* Remove arktos cmd/kube-scheduler, pkg/scheduler, staging/src/k8s.io/kube-scheduler code * Copy 1.18.5 cmd/kube-scheduler, pkg/scheduler, staging/src/k8s.io/kube-scheduler code * Add file modified comment for copyright header correctness * Fix copyright header for backported 1.18.5 scheduler code * run ./hack/update-bazel.sh * Use arktos pkg/version/verflag * Copy 1.18 staging/src/k8s.io/component-base/metrics/prometheus folder and staging/src/k8s.io/component-base/metrics/*_test.go * Copy 1.18.5 staging/src/k8s.io/apimachinery/pkg/runtime/serializer/ codec_factory.go, codec_test.go, testing/conversion.go * Copy 1.18.5 staging/src/k8s.io/csi-translation-lib folder * K8s PR 81946 - add cache-control headers to kube-apiserver * update-bazel & copyright header * K8s PR 80681 - add options for name and namespace of leaderelection object Only picked up changes required to compile scheduler * K8s PR 84129 - add lenient decoding path for v1aplpha kube-scheduler config * K8s 1.18 PR 77595 Volume Scheduling Limits - Due to Scheduler interface change. * K8s 1.18 PR 88401: refactor volume binder * K8s 1.18 PR 84274: feature-gate PDB informer starts * Skip PR 83394 as Arktos does not support CSINodes * K8s 1.18 PR 82465 - Move getPodPriority from /scheduler/util to /api/pod * Code change to skip PR 85863 - renaming, no performance impact * Code change to skip PR 81754 - Big change not performance related. * K8s PR 83578: scheduler policy API refactoring . Pick up api violation exception only TODO: add changes to test/integration/scheduler * K8s 1.18 PR 86229 - remove ds controller dependency on scheduler metadata * K8s 1.18 PR 84323-feat: remove suspendedDaemonPods from daemon controller * K8s 1.18 PR 86730: Break DS controller dependency on scheduler predicates and predicate errors * K8s 1.18 PR 84746 - move out const strings in pkg/scheduler/api/well_known_lablels.go * K8s 1.18 PR 88435 - move well known cloud provider taints to k8s.io/cloud-provider/api * K8s 1.18 PR 87342 - move generalpredicates logic to kubelet * K8s PR 87788 - reduce overhead of error message formatting and allocation for scheduler NodeResource filter * K8s 1.18 PR 84294 - remove predicates.NodeInfo dependency from kubelet * K8s 1.18 PR 82255 - volume scheduling: move metrics to a separate package to avoid import cycle * K8s 1.18 PR 76968 - add Overhead to PodSpec and RuntimeClass * update-bazel&copyright * Manual fix api-rules/violation_exceptions.list for make update * K8s PR 80933 partial - make Overhead validation unconditional * K8s 1.18 PR 88338 - upgrade PodOverhead to beta * K8s 1.19 PR 77327 - Even Pods Spread API changes * K8s 1.18 PR 83474 - CSI Topology GA * K8s 1.18 PR 82329 - init check for cloud node controller * K8s 1.18 PR 82848 - Break out of loop when NodeHostName is found Perf improvement of node_controller + easier merge * K8s 1.18 PR 84469 - improve error handling in cloud node controller Easier merge of node controller + better error message * K8s 1.18 RR 81431 - promote node zone/region topology labels to GA * K8s 1.18 PR 82049 - promote node instance type label to GA Due to scheduler backporting dependencies - LabelInstanceTypeStable * K8s 1.18 PR 88230 - volume binder: simplify API Due to scheduler backporting dependencies on SchedulerVolumeBinder.FindPodVolumes interface * K8s 1.18 PR 79971 - add check to reduce orphaned volume . Cleaner history of pv_controller merging . Performance improvement * K8s 1.18 PR 79696 - fix pv-controller sync check node affinity for scheduled claim * K8s 1.18 PR 82683 - Refactor CSI Translation Library into a struct that is injected into various components to simplify unit testing * K8s 1.18 PR 87250 - adding taint toleration error reasons * K8s 1.18 PR 80284 - fix a racing issue in client-go UpdateTransportConfig Metrics dependencies * make update * Pick k8s PR 79879 - [e2e] Refactor andMove node related methods to framework/node package [1] Due to scheduler backporting make quick-release dependencies. * Pick k8s PR 81043 - Add new flag for whitelisting node taints Due to scheduler backporting make quick release dependencies. * Pick k8s PR 86983 - Remove test/e2e dependency on scheduler/predicates package Due to scheduler backporting make quick-release dependencies. + dup log code to remove cycle dependencies * Remove scheduler/algorithm/priorities/util package * Adopting e2e to Arktos * PR 80200 - Tolerate the case if related event cannot be found * Code fix for Arktos customization Update pkg/registry/storage/rest/storage_storage.go to not "Propagate error from NewREST" * make update * Temp disable certain restclient metrics for backporting difficulty * Arktos change: set hashkey for ObjectMeta during workload generation * Apply Akrtos multi-tenancy changes to backported scheduler Added additional multi-tenancy changes for fake/listers.go compilation. Not found