Currently, the molecule CI runs sequentially. This is very slow, especially with the airgap test suite which downloads and copies big archives.
The limitation that causes that also means only one engineer can comfortably work on this repo at a time.
It also means the tests can often clash with the local dev as well and must run separate from it.
The reason for this is plenty of security group filtering relying on security_group_name (grep -r security_group_name molecule/)
This will clash when ran from multiple places (engineer's machine/another CI run/etc.)
We also need to check for other instances of such filtering (on instance names). It can be solved by suffixing the names with something unique. (eg. <username/GitHub job id>-<commit hash>-<platform>)
AC (Acceptance Criteria):
- all per-distro molecule tests run at the same time and don't error.
- multiple engineers can work on the repo and run converge/test steps without interfering with each other
This is low prio. We don't expect changes/CI runs here that often, so the payoff is not great.
Currently, the molecule CI runs sequentially. This is very slow, especially with the airgap test suite which downloads and copies big archives.
The limitation that causes that also means only one engineer can comfortably work on this repo at a time.
It also means the tests can often clash with the local dev as well and must run separate from it.
The reason for this is plenty of security group filtering relying on
security_group_name(grep -r security_group_name molecule/)This will clash when ran from multiple places (engineer's machine/another CI run/etc.)
We also need to check for other instances of such filtering (on instance names). It can be solved by suffixing the names with something unique. (eg.
<username/GitHub job id>-<commit hash>-<platform>)AC (Acceptance Criteria):
This is low prio. We don't expect changes/CI runs here that often, so the payoff is not great.