Component/s: HttpClient (classic)
• Assignee: Unassigned
• Reporter: Nicolai Stäger
• Created: 04/Jan/17 17:30
• Updated: 10/Jan/17 09:17
• Resolved: 07/Jan/17 11:40

Description

I am wondering, why the ServiceUnavailableRetryExec is before the RedirectExec in the execChain of the the InternalHttpClient. I am referring to the HttpClientBuilder.build() method where the execChain is put together something like this:

Client -> ... -> ServiceUnavailableRetryExec -> RedirectExec -> ...

We have the following scenario, where this causes an unexpected behavior and changing the order would solve this:

1. We are sending a request to endpoint A which returns a 303 with the endpoint B in the location-header
2. The RedirectExec redirects the request to B what returns a 503
3. Now the ServiceUnavailableRetryExec is doing its thing but retries to endpoint A

I would expect the behavior of the client to retry to endpoint B instead of A. Is there a reason why the chain is implemented this way around or is this a "bug".

Any help very appreciated!

Issue Links

• links to: GitHub Pull Request #62