fleet

module
v0.6.8 Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Aug 16, 2023 License: MIT

README

Fleet

GitHub release (latest by date) Go Report Card Build Status GitHub go.mod Go version codecov

Fleet Join/Leave is a feature that allows a member cluster to join and leave a fleet(Hub) in the fleet control plane.

Quick Start

This guide has two parts,

  1. Steps to run agents on Kind clusters
  2. Steps to run agents on AKS clusters

For the kind clusters everything needed to run the agents is already predefined in the makefile no changes are needed. But for the AKS cluster we need to make some changes to files within the fleet repo.


Prerequisites

Steps to run agents on Kind clusters

1. Clone the repo to your machine

$ git clone https://github.com/Azure/fleet

2. Navigate to fleet directory

$ cd fleet

3. Set up hub and member kind clusters

The makefile uses kindest/node:v1.23.3 if your version is higher/lower use the following command to pull the image for v1.23.3

$ docker pull kindest/node:v1.23.3

then run,

$ make create-hub-kind-cluster create-member-kind-cluster

4. Build and load images to kind clusters (since we are testing locally we don't have access to fleet packages)

 $ OUTPUT_TYPE=type=docker make docker-build-hub-agent docker-build-member-agent docker-build-refresh-token
 $ make load-hub-docker-image load-member-docker-image

5. Install hub and member agents helm charts

$ make install-member-agent-helm

6. Apply memberCluster to the hub cluster

$ kind export kubeconfig --name hub-testing
$ kubectl apply -f examples/fleet_v1alpha1_membercluster.yaml 

7. Check to make sure the memberCluster & internalMemberCluster resources status have been updated to 'Joined'

$ kubectl describe membercluster kind-member-testing
Result
Name:         kind-member-testing
Namespace:    
Labels:       <none>
Annotations:  <none>
API Version:  fleet.azure.com/v1alpha1
Kind:         MemberCluster
Metadata:
  Creation Timestamp:  2022-07-08T01:42:35Z
  Generation:          1
  Resource Version:  868
  UID:               67cec7a4-3386-4fd5-9de2-20397e7b0029
Spec:
  Identity:
    API Group:             
    Kind:                  ServiceAccount
    Name:                  member-agent-sa
    Namespace:             fleet-system
  Lease Duration Seconds:  30
  State:                   Join
Status:
  Allocatable:
    Cpu:     8
    Memory:  2032532Ki
  Capacity:
    Cpu:     8
    Memory:  2032532Ki
  Conditions:
    Last Transition Time:  2022-07-08T01:42:36Z
    Message:               
    Observed Generation:   1
    Reason:                InternalMemberClusterHeartbeatReceived
    Status:                True
    Type:                  HeartbeatReceived
    Last Transition Time:  2022-07-08T01:42:36Z
    Message:               
    Observed Generation:   1
    Reason:                MemberClusterJoined
    Status:                True
    Type:                  Joined
Events:
  Type    Reason                        Age   From           Message
  ----    ------                        ----  ----           -------
  Normal  NamespaceCreated              81s   memberCluster  Namespace was created
  Normal  InternalMemberClusterCreated  81s   memberCluster  Internal member cluster was created
  Normal  RoleCreated                   81s   memberCluster  role was created
  Normal  RoleBindingCreated            81s   memberCluster  role binding was created
  Normal  MemberClusterJoined           81s   memberCluster  member cluster is joined


