KEP-3063: dynamic resource allocation

keps/prod-readiness/sig-node/3063.yaml

@@ -0,0 +1,6 @@
+# The KEP must have an approver from the
+# "prod-readiness-approvers" group 
+# of http://git.k8s.io/enhancements/OWNERS_ALIASES
+kep-number: 3063
+alpha:
+  approver: "@johnbelamaric"

keps/sig-node/3063-dynamic-resource-allocation/Makefile

@@ -0,0 +1,20 @@
+IMAGES += components.png kubelet.png
+
+all: $(IMAGES)
+clean:
+	rm -f $(IMAGES)
+
+# We use the http://plantuml.com/plantuml server to generate
+# images. That way nothing needs to be installed besides Go.
+DOC_PLANTUML_GO = $(shell go env GOPATH)/bin/plantuml-go
+
+%.png: %.puml $(DOC_PLANTUML_GO)
+	$(DOC_PLANTUML_GO) -format png $<
+
+%.svg: %.puml $(DOC_PLANTUML_GO)
+	$(DOC_PLANTUML_GO) -format svg $<
+
+# Builds the binary in GOPATH/bin. Changing into / first avoids
+# modifying the project's go.mod file.
+$(DOC_PLANTUML_GO):
+	cd / && go get github.com/acarlson99/plantuml-go

keps/sig-node/3063-dynamic-resource-allocation/components.puml

@@ -0,0 +1,47 @@
+@startuml
+!theme reddress-lightblue
+skinparam componentStyle rectangle
+
+cloud "resource driver" {
+  component "resource driver controller" as drivercontroller
+  component "resource kubelet plugin" as driverplugin
+}
+
+component Kubernetes {
+  component apiserver {
+      file Pod
+      file ResourceClaim
+      file PodScheduling
+  }
+  component scheduler {
+    component "resource plugin" as k8sresourceplugin
+  }
+  component "controller-manager" as controllermanager {
+    component "resource claim controller" as k8sresourceclaimcontroller
+  }
+  component kubelet {
+    component "plugin manager" as pluginmanager
+    component "resource manager" as resourcemanager
+  }
+}
+
+controllermanager -[hidden]> kubelet
+drivercontroller -[hidden]> driverplugin
+
+Pod <. ResourceClaim: owned by\n(if created from template)
+Pod <. PodScheduling: owned by
+
+
+Pod -u-> k8sresourceclaimcontroller: read claim template\nfrom Pod spec
+ResourceClaim <-u- k8sresourceclaimcontroller: create claim,\nclean up users
+ResourceClaim <-u-> kubelet
+k8sresourceplugin <-u-> PodScheduling
+
+Pod <--> scheduler
+ResourceClaim <--> k8sresourceplugin
+
+ResourceClaim <-> drivercontroller
+pluginmanager <-> driverplugin
+resourcemanager <-> driverplugin
+PodScheduling <-> drivercontroller
+@enduml

keps/sig-node/3063-dynamic-resource-allocation/kep.yaml

@@ -0,0 +1,46 @@
+title: dynamic resource allocation
+kep-number: 3063
+authors:
+  - "@pohly"
+owning-sig: sig-node
+participating-sigs:
+  - sig-scheduling
+status: provisional
+creation-date: 2021-05-17
+reviewers:
+  - "@ahg-g"
+  - "@alculquicondor"
+  - "@klueska"
+approvers:
+  - "@sig-node-leads"
+
+see-also:
+replaces:
+
+# The target maturity stage in the current dev cycle for this KEP.
+stage: alpha
+
+# The most recent milestone for which work toward delivery of this KEP has been
+# done. This can be the current (upcoming) milestone, if it is being actively
+# worked on.
+latest-milestone: "v1.25"
+
+# The milestone at which this feature was, or is targeted to be, at each stage.
+milestone:
+  alpha: "v1.25"
+  beta: "v1.27"
+  stable: "v1.29"
+
+feature-gates:
+  - name: DynamicResourceAllocation
+    components:
+      - kube-apiserver
+      - kube-controller-manager
+      - kube-scheduler
+      - kubelet
+disable-supported: true
+
+metrics:
+  - resource_controller_create_total
+  - resource_controller_create_failures_total
+  - resource controller workqueue with name="resource_claim"

keps/sig-node/3063-dynamic-resource-allocation/kubelet.puml

@@ -0,0 +1,12 @@
+@startuml
+!theme reddress-lightblue
+
+(*) --> "Pod scheduled to the node"
+-->[ **Kubelet** pick up the pod ] "Pod references resource claim"
+-->[ **Kubelet** calls NodePrepareResource(ResourceClaimUID, AllocationParameters)\l **Resource Plugin** creates CDI JSON file(s) ] "CDI JSON files created"
+-->[ **Resource Plugin** returns device Ids ] "Resource prepared"
+-->[ **Kubelet** passes device Ids to the CRI runtime\l **CRI runtime** uses CDI JSON file(s) to update OCI container config\l **CRI runtime** creates pod containers ] "All Pod containers created"
+-->[ ... ] "Pod runs"
+-->[ ... ] "Pod terminated"
+-->[ **Kubelet** calls NodeUnprepareResource\l **Resource plugin** removes CDI JSON file(s) ] (*)
+@enduml