SNO Cluster Deployment

Prerequisite

Before you start, please create an inventory csv file by following the example inventory-manifest.csv.example file with the hardware information of your hosts. The first row of the file indicates the columns. Please keep this line and start a new line with the inventories.

Then provide the addons you would like to enable or disabled in all hosts in the acm-agent-addon.json file.

Create SNO clusters

Generating manifests

On the Bastion machine, run script create-manifests.sh to create the SNO clusters:

./create-manifests.sh inventory/csv/path pull/secret/path private/key/path cluster-image-set

If the script is exited without errors, manifests should be created for each inventory. The generated manifests are under /clusters. Under /clusters, directories will be created for each SNO clusters, with the directory name being the cluster name. Before continuing to the next step, we recommend spot checking the manifests of one of the generated clusters.

SiteConfig Support

If you want to use SiteConfig + argocd, use the following command to generate manifests:

./create-siteconfig.sh inventory/csv/path pull/secret/path private/key/path cluster-image-set

You can apply manifests as before. Note: only namespace & secrets will be applied, and ZTP will not start until you trigger a SiteConfig creation with gitops.

You will need to copy SiteConfig in clusters/siteconfigs directory into the gitops application repo, which will create all the required CRs.

Required arguments

The 4 required arguments for the script are:

1. inventory/csv/path - file path to the inventory csv file. E.g. /User/root/files/inventory.csv
2. pull/secret/path - file path to the OCP pull secret file. E.g. /User/root/files/pull_secret.txt
3. private/key/path - file path to your SSH private key. E.g. /User/root/.ssh/id_rsa
4. cluster-image-set - name of the cluster image set to use in the AgentClusterInstall resource. E.g. ocp-4.8

Additional options

Set the following variables ahead of running the "Generating manifests" section.

• enable_workload_partitioning - set to "true" to enable workload partitioning. Default is "false".
• enable_static_ip - set to "true" to use static ip, and "false" to use DHCP. Default is "true". If set "false", the nmstateconfig CR will not be generated, and the following fields in the inventory csv will be ignored, and you can set any values (left blank) for these fields:
  • ip
  • prefix
  • gateway
  • dns_resolver
• use_ipv4 - set to "true" to use ipv4 instead of ipv6. Default is "false". Will only matter when using static ip. DHCP won't be affected. (example ipv4 inventory csv)

Monitoring managed SNO clusters

Before applying the manifests, you can start the monitoring script that measures the progress of the installation will also be started in the background. Its output will be saved in managedsnocluster.csv:

./monitor-deployment.sh kubeconfig/path [interval_seconds]

You can specify an optional parameter interval_seconds to this script which is number of seconds to wait between pull stats (default value is 10).

Applying manifests

You can now run the script to apply these manifests for all clusters:

./apply-manifests.sh kubeconfig/path start_index end_index inventory_file [NUM_CONCURRENT_APPLY] [INTERVAL_SECOND]

You can specify two optional parameters to this script: number of concurrent applies (default value is 100) and the number of seconds (default value is 15) to wait in between each batch of concurrent applies.

Debug

You can check the logs generated when applying manifests with the log file: cluster/cluster-name/logs.

You can also check the progress of the Assisted Installer installation of a particular SNO cluster by checking status:

curl `oc get infraenv -n cluster-name cluster-name \
  -ojsonpath='{.status.isoDownloadURL}' | \
  sed 's~downloads/image~events~g'`

Below is an example output: TODO(tgu)

If the SNO cluster is created successfully, you can download the kubeconfig of a range of SNO clusters with:

./getkubeconfigs.sh start_index end_index

The kubeconfig file will be downloaded to cluster/cluster-name/kubeconfig.