Improvements of the scale testing framework

The SCALE code package is deployed across a wide variety of operating systems and utilizes several programming languages. Testing to ensure that there are no code regressions and that programs run as expected across all of these configurations can be challenging. In this paper, we describe the process of migrating SCALE's automated testing framework to an OpenStack cloud instance for managing continuous integration testing across Windows, MacOS, and Linux operating systems. Prior to this work, SCALE relied upon GitLab pipelines executed on a variety of physical machines. These machines were difficult to change and had little to no configuration control on individual machines; as such, over time, the configurations were susceptible to drift. In addition, the majority of SCALE's testing Windows machines were only 8 or 16 CPU cores and could take over five hours to run the test suite. To aid in the testing process, an OpenStack cloud instance was set up at Oak Ridge National Laboratory. All but one of these machines have been turned into hypervisor nodes that run a barebones operating system with Qemu/KVM. The final machine has been turned into a SCALE development box where developers can test changes. Pairing Qemu/KVM with a custom GitLab Runner allows the continuous integration (testing) pipeline to use a wide variety of operating systems and versions from a known set configuration. Having machines "destroyed" after each test allows there to be a controlled configuration. It also allows for quick and easy prototyping of new operating systems, compilers, and versions of third-party libraries. Testing time has been cut by almost 75% by utilizing the OpenStack instance and virtual machine testers. (author)