Description
Context
Managing the namePrefix token through the settings.json mitigated the risk of losing ownership on the globally unique resource names such as keyvaults.
It turns out that it is not enough since those names keep on being used by others, most likely carml users who forget to change the namePrefix value before running the pipelines. That is causing related pipelines to fail in the CARML testing environment due to resource names being already taken.
We currently need to change those names and further regenerate readmes containing 1:1 module samples to solve corresponding pipelines failures.
This is expected to get worse with the increase of the CARML usage and with the introduction of the new dependencies approach, since also dependencies will be removed and the number of failing pipelines is likely to increase.
Suggestion
- Create a new GH secret/ADO variable to host the namePrefix
- Keep the possibility to manage the token via the settings.json
- Implement a logic to give priority to the settings.json token if set, default to the secret otherwise
- In CARML comment out the namePrefix token in the settings.json and remove our value for the CI environment to leverage the env variable.
- Update the test module locally scripts accordingly
- Update documentation
cc: @ahmadabdalla, @MrMCake
Description
Context
Managing the namePrefix token through the settings.json mitigated the risk of losing ownership on the globally unique resource names such as keyvaults.
It turns out that it is not enough since those names keep on being used by others, most likely carml users who forget to change the namePrefix value before running the pipelines. That is causing related pipelines to fail in the CARML testing environment due to resource names being already taken.
We currently need to change those names and further regenerate readmes containing 1:1 module samples to solve corresponding pipelines failures.
This is expected to get worse with the increase of the CARML usage and with the introduction of the new dependencies approach, since also dependencies will be removed and the number of failing pipelines is likely to increase.
Suggestion
cc: @ahmadabdalla, @MrMCake