We want to run large scale benchmarks on our cloud infrastructure.
For that we will use 11.x in a Docker image.
In the context of the benchmark we know that we will know to do code changes:
- new importer classes
- i.e.: to create the documents randomly with the right meta-data
- new benchmark scenario
- i.e. for the new scenario we may want to create
- fix and optimization we will (probably) identify while testing
- i.e. how we handle FullText storage
My initial assertion is that it makes sense to do that work in a branch rather than in a dedicated addon:
- this is easier / faster to iterate
- we probably want to end up merging most of the changes in master
Ideally, we want to be able to quickly iterate and quickly rebuild a new Docker image to run the tests.
Compared to the existing pipeline, we will want to do some changes:
- take code from the target branch instead of master
- bypass some tests in order to speed up the pipeline
- add some packages or bundles inside the image
- add importer
- add webui
- add a studio project
- change the default configuration
- i.e. use sequence for uuids