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 knowledge supply and up to date guides for VMware Cloud
Director customers to deploy the required parts.

On this weblog put up, we will likely be putting in CSE 4.1 in an present VCD and creating and managing a TKGm cluster.

Getting ready the set up

To start with, we should be sure that all of the stipulations listed within the Terraform VCD Supplier documentation are met. CSE 4.1 requires a minimum of VCD 10.4.2, we are able to test our VCD model within the popup that exhibits up by clicking the About choice inside the assistance “(?)” button subsequent to our username within the prime proper nook:

Examine 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 stipulations

Step one of the set up mimics the UI wizard step during which stipulations are created:

We’ll do that precise step programmatically with Terraform. To try this, let’s clone the terraform-provider-vcd repository so we are able to obtain the required schemas, entities, and examples:

If we open 3.11-cse-install-2-cse-server-prerequisites.tf we are able to see that these configuration recordsdata create all of the RDE framework parts that CSE makes use of to work, consuming the schemas which might be hosted within the GitHub repository, plus all of the rights and roles which might be wanted. We received’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 permit 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’ll observe that the requested configuration values match.

Earlier than making use of all of the Terraform configuration recordsdata which might be obtainable on this folder, we’ll rename terraform.tfvars.instance to terraform.tfvars, and we’ll set the variables with appropriate values. The defaults that we are able to see in variables.tf and terraform.tfvars.instance match with these of the UI wizard, which ought to 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 take into consideration that the terraform.tfvars.instance is asking for a username and password to create a consumer that will likely 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 evaluate the configuration, we are able to safely full this step by operating:

The plan ought to show all the weather which might be 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 an alternative, it exhibits the CSE Server configuration we simply utilized:

Step 2: Configuring VMware Cloud Director and operating the CSE Server

We transfer to the subsequent step, which is positioned at examples/container-service-extension/v4.1/set up/step2 of our cloned repository.

This step is probably the most customizable one, because it will depend on our particular wants. Ideally, because the CSE documentation implies, there ought to be two Organizations: Options Group and Tenant Group, with Web entry so all of the required Docker photos 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 totally different recordsdata obtainable and alter every part that doesn’t match with our wants. For instance, if we already had the Group VDCs created, we may change from utilizing sources to utilizing knowledge sources as an alternative.

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 part from scratch. That is what the recordsdata on this folder do, they create a fundamental and minimal set of parts to make CSE 4.1 work.

Similar as earlier than, we rename terraform.tfvars.instance to terraform.tfvars and examine the file contents so we are able to set the right configuration. As we talked about, establishing the variables of this step will depend 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 remember that some constraints have to be met, just like the VM Sizing Insurance policies which might be required for CSE to work being printed to the VDCs, so let’s learn and perceive the set up information for that goal.

As soon as we evaluate the configuration, we are able to full this step by operating:

Now we should always evaluate 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 permit 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 part 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 got 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.

The cluster is created by the configuration file 3.11-cluster-creation.tf, by additionally utilizing the RDE framework. We encourage the readers to test 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’ll open 3.11-cluster-creation.tf and examine it, to right away see that it makes use of the JSON template positioned 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’ll take away the defaultStorageClassOptions block from it as we received’t use storage in our clusters.

The preliminary JSON template tkgmcluster.json.template seems like this now:

There’s nothing else that we are able to customise there, so we depart it like that.

The following factor that we discover is that we’d like a legitimate CAPVCD YAML, we are able to obtain it from right here. We’ll deploy a v1.25.7 Tanzu cluster, so we obtain this one to begin making ready it.

We open it with our editor and add the required snippets as acknowledged in the documentation. We begin with the type: Cluster blocks which might be required by the CSE Server to provision clusters:

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 through Terraform configuration.

Now we add the Machine Well being Examine block, which can permit 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:

Discover that the timeouts have an s because the values launched throughout set up had been in seconds. If we hadn’t put the worth in seconds, or we put the worth like 15m, we are able to take away the s suffix from these block choices.

Let’s add the final components, that are most related when specifying customized certificates in the course of the set up course of. In type: KubeadmConfigTemplate we should add the preKubeadmCommands and useExperimentalRetryJoin blocks underneath the spec > customers part:

In type: KubeadmControlPlane we should add the preKubeadmCommands and controllerManager blocks contained in the kubeadmConfigSpec part:

As soon as it’s accomplished, the ensuing YAML ought to be just like the one already offered 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. This can be a good train to test whether or not our YAML is appropriate earlier than continuing additional.

After we evaluate the crafted YAML, let’s create a tenant consumer with the Kubernetes Cluster Writer function. This consumer will likely be required to provision clusters:

Now, we are able to 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 test ours:

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 test that the downloaded CAPVCD YAML forces us to put double quotes on this worth when it isn’t used.

The tkr_version and tkg_version values had been obtained from the already offered in the documentation.

As soon as we’re pleased with the totally different choices, we apply the configuration:

Now we should always evaluate the plan as a lot as attainable to forestall errors. It ought to create the vcd_rde useful resource with the weather we offered.
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 offered for example:

Then we are able to do terraform refresh as many instances as we would like, to observe the occasions with:

As soon as computed_k8s_cluster_status states provisioned, this step will likely be completed and the cluster will likely be prepared to make use of. Let’s retrieve the Kubeconfig, which in CSE 4.1 is finished 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 are able to see a commented part that has a vcd_rde_behavior_invocation knowledge supply. If we uncomment these and do one other terraform apply, we should always be capable to get the Kubeconfig by operating

We are able to put it aside to a file to begin interacting with our cluster and kubectl.

Cluster replace

Instance use case: we realized that our cluster is just too small, so we have to scale it up. We’ll arrange 3 employee nodes.

To replace it, we have to make sure that it’s in provisioned standing. For that, we are able to use the identical mechanism that we used when the cluster creation began:

This could show provisioned. If that’s the case, we are able to 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 test each the vcd_rde documentation and the cluster administration information earlier than continuing. The necessary concept is that we should replace the input_entity argument with the knowledge that CSE saves within the computed_entity attribute, in any other case, we may break the cluster.

To try this, we are able to use the next output that may return the computed_entity attribute:

Then we run this command to reserve it to a file for a greater studying:

Let’s open computed.json for inspection. We are able to simply see that it seems just about the identical as tkgmcluster.json.template however with the addition of a giant "standing" object that comprises important details about the cluster. This should 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 are able to change worker_machine_count = 1 to worker_machine_count = 3 within the present terraform.tfvars, to finish the replace course of with:

Now it’s essential to confirm and make sure that the output plan exhibits that the "standing" is being added to the input_entity payload. If that isn’t the case, we should always cease the operation instantly and test what went incorrect. If "standing" is seen within the plan as being added, you possibly can full the replace operation by writing sure to the immediate.

Cluster deletion

The primary concept of deleting a TKGm cluster is that we should always not use terraform destroy for that, even when that’s the first concept we bear in mind. The reason being that the CSE Server creates loads of components (VMs, Digital Providers, and so on) 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 comprises two particular arguments, that mimic the deletion choice from UI:

To set off an asynchronous deletion course of we should always change them to true and execute terraform apply to carry out an replace. We should additionally introduce the latest "standing" object to the tkgmcluster.json.template when making use of, just about like within the replace state of affairs 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.