The current CI pipeline running at LLNL was a good first aim and proof of concept. The issue is that is has become difficult to maintain with the container builds and often fails tests.
This issue proposes that instead of building containers with on LLNL machines, we test the Docker Hub images using LLNL machines.
This Issue also Proposes that doing a fresh spack build for every pipeline is in efficient and unnecessary. Instead, we can look into a process to uninstall only the ExaWorks packages from an already existent spack installation, then reinstall them at the latest versions to capture any code changes.
Both of these changes will vastly reduce CI/CD run times and complexity, making development easier and results more reliable.
The current CI pipeline running at LLNL was a good first aim and proof of concept. The issue is that is has become difficult to maintain with the container builds and often fails tests.
This issue proposes that instead of building containers with on LLNL machines, we test the Docker Hub images using LLNL machines.
This Issue also Proposes that doing a fresh spack build for every pipeline is in efficient and unnecessary. Instead, we can look into a process to uninstall only the ExaWorks packages from an already existent spack installation, then reinstall them at the latest versions to capture any code changes.
Both of these changes will vastly reduce CI/CD run times and complexity, making development easier and results more reliable.