Don’t handle web requests until the server is started

In  undertow

Overview

WFLY-6493 changed the default behaviour of the server to queue requests while the server is starting, instead of just rejecting them (or potentially letting a request through to an incomplete deployments).

This feature is about making this feature configurable, so that a user can choose between the current behaviour of queueing requests until the server has started, or between rejecting requests until the server has started.

Note that neither of these selections will allow request processing to start until the server has completely started, which is an improvement on the original behaviour.

Issue Metadata

Issue

Dev Contacts

QE Contacts

Affected Projects or Components

  • Undertow

Requirements

Hard Requirements

This feature will provide an attribute on the host called queue-requests-on-start, this attribute will default to true.

If this attribute is true then there will be no deviation from the current behaviour, requests that arrive when the server is starting will be held until the server is ready.

If this attribute is false then requests that arrive before the server has completely started will be rejected with the default response code.

No matter the value of this attribute request processing will not start until the server is completely started, making it impossible for requests to be processed by an incomplete deployment.

Test Plan

This is complex to test, as our existing test infrastructure is not set up to perform tests while the server is starting up. As such tests will be confined to the QE internal test suites, rather than being part of the main WildFly test suite.