The Terraform VMware Cloud Director Supplier v3.11.0 now helps putting in and managing Container Service Extension (CSE) 4.1, with a brand new set of enhancements, the brand new vcd_rde_behavior_invocation information supply and up to date guides for VMware Cloud
Director customers to deploy the required elements.
On this weblog put up, we can be putting in CSE 4.1 in an current VCD and creating and managing a TKGm cluster.
Making ready the set up
To start with, we should ensure that all of the conditions listed within the Terraform VCD Supplier documentation are met. CSE 4.1 requires at the very least VCD 10.4.2, we will verify our VCD model within the popup that reveals up by clicking the About possibility inside the assistance “(?)” button subsequent to our username within the prime proper nook:
Test that you simply even have ALB controllers obtainable to be consumed from VMware Cloud Director, because the created clusters require them for load-balancing functions.
Step 1: Putting in the conditions
Step one of the set up mimics the UI wizard step by which conditions are created:
We are going to do that precise step programmatically with Terraform. To try this, let’s clone the terraform-provider-vcd repository so we will obtain the required schemas, entities, and examples:
git clone https://github.com/vmware/terraform-provider-vcd.git
cd terraform-provider-vcd
git checkout v3.11.0
cd examples/container-service-extension/v4.1/set up/step1
If we open 3.11-cse-install-2-cse-server-prerequisites.tf we will see that these configuration recordsdata create all of the RDE framework elements that CSE makes use of to work, consuming the schemas which are hosted within the GitHub repository, plus all of the rights and roles which are wanted. We gained’t customise something inside these recordsdata, as they create the identical objects because the UI wizard step proven within the above screenshot, which doesn’t enable customization both.
Now we open 3.11-cse-install-3-cse-server-settings.tf, this one is equal to the next UI wizard step:
We are able to observe that the UI wizard permits us to set some configuration parameters, and if we glance to terraform.tfvars.instance we are going to observe that the requested configuration values match.
Earlier than making use of all of the Terraform configuration recordsdata which are obtainable on this folder, we are going to rename terraform.tfvars.instance to terraform.tfvars, and we are going to set the variables with appropriate values. The defaults that we will see in variables.tf and terraform.tfvars.instance match with these of the UI wizard, which must be good for CSE 4.1. In our case, our VMware Cloud Director has full Web entry, so we’re not setting any customized Docker registry or certificates right here.
We also needs to consider that the terraform.tfvars.instance is asking for a username and password to create a person that can be used to provision API tokens for the CSE Server to run. We additionally depart these as they’re, as we just like the "cse_admin" username.
As soon as we overview the configuration, we will safely full this step by working:
terraform init
terraform apply
The plan ought to show all the weather which are going to be created. We full the operation (by writing sure to the immediate) so step one of the set up is completed. This may be simply checked within the UI as now the wizard doesn’t ask us to finish this step, as a substitute, it reveals the CSE Server configuration we simply utilized:
Step 2: Configuring VMware Cloud Director and working the CSE Server
We transfer to the following step, which is situated at examples/container-service-extension/v4.1/set up/step2 of our cloned repository.
cd examples/container-service-extension/v4.1/set up/step2
This step is probably the most customizable one, because it relies on our particular wants. Ideally, because the CSE documentation implies, there must be two Organizations: Options Group and Tenant Group, with Web entry so all of the required Docker pictures and packages could be downloaded (or with entry to an inner Docker registry if we had chosen a customized registry within the earlier step).
We are able to examine the completely different recordsdata obtainable and alter every little thing that doesn’t match with our wants. For instance, if we already had the Group VDCs created, we might change from utilizing assets to utilizing information sources as a substitute.
In our case, the VMware Cloud Director equipment the place we’re putting in CSE 4.1 is empty, so we have to create every little thing from scratch. That is what the recordsdata on this folder do, they create a primary and minimal set of elements to make CSE 4.1 work.
