You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
To agent CIs should be hosted on CFF accounts, so that all approvers have access to it. Currently, it is host on Broadcom infrastructure and only Broadcom approvers can create new releases.
The agent supports also Windows and there are Windows integration tests executed during the validation. The move will require to explore options to execute the Windows integration tests on GCP where all the BOSH validations are hosted.
There is a Slack discussion here which has some more details.
If we are ok with the cost, the simplest thing to do would be to deploy a single windows 2019 worker in the FIWG concourse and move the pipeline over as-is. Anything else would require refactoring of the pipeline most likely.
@selzoc which GCP instance type we will need for this? We could also have an approach like for the sole tenancy resource support I refactored in cloudfoundry/bosh-google-cpi-release#367. The sole tenancy instance is very expensive and it is started only for the integration tests execution and destroyed afterwards.
Also something I just realized is that the integration tests rely on a persistent bosh director that we have deployed, so we'd need either one of those in the CFF account or add a "deploy director" step into the pipeline.
Actually, the Application Platform WG has also Windows validation tests. We could check with them how they use Windows Workers. Additionally, there is the initiative to have a central Concourse Instance provided by the Concourse WG and with this they could provide a Windows worker used by all CF components to execute Windows validations there. That is why when moving the pipelines we should also consider to move them directly to the central Concourse if it is available.
Activity
selzoc commentedon Sep 22, 2025
If we are ok with the cost, the simplest thing to do would be to deploy a single windows 2019 worker in the FIWG concourse and move the pipeline over as-is. Anything else would require refactoring of the pipeline most likely.
beyhan commentedon Sep 23, 2025
@selzoc which GCP instance type we will need for this? We could also have an approach like for the sole tenancy resource support I refactored in cloudfoundry/bosh-google-cpi-release#367. The sole tenancy instance is very expensive and it is started only for the integration tests execution and destroyed afterwards.
selzoc commentedon Sep 23, 2025
Our internal windows workers are configured like
which is roughly equivalent to
e2-standard-8I think?beyhan commentedon Sep 24, 2025
Yes, I think that should be the recommended instance type.
Here's a quick comparison for the total estimated monthly cost of an
n2-standard-8+ Windows + 512 GiB Disk inus-central1:selzoc commentedon Sep 24, 2025
Also something I just realized is that the integration tests rely on a persistent bosh director that we have deployed, so we'd need either one of those in the CFF account or add a "deploy director" step into the pipeline.
beyhan commentedon Sep 25, 2025
Actually, the Application Platform WG has also Windows validation tests. We could check with them how they use Windows Workers. Additionally, there is the initiative to have a central Concourse Instance provided by the Concourse WG and with this they could provide a Windows worker used by all CF components to execute Windows validations there. That is why when moving the pipelines we should also consider to move them directly to the central Concourse if it is available.
aramprice commentedon Sep 26, 2025
Some YAML snippets which might be helpful for getting windows workers setup with the FIWG concourse
pipeline.ymlsnippet:A bunch of details are left out of of this snippet because they area iaas specific
windows-worker-config.ymlconcourse-use-latest-release.yml