data-plane-osd-cluster-options.md

Data Plane OSD cluster Options

Using an existing OSD Cluster

Any OSD cluster can be used by the service, it does not have to be created with the service itself. If you already have an existing OSD cluster, you will need to register it in the database so that it can be used by the service for incoming Kafka requests. The cluster must have been created with multizone availability.

Get the ID of your cluster (e.g. 1h95qckof3s31h3622d35d5eoqh5vtuq) using the CLI - Run ocm list clusters - The ID should be displayed under the ID column

NOTE: By default the auto scaling feature is disabled in development mode. You can enable it by passing --dataplane-cluster-scaling-type=auto when starting a built binary. Or changing the default value in default development environment flags file when using make run to start the API.

Using an existing OSD cluster with manual scaling enabled

You can manually add the cluster in the dataplane-cluster-configuration.yaml file.

A content of the file is:

• A list of clusters for kas fleet manager
• All clusters are created with cluster_provisioning status if they are not already in kas fleet manager
• All clusters in kas fleet manager DB already but are missing in the list will be marked as deprovisioning and will later be deleted.
• This list is ordered, any new cluster should be appended at the end.

From the cluster ID taken from step (1), use ocm CLI to get cluster details i.e name, id which is cluster_id, multi_az, cloud_provider and region.

ocm get /api/clusters_mgmt/v1/clusters/$ID | jq ' .name, .id, .multi_az, .cloud_provider.id, .region.id '

"cluster-name"
"1jp6kdr7k0sjbe5adck2prjur8f39378" # or a value matching your cluster ID 
true
"aws" # or any cloud provider
"us-east-1" # or any region

From the above command, the resulting dataplane-cluster-configuration.yaml file content will

clusters:
  - name: cluster-name
    cluster_id: 1jp6kdr7k0sjbe5adck2prjur8f39378
    cloud_provider: aws
    region: us-east-1
    multi_az: true
    schedulable: true # change this to false if you do not want the cluster to be schedulable
    kafka_instance_limit: 2 # change this to match any value of configuration
    supported_instance_type: "standard,developer" # could be "developer", "standard" or both i.e "standard,developer" or "developer,standard". Defaults to "standard,developer" if not set

Connecting to a standalone cluster

kas-fleet-manager allows provisioning of kafkas in an already preexisting standalone dataplane cluster. To do so, add the standalone cluster in the dataplane-cluster-configuration.yaml giving the:

• name of the kubeconfig context to use. This option is required and it has to be an existing name of a context in kubeconfig
• provider_type must be set to standalone. This is required to indicate that we are using a standalone
• cluster_dns This will be used to build kafka bootstrap url and to communicate with clusters e.g apps.example.dns.com. This option is required.
• cloud_provider the cloud provider where the standalone cluster is provisioned in
• region the cloud region where the standalone cluster is provisioned
• ... rest of the options

NOTE: kubeconfig path can be configured via the --kubeconfig CLI flag. Otherwise is defaults to $HOME/.kube/config

NOTE: OLM in the destination standalone cluster/s is a prerequisite to be able to install strimzi and kas-fleetshard operators

Configuring OSD Cluster Creation and AutoScaling

To configure auto scaling, use the --dataplane-cluster-scaling-type=auto. Once auto scaling is enabled this will activate the scaling up/down of compute nodes for existing clusters, dynamic creation and deletion of OSD dataplane clusters as explained in the dynamic scaling architecture documentation

NOTE: It's worth noting that turning on OSD autoscaling will result into the auto creation and deletion of data plane clusters. Have a look at dynamic scaling OSD creation and deletion architecture to understand more.

Registering an existing cluster in the Database

NOTE: This should only be done if auto scaling is enabled. If manual scaling is enabled, please follow the guide for using an existing cluster with manual scaling instead.

1. Register the cluster to the service

   • Run the following command to generate an INSERT command:

     make db/generate/insert/cluster CLUSTER_ID=<your-cluster-id>
     

   • Run the command generated above in your local database:
     • Login to the local database using make db/login
     • Execute SQL command from previous steps (kas-fleet-manager will populate any blank values when it reconciles)
     • Ensure that the clusters table is available.
       • Create the binary by running make binary
       • Run ./kas-fleet-manager migrate
     • Once the table is available, the generated INSERT command can now be run.

2. Ensure the cluster is ready to be used for incoming Kafka requests.

   • Take note of the status of the cluster, cluster_provisioned, when you registered it to the database in step 2. This means that the cluster has been successfully provisioned but still have remaining resources to set up (i.e. Strimzi operator installation).
   • Run the service using make run and let it reconcile resources required in order to make the cluster ready to be used by Kafka requests.
   • Once done, the cluster status in your database should have changed to ready. This means that the service can now assign this cluster to any incoming Kafka requests so that the service can process them.