Identical as earlier than, we rename terraform.tfvars.instance to terraform.tfvars and examine the file contents so we will set the proper configuration. As we talked about, organising the variables of this step relies on our wants and the way we need to arrange the networking, the NSX ALB, and which TKGm OVAs we need to present to our tenants. We also needs to bear in mind that some constraints must be met, just like the VM Sizing Insurance policies which are required for CSE to work being revealed to the VDCs, so let’s learn and perceive the set up information for that objective.
As soon as we overview the configuration, we will full this step by working:
terraform init
terraform apply
Now we must always overview that the plan is appropriate and matches to what we need to obtain. It ought to create the 2 required Organizations, our VDCs, and most significantly, the networking configuration ought to enable Web visitors to retrieve the required packages for the TKGm clusters to be provisioned with out points (do not forget that within the earlier step, we didn’t set any inner registry nor certificates). We full the operation (by writing sure to the immediate) so the second step of the set up is completed.
We are able to additionally double-check that every little thing is appropriate within the UI, or do a connectivity take a look at by deploying a VM and utilizing the console to ping an outside-world web site.
Cluster creation with Terraform
Provided that we’ve completed the set up course of and we nonetheless have the cloned repository from the earlier steps, we transfer to examples/container-service-extension/v4.1/cluster.
cd examples/container-service-extension/v4.1/cluster
The cluster is created by the configuration file 3.11-cluster-creation.tf, by additionally utilizing the RDE framework. We encourage the readers to verify each the vcd_rde documentation and the cluster administration information earlier than continuing, because it’s necessary to understand how this useful resource works in Terraform, and most significantly, how CSE 4.1 makes use of it.
We are going to open 3.11-cluster-creation.tf and examine it, to right away see that it makes use of the JSON template situated at examples/container-service-extension/v4.1/entities/tkgmcluster.json.template. That is the payload that the CSE 4.1 RDE requires to initialize a TKGm cluster. We are able to customise this JSON to our wants, for instance, we are going to take away the defaultStorageClassOptions block from it as we gained’t use storage in our clusters.
The preliminary JSON template tkgmcluster.json.template appears to be like like this now:
{
“apiVersion”: “capvcd.vmware.com/v1.1”,
“form”: “CAPVCDCluster”,
“identify”: “${identify}”,
“metadata”: {
“identify”: “${identify}”,
“orgName”: “${org}”,
“web site”: “${vcd_url}”,
“virtualDataCenterName”: “${vdc}”
},
“spec”: {
“vcdKe”: {
“isVCDKECluster”: true,
“markForDelete”: ${delete},
“forceDelete”: ${force_delete},
“autoRepairOnErrors”: ${auto_repair_on_errors},
“safe”: {
“apiToken”: “${api_token}”
}
},
“capiYaml”: ${capi_yaml}
}
}
There’s nothing else that we will customise there, so we depart it like that.
The following factor that we discover is that we’d like a legitimate CAPVCD YAML, we will obtain it from right here. We’ll deploy a v1.25.7 Tanzu cluster, so we obtain this one to start out getting ready it.
We open it with our editor and add the required snippets as acknowledged in the documentation. We begin with the form: Cluster blocks which are required by the CSE Server to provision clusters:
apiVersion: cluster.x-k8s.io/v1beta1
form: Cluster
metadata:
identify: ${CLUSTER_NAME}
namespace: ${TARGET_NAMESPACE}
labels: # We add this block
cluster-role.tkg.tanzu.vmware.com/administration: “”
tanzuKubernetesRelease: ${TKR_VERSION}
tkg.tanzu.vmware.com/cluster-name: ${CLUSTER_NAME}
annotations: # We add this block
TKGVERSION: ${TKGVERSION}
# …
We added the 2 labels and annotations blocks, with the required placeholders TKR_VERSION, CLUSTER_NAME, and TKGVERSION. These placeholders are used to set the values by way of Terraform configuration.
