Rebase to upstream/cluster-autoscaler-release-1.14 #107
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
VPA - e2e tests for v1beta2 API
Fix readme to reflect addon-resizer version 2.1.
VPA - fix names of files with junit test results
Vertical Pod Autoscaler version 0.4.0
VPA - disable errexit in run e2e scripts
Get HTTP responses together with error, so that we could get full error messages. It also fixes some edge cases for success responses.
Update VPA to beta in main README
Add version field to spec
VPA - check ConfigDeprecated condition for v1beta1 e2e tests
Fix error message for long-waiting operations
Update CA godeps to Kubernetes v1.15.0-alpha.0
Add the ability to better pass additional arguments to containers
Change default version Update docs
VPA switches to 0.4.0
Finds a machine using provider ID.
…r ID Change findMachineByNodeProviderID() to first match using ProviderID values. If there is no match (because actuator does not set the value on the machine object) then fallback to matchin based on the "machine" annotation from the node object. Update unit tests to reflect new behaviour.
Add new test that verifies that machines without a ProviderID value can still be found if there is a "machine" annotation on the node object.
When mapping from a machine to a node, prioritize the search based on ProviderID, falling back to using machine.Status.NodeRef.Name. This value is populated by the nodelink-controller.
Emit MaxNodesTotalReached in scale up when the condition is reached.
Be less specific as to when the event is generated. The qualifying criteria is when the adding a node would reach the --max-nodes-total specified.
We previously only considered a MachineSet or MachineDeployment
suitable for scaling if it had positive scaling bounds.
For example:
annotations:
machine.openshift.io/cluster-api-autoscaler-node-group-min-size: "1"
machine.openshift.io/cluster-api-autoscaler-node-group-max-size: "6"
On us-west-2 we have:
$ kubectl get machinesets
NAME DESIRED CURRENT READY AVAILABLE AGE
amcdermo-ca-5ws2b-worker-us-west-2a 1 1 1 1 4h23m
amcdermo-ca-5ws2b-worker-us-west-2b 1 1 1 1 4h23m
amcdermo-ca-5ws2b-worker-us-west-2c 1 1 1 1 4h23m
amcdermo-ca-5ws2b-worker-us-west-2d 0 0 4h23m
but the autoscaler will fail to scale up any of these because the last
machineset has a replica count of 0. This change modifies the criteria
for where we return a nodegroup (i.e., machine set) to must have:
- positive scaling bounds, and
- the underlying resource must have a replica count that is > 0.
When creating a test config I had previously allowed for the number of nodes to be different to the number of replicas. We don't use this distinction so dropping that entirely from all the unit tests.
This invokes the e2e tests and is mainly invoked by CI.
frobware
force-pushed
the
rebase-to-upstream-1.14
branch
from
9216620 to
6efd972
Compare
|
The only difference between what we had previously for the cloudprovider is: $ git diff openshift/master ./cluster-autoscaler/cloudprovider/openshiftmachineapi/
diff --git a/cluster-autoscaler/cloudprovider/openshiftmachineapi/machineapi_nodegroup.go b/cluster-autoscaler/cloudprovider/openshiftmachineapi/machineapi_nodegroup.go
index 383e527e5..84c1998de 100644
--- a/cluster-autoscaler/cloudprovider/openshiftmachineapi/machineapi_nodegroup.go
+++ b/cluster-autoscaler/cloudprovider/openshiftmachineapi/machineapi_nodegroup.go
@@ -24,7 +24,7 @@ import (
machinev1beta1 "github.com/openshift/cluster-api/pkg/client/clientset_generated/clientset/typed/machine/v1beta1"
apiv1 "k8s.io/api/core/v1"
"k8s.io/autoscaler/cluster-autoscaler/cloudprovider"
- schedulercache "k8s.io/kubernetes/pkg/scheduler/cache"
+ schedulernodeinfo "k8s.io/kubernetes/pkg/scheduler/nodeinfo"
)
const (
@@ -188,7 +188,7 @@ func (ng *nodegroup) Nodes() ([]cloudprovider.Instance, error) {
// allocatable information as well as all pods that are started on the
// node by default, using manifest (most likely only kube-proxy).
// Implementation optional.
-func (ng *nodegroup) TemplateNodeInfo() (*schedulercache.NodeInfo, error) {
+func (ng *nodegroup) TemplateNodeInfo() (*schedulernodeinfo.NodeInfo, error) {
return nil, cloudprovider.ErrNotImplemented
}This represents a change made in upstream/cluster-autoscaler-release-1.14. |
frobware
changed the title
openshift-ci-robot
removed
the
do-not-merge/work-in-progress
|
Key output from jenkins job: I0620 10:48:53.232455 1 main.go:345] Cluster Autoscaler 1.14.3 |
|
/hold Would like some feedback (if possible) from @joelsmith and/or @sjenning to ensure I haven't broken the VPA in any way. |
openshift-ci-robot
added
the
do-not-merge/hold
|
Key output from CI/openshift: I0620 11:16:48.931790 1 main.go:345] Cluster Autoscaler 1.14.3 |
|
/hold cancel |
openshift-ci-robot
removed
the
do-not-merge/hold
Merged
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The PR was created by first taking upstream/cluster-autoscaler-release-1.14 as the base then applying UPSTREAM: patches on top. The set of picks applied was derived from:
git log --oneline --no-merges 810bb14c0a53261dfa3f2afb8e563ce84f8bd566..openshift/masterwhere 810bb14 reflects the changes since our last rebase (which was upstream/cluster-autoscaler-release-1.13).
And in that set of picks anything that was related to vendoring in
test/openshiftwas dropped because we have moved to cloning the https://github.com/openshift/cluster-api-actuator-pkg when running the e2e tests. There is no value in picking all these historical commits as they were made redundant in PR #104 .Equally, anything that was previously vendored in
cluster-autoscaler/vendorhas been dropped. We have to run thecluster-autoscaler/fix_gopath.shscript again to pick up the new kubernetes version (defined incluster-autoscaler/kubernetes.sync) and as part of that process you have torm -rf vendor Godeps.To create the merge commit I have used the following steps:
Picks were as follows:
For details on the merge^{tree} syntax please read this documentation.
The result of git commit-tree in this PR is b68c8ee.
The previous rebase to upstream/cluster-autoscaler-release-1.13 was done in PR #76.