changes in pkg/scheduler/factory/factory.go Needs review * Arktos changes: Partial runtime readiness support at node agent (#339) - Yunwen Rewrite following scheduler plugin pattern. Needs CR. 9895836 * Remove tenant from PodMatchesTermsNamespaceAndSelector as it is missing in Arktos and needs more changes. * Arktos change - bug fix: scheduler node runtime readiness predicate should respect pod no-schedule toleration (#593) h-w-chen Skipped predicates_test.go as there is no placeholder for the test. b012cae * Arktos changes - log QPS * Arktos POC changes in PR 991 Has some later changes, not compilable * Arktos POC: script change to start scale out cluster in local * Arktos POC: KCM/Scheduler/Kubelet code changes for scale out 2RPs * make update * Disable feature CSINodeInfo as it is introduced by 1.18.5 scheduler backporting and not supported in Arktos. * Manual for local scale out cluster setup. * Kubeup & Kubemark changes for multiple RPs. * Perf test changes for multiple RPs. * make update * Force setting nodes to be evaluated as 500. (#1072) * Fix perf-test clusterloader metrics - add back DeprecatedSchedulingLatencyName * fix error: User "system:kube-scheduler" cannot create resource "events" * fix scheduler error: event_recorder.go:55] Could not construct reference to: '<nil>' due to: 'can't reference a nil object' * using gcr.io as docker registry for flannel ds * VM start and stop with resource deallocation and unbinding from assigned node (#250) * (back-port to scheduler 530) start and stop with resource eviction from node * (back-port to scheduler 530) added a noschedule pod phase and removed most VirtualMachine logic from scheduler. a controller in charge (#273) of VM state change log is next so that the left-over logic can be removed from scheduler once and for all. * (back-port to scheduler 530) vmPodShouldSleep change of 04013de (added vm-pod-controller for VM state management (#292)) * (back-port to scheduler 530) Scheduler changes to use CommonInfo * (back-port to scheduler 530) scheduler production part of commit 23cbc84 (In-place Pod Vertical Scaling for container pods (#275)) * Fix unit tests for scheduler common info * Remove debug imports * Controller changes to use CommonInfo * K8s 1.18 RP 79837 - add fakes for events package, add startEventWatcher to event interface * Copy K8s 1.18 staging/src/k8s.io/client-go/tools/leaderelection folder over. Except interface changes to restclient Due to UTs in cmd/kube-scheduler/app/options * K8s PR 80681 - add options for name and namespace of leaderelection object - Partial picked up earlier for scheduler backporting. Add complete PR due to UTs. * Multiple RP support - bug fix - clone nodeInfo with rpId * Bug fix: add missing plugin code for partial runtime readiness support at node agent * Fix UT pkg/master/TestStorageVersionHashes - comment out csinodes in storage hash as it was disabled due to scheduler backporting * Daemonset UT fix: copied daemon_controller_test from 1.18.5. Applied multi-tenancy & Vertical Scaling changes * Align arktos scheduler multi-tenancy UTs with 1.18 * Fix UTs for multiple RP support & scheduler back porting & Disable CSIDriver related UTs & Update cluster-roles testdata * Bugfix for backporting In-place Pod Vertical Scaling for container pods #275 * Add back mistakenly removed updatePod code during cherry-pick from arktos scheduler changes * Fix arktos-up.sh * using commonInfo in scheduler resource predicate and priority plugin * add back vertical scheduling feature in scheduler code * fix a couple UT for vertical scheduling * add noderuntime readiness as default filters * Remove outdated OWNERS * Pick up scheduler changes caused daemonset integraion test changes * K8s PR 82795 - graduate ScheduleDaemonSetPods to GA Affecting integration tests * integration test: scheduler: refactor with scheduler back-ported * format or build change by make update * integration test: volume: minor refactor with RP/TP scale out feature * bug fix: not to start node informaer twice for tp/rp solo(combined) cluster * scheduler perf tests and test framework changes * provided test utility code of waitCachedPodsStable * integration test: scheduler: minor - import orderings and commented code removal * Address CR comment: sort import, rewrite comment, go-fmt, string fmt. * Return error when node list fail - PR comment * make update + copyrights correction * CR feedback changes: comment, remove obsoleted commented code. Co-authored-by: sonyafenge <sonyafenge@gmail.com> Co-authored-by: Yunwen Bai <yunwen.bai@futurewei.com> Co-authored-by: Peng Du <rick.peng.du@gmail.com> Co-authored-by: Hongwei Chen <hong.chen@futurewei.com> Co-authored-by: Vinay Kulkarni <vinay.kulkarni@huawei.com>

Consisten etcd key paths for the system tenants and regular tenants. #339

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions