Use bigger machines for install on bootstrap conformance tests

[fabriziopandini]

/kind feature

Describe the solution you'd like
Conformance tests are using bigger machines to increase speed/reduce flakes, but install on bootstrap conformance tests are not.

We should start using bigger machines for install on bootstrap conformance tests too

Anything else you would like to add:

The existing patch cannot be used (it applies to legacy clusters. install on bootstrap uses a classy cluster)
Also, this is a problem that exists for both govimomi and supervisor templates.

[sbueringer]

I'm fine with doing that. But I'm a bit confused why this was never a problem with govmomi (tests are absolutely stable there afaik) or are we already using bigger machines there?

[fabriziopandini]

are we already using bigger machines there? (in the govmomi test)

AFAIK no

why this was never a problem with govmomi (tests are absolutely stable there afaik)

I don't have full historical context on why conformance is using bigger machines, I'm just assuming the two conformance tests have the same requirements.

We can also put this on hold and wait for when we have data about actual resource consumption (I know we have those data for the jobs migrated to the eks cluster, I'm not sure if there is something similar for the jobs running on the google cluster)

[sbueringer]

I think we're talking about different resources. Wasn't the idea to change the size of the VMs created in VMC? Not the resources for the ProwJob?

[fabriziopandini]

It is the VM size, rif https://github.com/kubernetes-sigs/cluster-api-provider-vsphere/blob/main/test/e2e/data/infrastructure-vsphere-govmomi/main/conformance/worker-node-size.yaml
(and you are right, prow monitor will not provide the data we need)

[sbueringer]

I'm fine either way. I think we can either wait to see if the conformance tests are flaky or fix it directly