Now we add the Machine Well being Test block, which is able to enable to make use of one of many new highly effective options of CSE 4.1, that remediates nodes in failed standing by changing them, enabling cluster self-healing:
apiVersion: cluster.x-k8s.io/v1beta1
form: MachineHealthCheck
metadata:
identify: ${CLUSTER_NAME}
namespace: ${TARGET_NAMESPACE}
labels:
clusterctl.cluster.x-k8s.io: “”
clusterctl.cluster.x-k8s.io/transfer: “”
spec:
clusterName: ${CLUSTER_NAME}
maxUnhealthy: ${MAX_UNHEALTHY_NODE_PERCENTAGE}%
nodeStartupTimeout: ${NODE_STARTUP_TIMEOUT}s
selector:
matchLabels:
cluster.x-k8s.io/cluster-name: ${CLUSTER_NAME}
unhealthyConditions:
– sort: Prepared
standing: Unknown
timeout: ${NODE_UNKNOWN_TIMEOUT}s
– sort: Prepared
standing: “False”
timeout: ${NODE_NOT_READY_TIMEOUT}s
—
Discover that the timeouts have an s because the values launched throughout set up have been in seconds. If we hadn’t put the worth in seconds, or we put the worth like 15m, we will take away the s suffix from these block choices.
Let’s add the final elements, that are most related when specifying customized certificates through the set up course of. In form: KubeadmConfigTemplate we should add the preKubeadmCommands and useExperimentalRetryJoin blocks beneath the spec > customers part:
preKubeadmCommands:
– mv /and many others/ssl/certs/custom_certificate_*.crt
/usr/native/share/ca-certificates && update-ca-certificates
useExperimentalRetryJoin: true
In form: KubeadmControlPlane we should add the preKubeadmCommands and controllerManager blocks contained in the kubeadmConfigSpec part:
preKubeadmCommands:
– mv /and many others/ssl/certs/custom_certificate_*.crt
/usr/native/share/ca-certificates && update-ca-certificates
controllerManager:
extraArgs:
enable-hostpath-provisioner: “true”
As soon as it’s accomplished, the ensuing YAML must be much like the one already supplied within the examples/cluster folder, cluster-template-v1.25.7.yaml, because it makes use of the identical model of Tanzu and has all of those additions already launched. It is a good train to verify whether or not our YAML is appropriate earlier than continuing additional.
After we overview the crafted YAML, let’s create a tenant person with the Kubernetes Cluster Creator function. This person can be required to provision clusters:
information “vcd_global_role” “k8s_cluster_author” {
identify = “Kubernetes Cluster Creator”
}
useful resource “vcd_org_user” “cluster_author” {
identify = “cluster_author”
password = “dummyPassword” # This one must be in all probability a wise variable and a bit safer.
function = information.vcd_global_role.k8s_cluster_author.identify
}
Now, we will full the customization of the configuration file 3.11-cluster-creation.tf by renaming terraform.tfvars.instance to terraform.tfvars and configuring the parameters of our cluster. Let’s verify ours:
vcd_url = “https://…”
cluster_author_user = “cluster_author”
cluster_author_password = “dummyPassword”
cluster_author_token_file = “cse_cluster_author_api_token.json”
k8s_cluster_name = “instance”
cluster_organization = “tenant_org”
cluster_vdc = “tenant_vdc”
cluster_routed_network = “tenant_net_routed”
control_plane_machine_count = “1”
worker_machine_count = “1”
control_plane_sizing_policy = “TKG small”
control_plane_placement_policy = “”””
control_plane_storage_profile = “*”
worker_sizing_policy = “TKG small”
worker_placement_policy = “”””
worker_storage_profile = “*”
disk_size = “20Gi”
tkgm_catalog = “tkgm_catalog”
tkgm_ova_name = “ubuntu-2004-kube-v1.25.7+vmware.2-tkg.1-8a74b9f12e488c54605b3537acb683bc”
pod_cidr = “100.96.0.0/11”
service_cidr = “100.64.0.0/13”
tkr_version = “v1.25.7—vmware.2-tkg.1”
tkg_version = “v2.2.0”
auto_repair_on_errors = true
We are able to discover that control_plane_placement_policy = """", that is to keep away from errors after we don’t need to use a VM Placement Coverage. We are able to verify that the downloaded CAPVCD YAML forces us to put double quotes on this worth when it’s not used.
The tkr_version and tkg_version values have been obtained from the already supplied in the documentation.