$ kubectl describe internalmembercluster kind-member-testing -n fleet-member-kind-member-testing
Result
Name:         kind-member-testing
Namespace:    fleet-kind-member-testing
API Version:  fleet.azure.com/v1alpha1
Kind:         InternalMemberCluster
Metadata:
  Creation Timestamp:  2022-07-08T01:42:36Z
  Generation:          1
  Owner References:
    API Version:     fleet.azure.com/v1alpha1
    Controller:      true
    Kind:            MemberCluster
    Name:            kind-member-testing
    UID:             67cec7a4-3386-4fd5-9de2-20397e7b0029
  Resource Version:  865
  UID:               1b544873-81b8-4bac-9624-d4208aa21fd1
Spec:
  Lease Duration Seconds:  60
  State:                   Join
Status:
  Allocatable:
    Cpu:     8
    Memory:  2032532Ki
  Capacity:
    Cpu:     8
    Memory:  2032532Ki
  Conditions:
    Last Transition Time:  2022-07-08T01:42:36Z
    Message:               
    Observed Generation:   1
    Reason:                InternalMemberClusterHeartbeatReceived
    Status:                True
    Type:                  HeartbeatReceived
    Last Transition Time:  2022-07-08T01:42:36Z
    Message:               
    Reason:                ReconcileSuccess
    Status:                True
    Type:                  Synced
    Last Transition Time:  2022-07-08T01:42:36Z
    Message:               
    Observed Generation:   1
    Reason:                InternalMemberClusterJoined
    Status:                True
    Type:                  Joined
    Last Transition Time:  2022-07-08T01:42:36Z
    Message:               
    Observed Generation:   1
    Reason:                InternalMemberClusterHealthy
    Status:                True
    Type:                  Healthy
Events:
  Type    Reason                                  Age                From                             Message
  ----    ------                                  ----               ----                             -------
  Normal  InternalMemberClusterHeartbeatReceived  8s (x6 over 4m8s)  InternalMemberClusterController  internal member cluster heartbeat received
  Normal  InternalMemberClusterJoined             8s (x6 over 4m8s)  InternalMemberClusterController  internal member cluster has joined
  Normal  InternalMemberClusterHealthy            8s (x6 over 4m8s)  InternalMemberClusterController  internal member cluster healthy


8. Change the state for memberCluster yaml file to be Leave and apply the change

$ kubectl apply -f examples/fleet_v1alpha1_membercluster.yaml 

9. Check to make sure the memberCluster resource status have been updated to 'Left'

$ kubectl describe membercluster kind-member-testing
Result
Name:         kind-member-testing
Namespace:    
Labels:       <none>
Annotations:  <none>
API Version:  fleet.azure.com/v1alpha1
Kind:         MemberCluster
Metadata:
  Creation Timestamp:  2022-07-08T01:42:35Z
  Generation:          2
  Resource Version:  1565
  UID:               67cec7a4-3386-4fd5-9de2-20397e7b0029
Spec:
  Identity:
    API Group:             
    Kind:                  ServiceAccount
    Name:                  member-agent-sa
    Namespace:             fleet-system
  Lease Duration Seconds:  30
  State:                   Leave
Status:
  Allocatable:
    Cpu:     8
    Memory:  2032532Ki
  Capacity:
    Cpu:     8
    Memory:  2032532Ki
  Conditions:
    Last Transition Time:  2022-07-08T01:42:36Z
    Message:               
    Observed Generation:   1
    Reason:                InternalMemberClusterHeartbeatReceived
    Status:                True
    Type:                  HeartbeatReceived
    Last Transition Time:  2022-07-08T01:49:10Z
    Message:               
    Observed Generation:   2
    Reason:                MemberClusterLeft
    Status:                False
    Type:                  Joined
    Last Transition Time:  2022-07-08T01:49:10Z
    Message:               
    Reason:                ReconcileSuccess
    Status:                True
    Type:                  Synced
Events:
  Type    Reason                            Age    From           Message
  ----    ------                            ----   ----           -------
  Normal  NamespaceCreated                  6m56s  memberCluster  Namespace was created
  Normal  InternalMemberClusterCreated      6m56s  memberCluster  Internal member cluster was created
  Normal  RoleCreated                       6m56s  memberCluster  role was created
  Normal  RoleBindingCreated                6m56s  memberCluster  role binding was created
  Normal  MemberClusterJoined               6m56s  memberCluster  member cluster is joined
  Normal  InternalMemberClusterSpecUpdated  22s    memberCluster  internal member cluster spec is marked as Leave
  Normal  NamespaceDeleted                  22s    memberCluster  namespace is deleted for member cluster
  Normal  MemberClusterLeft                 22s    memberCluster  member cluster has left


Cleanup

delete kind clusters setup

$ make clean-e2e-tests

Steps to run agents on AKS clusters


Before starting create a text file and have a list of variables and their associated values for anything with double quotes put it in this text file to keep track, since we need to make changes to most of the commands before using them.

