Install build strategies via operator #171
Conversation
openshift-ci
bot
added
release-note
do-not-merge/work-in-progress
pull-request-size
bot
added
the
size/XXL
| if !crdExists { | ||
| return true, nil | ||
| } | ||
| // Apply the provided manifest containing the build strategies |
There was a problem hiding this comment.
After checkiing crds, shouldnt we wait also for the webhook svc to be up, otherwise we will face a connection refused error when we apply the stragetgies
I have seen this issue happening when the webhook is not up, and with parallel controllers, now that the logic is after the webhook install, we may not have the issue
conversion webhook for shipwright.io/v1beta1, Kind=ClusterBuildStrategy failed: Post \"https://shp-build-webhook.shipwright-operator.svc:443/convert?timeout=30s\": dial tcp 10.96.125.48:443: connect: connection refused
There was a problem hiding this comment.
Great point! I think this explains a lot, especially with the tests driven by EnvTest. Unlike a real cluster, we have no pods being deployed for the webhook (or service, for that matter). I'll switch to using the v1alpha1 samples to see if that fixes the issue by avoiding the conversion webhook invocation.
Add shipwright-io/build v0.12.0 as a dependency, with additional updates via `go get`: - Update ginkgo to v2.13.0 - Update k8s.io/* to v0.26.9
adambkaplan
force-pushed
the
install-build-strategies
branch
from
b72656a to
5260204
Compare
Install the build strategy samples from shipwright-io/build, at the v0.12.0 tag. Only cluster build strategies were added - namespace scoped build strategies were dropped. Red Hat-specific strategies were also dropped, in part to avoid potential trademark issues. The build strategies use v1alpha1 APIs because this is the stored version in v0.12.0. Testing revealed that if we reconciled v1beta1 APIs, we would need to wait for (and mock out!) the conversion webhook deployment. Lastly, the operator was granted RBAC permission to administer all ClusterBuildStrategies. Reconciling build strategies required adding logic to wait for the required CRDs to be installed on the cluster first. If a requeue is required, the operator will report the Ready status condition as "Unknown." The manifestival library code also had to enhanced to optionally recurse a directory for manifests to deploy. Finally, development of this feature revealed refactoring opportunities with respect to our use of k8s and controller-runtime client libraries, and the way we are organizing/managing Manifestival-driven reconcilers. Some refactoring of test code was included to simplify future testing efforts.
adambkaplan
force-pushed
the
install-build-strategies
branch
from
5260204 to
d423d71
Compare
adambkaplan
changed the title
openshift-ci
bot
removed
the
do-not-merge/work-in-progress
|
/assign @jkhelil Thanks for pointing me in the right direction wrt. the conversion webhook. Reconciling with the "stored" api version bypasses the need to wait for the conversion webhook to be ready. |
|
/approve |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: jkhelil 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 |
openshift-ci
bot
added
the
approved
|
/lgtm |
Changes
Enhance the operator to install a set of sample cluster build strategies. These were hand-copied from shipwright-io/build samples, only including
ClusterBuildStrategyobjects.Fixes #155
/kind feature
Submitter Checklist
See the contributor guide
for details on coding conventions, github and prow interactions, and the code review process.
Release Notes