When we fallback to different AZ's with test resources using Capacity Reservation, runner is already created and might be in different AZ This infers additional costs.
Idea is to first reserve capacity and after that create sct-runner in the same AZ as capacity was reserved.
Another idea would be to make 'provision step' to provision also sct-runner. I think we would need to fix only setup of syslogng to point to host (e.g. sct-runner) and after provisioning instances add ip of sct-runner to /etc/hosts in each provisioned instance so it points to valid host. This way we could shave another 1minute from test run.
@soyacz are we doing fallback now if capacity reservation fails ? if not, we can postpone this to when we would want to do fallbacks to other AZ or regions
When we fallback to different AZ's with test resources using Capacity Reservation, runner is already created and might be in different AZ This infers additional costs.
Idea is to first reserve capacity and after that create sct-runner in the same AZ as capacity was reserved.
Another idea would be to make 'provision step' to provision also sct-runner. I think we would need to fix only setup of syslogng to point to host (e.g. sct-runner) and after provisioning instances add ip of sct-runner to /etc/hosts in each provisioned instance so it points to valid host. This way we could shave another 1minute from test run.