List of variables that's supposed to be in your text file as you go through the commands,

  • "RegistryName"
  • "RegistryLoginServer"
  • "ResourceGroupName"
  • "PRINCIPAL_ID"
  • "CLIENT_ID"
  • "HUB_URL"
  • "MemberClusterCRName"
Prerequisites
  • Valid Azure subscription to create AKS clusters setup subscription
  • Resource Group under subscription setup resource group ("ResourceGroupName" name of the resource group created)
  • ACR inside the resource group to build & push docker images setup ACR ("RegistryLoginServer" login server field when you navigate to your registry on Azure portal, "RegistryName" name of Registry created)
1. Create hub cluster with AAD, RBAC enabled
$ az aks create --resource-group "ResourceGroupName" --name hubCluster --attach-acr "RegistryName" --node-count 1 --generate-ssh-keys --enable-aad --enable-azure-rbac
2. Create member cluster with managed identity enabled
$ az aks create --resource-group "ResourceGroupName" --name memberCluster --attach-acr "RegistryName" --node-count 1 --generate-ssh-keys --enable-managed-identity
3. Admin access to install helm charts and apply CRs
$ az aks get-credentials --resource-group "ResourceGroupName" --name hubCluster --admin
$ az aks get-credentials --resource-group "ResourceGroupName" --name memberCluster --admin
4. Switching contexts between clusters

switching contexts provides access to the corresponding clusters. The contexts are defined in a config file, it's usual file path is /Users/username/.kube/config

$ kubectl config use-context hubCluster-admin
$ kubectl config use-context memberCluster-admin
5. Login to ACR
$ az acr login -n "RegistryName"
6. Build and push docker images

In the Makefile which exists in fleet/Makefile and values.yaml files for both helm charts which exists in fleet/charts/hub-agent/values.yaml & fleet/charts/member-agent/values.yaml

make the necessary changes.

Makefile hub-agent/values.yaml member-agent/values.yaml

From the fleet directory run the following commands. This builds the docker images from local fleet directory and pushes the images to ACR.

$ make docker-build-hub-agent
$ make docker-build-member-agent
$ make docker-build-refresh-token
7. Install helm charts and CRs:

From the fleet directory run the following commands,

$ helm install hub-agent ./charts/hub-agent/

Each time we create an AKS cluster a resource group gets auto generated for us MC_ResourceGroupName_ClusterName_Location find the resource group and then go and click the agent pool MSI object and get the "PRINCIPAL_ID" which will be the name of the identity for the memberCluster CR, we can also find the "CLIENT_ID" here

copy the code below and navigate to fleet/examples/fleet_v1alpha1_membercluster.yaml, paste the code and replace the "PRINCIPAL_ID"
apiVersion: fleet.azure.com/v1alpha1
kind: MemberCluster
metadata:
  name: membercluster-sample
spec:
  state: Join
  identity:
    name: "PRINCIPAL_ID"
    kind: User
    namespace: fleet-system
    apiGroup: rbac.authorization.k8s.io
  leaseDurationSeconds: 30

then apply the CR,

$ kubectl apply -f ./examples/fleet_v1alpha1_membercluster.yaml

switch cluster context to member cluster and run, "CLIENT_ID" is clientId from the agent pool MSI object, "HUB_URL" can be found in the .kube/config file in the hub cluster context section.

