More support for cluster-api #17636
More support for cluster-api #17636
Conversation
k8s-ci-robot
added
do-not-merge/work-in-progress
justinsb
force-pushed
the
clusterapi_controllers
branch
2 times, most recently
from
c2c7c26 to
5e7c5bb
Compare
hakman
requested review from
hakman
and removed request for
olemarkus and
zetaab
justinsb
force-pushed
the
clusterapi_controllers
branch
4 times, most recently
from
80b82ec to
855ef49
Compare
justinsb
force-pushed
the
clusterapi_controllers
branch
2 times, most recently
from
27f468b to
3e80afa
Compare
|
/test pull-kops-scenario-clusterapi-gcp Let's try this new test :-) |
|
/test pull-kops-scenario-clusterapi-gcp |
hakman
force-pushed
the
clusterapi_controllers
branch
from
b8bc822 to
a4fbb3d
Compare
|
/test pull-kops-scenario-clusterapi-gcp |
hakman
force-pushed
the
clusterapi_controllers
branch
from
a4fbb3d to
5496816
Compare
|
/test pull-kops-scenario-clusterapi-gcp |
justinsb
force-pushed
the
clusterapi_controllers
branch
2 times, most recently
from
a3ef085 to
fb68367
Compare
|
/test pull-kops-scenario-clusterapi-gcp |
justinsb
force-pushed
the
clusterapi_controllers
branch
from
fb68367 to
b7f6dc3
Compare
|
/test pull-kops-scenario-clusterapi-gcp |
justinsb
force-pushed
the
clusterapi_controllers
branch
from
4809bb4 to
a4ef300
Compare
|
/test pull-kops-scenario-clusterapi-gcp Doh! Looks like the bare-metal tasks were running out of disk space - finally spotted the warning message! Found a script from the Apache Flink project (Apache licensed) and adapted it - e.g. we want to keep gcloud / azure-cli etc! |
@justinsb unrelated but you can now run those bare-metal tests in prow |
Well, not entirely unrelated given the pain :-). I will look into that! |
Just make sure you pick the right nodepool for the gcloud container node-pools list --cluster=prow-build --project k8s-infra-prow-build --region us-central1 --format='value(name,config.advancedMachineFeatures)'
pool6-20251021213510079600000001 enableNestedVirtualization=True
pool7-arm64-20251021220031252300000001
pool7-20251021221638051700000001Also, GKE Node could include a label |
The bare-metal tests were running out of disk space, which leads to confusing errors.
As we're struggling with disk space, we don't want to install all the recommended packages.
Less hacky support for GCP, encode more of the logic into controllers. Co-authored-by: Ciprian Hacman <[email protected]>
Not the cleanest presentation, but this is the thing that is causing the most trouble right now.
We are unifying our asset-building paths, and the new path does not (yet) support filtering architectures inside kops-controller.
There's no logging from systemctl restart, so waiting for it to complete actually makes diagnostics harder.
justinsb
force-pushed
the
clusterapi_controllers
branch
from
a4ef300 to
7868e58
Compare
|
/test pull-kops-scenario-clusterapi-gcp OK, looks like bare-metal is the disk space. So now did I break CAPI while trying to fix bare-metal? :-) |
|
/lgtm |
k8s-ci-robot
added
do-not-merge/hold
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: hakman The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
k8s-ci-robot
added
the
approved
|
Awesome. This has been a journey so I think we should unhold and iterate! Thanks for reviewing! /hold cancel |
k8s-ci-robot
removed
the
do-not-merge/hold
|
Amazing progress, and quite fun to dig into CAPI. |
4 participants
Less hacky support for GCP, encode more of the logic into controllers.