AUTOSCALE-335,AUTOSCALE-336: 1.34.0 upstream rebase #386
Conversation
fix(VPA): Do not update webhook CA when registerWebhook is disabled
Signed-off-by: Yuriy Losev <[email protected]>
Signed-off-by: Yuriy Losev <[email protected]>
Signed-off-by: Yuriy Losev <[email protected]>
[VPA] Use factory start to fill caches instead of separate informers
…t-success OCI provider: Avoid interpreting HTTP 404 as success on delete
…-cloud-endpoint-reloving fix bug 8168 GetEndpoint resolving fail
…e-terminate-by-default feat: cordon node before terminate by default
this change adds debug logs at level 5 to aid in triaging failed node balancing. It adds logs to help determine why two node groups are not considered as similar. These logs can be quite noisy so the logging level has been set to 5 by default.
AEP-7862: Decouple Startup CPU Boost from VPA modes - updates
* add h4d pricing * fix go fmt * revert gofmt on other files
cluster-autoscaler: add logging for failed node balancing
./hack/update-deps.sh v1.34.0-alpha.1 v1.34.0-alpha.1 https://github.com/kubernetes/kubernetes.git
Following kubernetes#7195
hack/update-codegen.sh
This reverts commit 897989f.
As discussed in sig-autoscaling meeting on 2025-06-30, this is to try follow a similar pattern to the KEP process by getting a tech lead's buy in before merging an AEP.
… values in real apis not necessary
CA: remove azure UT cases
…s-approvers-for-aeps Give sig-autoscaling-leads approval of the AEP directory
chore: bump golangci lint to v2
…ode-groups-from-balancing Filter out non-existing node-groups before scale-up balancing
Signed-off-by: bo.jiang <[email protected]>
Fix capacity buffers injector order in pod list processor
… registered nodes
…test-in-docker` `make test-in-docker` was changed to disable the printf analyzer, but `make test-unit` wasn't for some reason. The current master isn't compatible with the printf analyzer, so `make test-unit` fails on master without this change.
…erry-pick-8552-to-cluster-autoscaler-release-1.34 [cluster-autoscaler-release-1.34] Allow atomic scale down of partially healthy node groups
TestNodeLoadFromExistingTaints creates a currentTime variable set to time.Now(), and a bunch of test objects with time values offset from that variable. This is all standard practice, but then the test iterates over test cases, calls t.Parallel(), and overwrites currentTime with time.Now() again. This makes go test -race fail, because multiple goroutines are writing currentTime at once. It also doesn't seem to make sense in the context of the test, because the other test objects are still offset from the original value. Removing the second write to currentTime seems to be the correct fix here. Also renamed one import because it collided with a local variable name used throughout this test file.
…erry-pick-8584-to-cluster-autoscaler-release-1.34 [cluster-autoscaler-release-1.34] Change `make test-unit` to have the same go test parameters as `make test-in-docker`
…erry-pick-8588-to-cluster-autoscaler-release-1.34 [cluster-autoscaler-release-1.34] Fix a race condition in TestNodeLoadFromExistingTaints
The DRA scheduler plugin is enabled by default since 1.34. We have to hack it to be disabled if the CA DRA logic is disabled via the flag. Without this, the DRA scheduler plugin is enabled but not set up properly, and panics.
…erry-pick-8598-to-cluster-autoscaler-release-1.34 [cluster-autoscaler-release-1.34] Fix DRA enablement logic
openshift-ci-robot
added
the
jira/valid-reference
|
@joelsmith: This pull request references AUTOSCALE-335 which is a valid jira issue. In response to this:
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the openshift-eng/jira-lifecycle-plugin repository. |
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: The full list of commands accepted by this bot can be found here.
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
This change carries files and modifications that are used by OpenShift release infrastructure and related files. * spec file * dockerfiles * vertical-pod-autoscaler/Dockerfile.rhel * vertical-pod-autoscaler/Dockerfile.openshift * images/cluster-autoscaler/Dockerfile * images/cluster-autoscaler/Dockerfile.rhel * hack scripts (ci and build related) * Makefile * JUnit tools * update gitignore * update/remove OWNERS files * ci-operator config yaml * remove gitignore file from vertical-pod-autoscaler (allow vendor addition) * add Snyk file to exclude vendor directories and problematic cloud providers on scan
Add vendor folders * cluster-autoscaler * balancer * vertical-pod-autoscaler * vertical-pod-autoscaler/e2e for i in cluster-autoscaler balancer vertical-pod-autoscaler vertical-pod-autoscaler/e2e; do pushd $i; go mod tidy; go mod vendor; popd; done
…otation The delete annotation upstream has a different format, but is now inferred dynamically from the API group. If we update this in MAO to use the new format, we can drop this old key
This change re-adds the machine api support for labels and taints on node groups. The code was removed upstream as it is openshift specific, see this pull request[0]. It also adds in the functionality of the upstream override annotation for labels and taints[1] to support https://issues.redhat.com/browse/MIXEDARCH-259 [0]: kubernetes#5249 [1]: kubernetes#5382
the upstream annotations for the scale from zero capacity resources is slighty different than the openshift implementation. the largest difference is the addition of a gpu type annotation. openshift does not yet utilize this annotation and thus this patch should be carried until the machineset controllers for the various providers on openshift have been modified to use the new annotations. another important change is the modification of the memory annotation. previously in openshift we expected this value to be a count of memory in Mebibytes. the conversion function and tests have been modified to allow continued openshift operation. this change can be dropped when the annotations in openshift have been updated, the progress for this effort can be followed at https://issues.redhat.com/browse/OCPCLOUD-944
|
@joelsmith: This pull request references AUTOSCALE-335 which is a valid jira issue. In response to this:
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the openshift-eng/jira-lifecycle-plugin repository. |
…ider This change corrects the behavior for parsing taints from the unstructured scalable resource. This is required on OpenShift as our implementation is slightly different from the upstream.
Also: * Add unit tests for upstream annotations * Update unit tests using upstream annotations new values
joelsmith
changed the title
|
@joelsmith: This pull request references AUTOSCALE-335 which is a valid jira issue. This pull request references AUTOSCALE-336 which is a valid jira issue. In response to this:
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the openshift-eng/jira-lifecycle-plugin repository. |
|
@joelsmith: The following tests failed, say
Full PR test history. Your PR dashboard. Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository. I understand the commands that are listed here. |
Started out with this command:
After it was done, I removed the cherry-picks and manually cherry-picked the set of patches it identified so that I could manually resolve merge conflicts.
I squashed "Remove OWNERS automation preamble" into "configure repository for OpenShift releases"
Most of the cherry-picks required minimal or no changes, but "Fix unstructured taint parsing in Cluster API provider" required substantial changes due to the upstream PR kubernetes#8536 which refactored a lot of the cluster API provider's test framework.