$ helm install member-agent ./charts/member-agent/ --set azure.clientid="CLIENT_ID" --set config.provider=azure --set config.hubURL="HUB_URL" --set config.memberClusterName="MemberClusterCRName"

check events to see if member cluster has Joined.

$ kubectl describe membercluster "MemberClusterCRName"

After applying the member cluster CR the Join workflow completes and the member cluster gets marked as Joined with a condition.

To trigger the leave workflow change the state from Join to Leave in the member cluster CR or change the CR's spec to Leave in the fleet/examples/fleet_v1alpha1_membercluster.yaml and apply the CR again.

$ kubectl edit membercluster "MemberClusterCRName"

check events to see if member cluster has Left.

$ kubectl describe membercluster "MemberClusterCRName"
8. Verify the token file exists in the member cluster

switch cluster context to member cluster,

upgrade the AKS member cluster to use kubernetes version greater than 1.22 because we need ephemeral containers to access the token. we can check the kubernetes version for the AKS cluster by running,

$ kubectl get nodes -A

run this command to get possible upgrades for your cluster if kubernetes version for cluster is less than 1.23,

$ az aks get-upgrades --resource-group "ResourceGroupName" --name memberCluster --output table

then use a version greater than 1.22 replace the KUBERNETES_VERSION variable in command below this is done because ephemeral debug containers were introduced in 1.23 and run,

az aks upgrade --resource-group "ResourceGroupName" --name memberCluster --kubernetes-version KUBERNETES_VERSION

after the upgrade run,

$ kubectl debug node/nodeName -it --image=busybox

this opens a shell to access the files present within the node then run,

$ find . -name provider-token

which returns something similar to this,

Result
/host/var/lib/kubelet/pods/podName/volumes/kubernetes.io~empty-dir/provider-token

navigate to the directory to find a file called token use vim to open it.

9. Cleanup

Delete the resource group under which every resource was created. This might take some time.

Code of Conduct


This project has adopted the [Microsoft Open Source Code of Conduct][8]. For more information, see the [Code of Conduct FAQ][9] or contact [opencode@microsoft.com][19] with any additional questions or comments.

Contributing


Support


Azure fleet is an open source project that is [not covered by the Microsoft Azure support policy][10]. [Please search open issues here][11], and if your issue isn't already represented please [open a new one][12]. The project maintainers will respond to the best of their abilities.

Trademarks

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.