As soon as we’re pleased with the completely different choices, we apply the configuration:
terraform init
terraform apply
Now we must always overview the plan as a lot as potential to forestall errors. It ought to create the vcd_rde useful resource with the weather we supplied.
We full the operation (by writing sure to the immediate) so the cluster ought to begin getting created. We are able to monitor the method both in UI or with the 2 outputs supplied for instance:
locals
output “computed_k8s_cluster_status” {
worth = native.has_status && !native.being_deleted ? native.k8s_cluster_computed[“status”][“vcdKe”][“state”] : null
}
output “computed_k8s_cluster_events” {
worth = native.has_status && !native.being_deleted ? native.k8s_cluster_computed[“status”][“vcdKe”][“eventSet”] : null
}
Then we will do terraform refresh as many instances as we would like, to observe the occasions with:
terraform output computed_k8s_cluster_status
terraform output computed_k8s_cluster_events
As soon as computed_k8s_cluster_status states provisioned, this step can be completed and the cluster can be prepared to make use of. Let’s retrieve the Kubeconfig, which in CSE 4.1 is completed utterly in a different way than in 4.0, as we’re required to invoke a Habits to get it. In 3.11-cluster-creation.tf we will see a commented part that has a vcd_rde_behavior_invocation information supply. If we uncomment these and do one other terraform apply, we must always be capable of get the Kubeconfig by working
terraform output kubeconfig
We are able to reserve it to a file to start out interacting with our cluster and kubectl.
Cluster replace
Instance use case: we realized that our cluster is simply too small, so we have to scale it up. We’ll arrange 3 employee nodes.
To replace it, we have to ensure that it’s in provisioned standing. For that, we will use the identical mechanism that we used when the cluster creation began:
terraform output computed_k8s_cluster_status
This could show provisioned. If that’s the case, we will proceed with the replace.
As with the cluster creation, we first want to grasp how the vcd_rde useful resource works to keep away from errors, so it’s inspired to verify each the vcd_rde documentation and the cluster administration information earlier than continuing. The necessary thought is that we should replace the input_entity argument with the data that CSE saves within the computed_entity attribute, in any other case, we might break the cluster.
To try this, we will use the next output that can return the computed_entity attribute:
output “computed_k8s_cluster” {
worth = vcd_rde.k8s_cluster_instance.computed_entity # References the created cluster
}
Then we run this command to reserve it to a file for a greater studying:
|
|
terraform output –json computed_k8s_cluster > computed.json |
Let’s open computed.json for inspection. We are able to simply see that it appears to be like just about the identical as tkgmcluster.json.template however with the addition of a giant "standing" object that incorporates very important details about the cluster. This have to be despatched again on updates, so we copy the entire "standing" object as it’s and we place it within the unique tkgmcluster.json.template.
After that, we will change worker_machine_count = 1 to worker_machine_count = 3 within the current terraform.tfvars, to finish the replace course of with:
terraform apply
Now it’s essential to confirm and ensure that the output plan reveals that the "standing" is being added to the input_entity payload. If that’s not the case, we must always cease the operation instantly and verify what went unsuitable. If "standing" is seen within the plan as being added, you’ll be able to full the replace operation by writing sure to the immediate.
Cluster deletion
The primary thought of deleting a TKGm cluster is that we must always not use terraform destroy for that, even when that’s the first thought we take note of. The reason being that the CSE Server creates lots of components (VMs, Digital Companies, and many others) that might be in an “orphan” state if we simply delete the cluster RDE. We have to let the CSE Server do the cleanup for us.
For that matter, the vcd_rde current in 3.11-cluster-creation.tf incorporates two particular arguments, that mimic the deletion possibility from UI:
delete = false # Make this true to delete the cluster
force_delete = false # Make this true to forcefully delete the cluster
To set off an asynchronous deletion course of we must always change them to true and execute terraform apply to carry out an replace. We should additionally introduce the newest "standing" object to the tkgmcluster.json.template when making use of, just about like within the replace situation described within the earlier part.
Ultimate ideas
We hope you loved the method of putting in CSE 4.1 in your VMware Cloud Director equipment. For a greater understanding of the method, please learn the present set up and cluster administration guides.