Directories

Path Synopsis
apis
cluster/v1beta1
+kubebuilder:object:generate=true +k8s:deepcopy-gen=package,register +groupName=cluster.kubernetes-fleet.io
+kubebuilder:object:generate=true +k8s:deepcopy-gen=package,register +groupName=cluster.kubernetes-fleet.io
placement/v1beta1
+kubebuilder:object:generate=true +k8s:deepcopy-gen=package,register +groupName=placement.kubernetes-fleet.io
+kubebuilder:object:generate=true +k8s:deepcopy-gen=package,register +groupName=placement.kubernetes-fleet.io
v1alpha1
+kubebuilder:object:generate=true +k8s:deepcopy-gen=package,register +groupName=fleet.azure.com
+kubebuilder:object:generate=true +k8s:deepcopy-gen=package,register +groupName=fleet.azure.com
cmd
hack
pkg
controllers/clusterresourceplacement
Package clusterresourceplacement features a controller to reconcile the clusterResourcePlacement changes.
Package clusterresourceplacement features a controller to reconcile the clusterResourcePlacement changes.
controllers/clusterresourceplacementwatcher
Package clusterresourceplacementwatcher features a controller to watch the clusterResourcePlacement changes.
Package clusterresourceplacementwatcher features a controller to watch the clusterResourcePlacement changes.
controllers/clustershedulingpolicysnapshot
Package clustershedulingpolicysnapshot features a controller to reconcile the clusterSchedulingPolicySnapshot object.
Package clustershedulingpolicysnapshot features a controller to reconcile the clusterSchedulingPolicySnapshot object.
controllers/rollout
Package rollout features a controller to do rollout.
Package rollout features a controller to do rollout.
controllers/workgenerator
Package workgenerator features a controller to generate work objects based on resource binding objects.
Package workgenerator features a controller to generate work objects based on resource binding objects.
scheduler
Package scheduler features the scheduler for Fleet workloads.
Package scheduler features the scheduler for Fleet workloads.
scheduler/clustereligibilitychecker
Package clustereligibilitychecker features a utility for verifying if a member cluster is eligible for resource placement.
Package clustereligibilitychecker features a utility for verifying if a member cluster is eligible for resource placement.
scheduler/framework
Package framework features the scheduler framework, which the scheduler runs to schedule a placement to most appropriate clusters.
Package framework features the scheduler framework, which the scheduler runs to schedule a placement to most appropriate clusters.
scheduler/framework/parallelizer
Package parallelizer features some utilities to help run tasks in parallel.
Package parallelizer features some utilities to help run tasks in parallel.
scheduler/framework/plugins/clusteraffinity
Package clusteraffinity features a scheduler plugin that enforces cluster affinity (if any) defined on a CRP.
Package clusteraffinity features a scheduler plugin that enforces cluster affinity (if any) defined on a CRP.
scheduler/framework/plugins/clustereligibility
Package clustereligibility features a scheduler plugin that filters out clusters that are not eligible for resource placement.
Package clustereligibility features a scheduler plugin that filters out clusters that are not eligible for resource placement.
scheduler/framework/plugins/sameplacementaffinity
Package sameplacementaffinity features a scheduler plugin that filters out any cluster that has been already scheduled/bounded to the resource placement and prefers the same cluster which has an obsolete binding.
Package sameplacementaffinity features a scheduler plugin that filters out any cluster that has been already scheduled/bounded to the resource placement and prefers the same cluster which has an obsolete binding.
scheduler/framework/plugins/topologyspreadconstraints
Package topologyspreadconstraints features a scheduler plugin that enforces the topology spread constraints (if any) defined on a CRP.
Package topologyspreadconstraints features a scheduler plugin that enforces the topology spread constraints (if any) defined on a CRP.
scheduler/framework/uniquename
package uniquename features some utilities that are used to generate unique names in use by the scheduler.
package uniquename features some utilities that are used to generate unique names in use by the scheduler.
scheduler/profile
Package profile holds the definition of a scheduling Profile.
Package profile holds the definition of a scheduling Profile.
scheduler/queue
Package queue features a scheduling queue, which keeps track of all placements for the scheduler to schedule.
Package queue features a scheduling queue, which keeps track of all placements for the scheduler to schedule.
scheduler/watchers/clusterresourceplacement
Package clusterresourceplacement features a controller that enqueues CRPs for the scheduler to process where the CRP is marked for deletion.
Package clusterresourceplacement features a controller that enqueues CRPs for the scheduler to process where the CRP is marked for deletion.
scheduler/watchers/clusterschedulingpolicysnapshot
Package clusterschedulingpolicysnapshot features a controller that enqueues CRPs for the scheduler to process where there is a change in their scheduling policy snapshots.
Package clusterschedulingpolicysnapshot features a controller that enqueues CRPs for the scheduler to process where there is a change in their scheduling policy snapshots.
scheduler/watchers/membercluster
Package membercluster features a controller that enqueues CRPs on member cluster changes.
Package membercluster features a controller that enqueues CRPs on member cluster changes.
utils/annotations
Package annotations provides the utils related to object annotations.
Package annotations provides the utils related to object annotations.
utils/condition
Package condition provides condition related utils.
Package condition provides condition related utils.
utils/labels
Package labels provides utils related to object labels.
Package labels provides utils related to object labels.
test
utils/controller
Package controller provides a fake controller for testing.
Package controller provides a fake controller for testing.
utils/keycollector
Package keycollector features a utility that is used for testing scheduler-related logic.
Package keycollector features a utility that is used for testing scheduler-related logic.

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL