The steps for performing a UPI-based install are outlined here. Several CloudFormation templates are provided to assist in completing these steps or to help model your own. You are also free to create the required resources through other methods; the CloudFormation templates are just an example.
Create an install configuration as for the usual approach:
$ openshift-install create install-config
? SSH Public Key /home/user_id/.ssh/id_rsa.pub
? Platform aws
? Region us-east-2
? Base Domain example.com
? Cluster Name openshift
? Pull Secret [? for help]The IPI-based installer creates an encrypted AMI by default. If you wish to have an encrypted AMI for UPI-based installs, you will need to create it directly. See CoreOS bootimages for more information about bootimages, including how to find the AMI identifiers.
You will make an encrypted copy of the AMI according to the AWS documentation.
With the new AMI, you can customize the install-config created on the previous step to override the default. Additionally, you would pass it to the templates or EC2 launch instance commands according to how you intend to launch your hosts.
We'll be providing the control-plane and compute machines ourselves, so edit the resulting install-config.yaml to set replicas to 0 for the compute pool:
python -c '
import yaml;
path = "install-config.yaml";
data = yaml.load(open(path));
data["compute"][0]["replicas"] = 0;
open(path, "w").write(yaml.dump(data, default_flow_style=False))'Use a staged install to make some adjustments which are not exposed via the install configuration.
$ openshift-install create manifests
INFO Consuming "Install Config" from target directoryRemove the control-plane Machines and compute MachineSets, because we'll be providing those ourselves and don't want to involve the machine-API operator:
$ rm -f openshift/99_openshift-cluster-api_master-machines-*.yaml openshift/99_openshift-cluster-api_worker-machineset-*.yaml 99_openshift-machine-api_master-control-plane-machine-set.yamlYou are free to leave the compute MachineSets in if you want to create compute machines via the machine API, but if you do you may need to update the various references (subnet, etc.) to match your environment.
Currently emptying the compute pools makes control-plane nodes schedulable. But due to a Kubernetes limitation, router pods running on control-plane nodes will not be reachable by the ingress load balancer. Update the scheduler configuration to keep router pods and other workloads off the control-plane nodes:
python -c '
import yaml;
path = "manifests/cluster-scheduler-02-config.yml"
data = yaml.load(open(path));
data["spec"]["mastersSchedulable"] = False;
open(path, "w").write(yaml.dump(data, default_flow_style=False))'The ingress operator is able to manage DNS records on your behalf. Depending on whether you want operator-managed DNS or user-managed DNS, you can choose to identify the internal DNS zone or remove DNS zones from the DNS configuration.
If you want the ingress operator to manage DNS records on your behalf, adjust the privateZone section in the DNS configuration to identify the zone it should use.
By default it will use a kubernetes.io/cluster/{infrastructureName}: owned tag, but that tag is only appropriate if openshift-install destroy cluster should remove the zone.
For user-provided zones, you can remove tags completely and use the zone ID instead:
python -c '
import yaml;
path = "manifests/cluster-dns-02-config.yml";
data = yaml.load(open(path));
del data["spec"]["privateZone"]["tags"];
data["spec"]["privateZone"]["id"] = "Z21IZ5YJJMZ2A4";
open(path, "w").write(yaml.dump(data, default_flow_style=False))'If you don't want the ingress operator to manage DNS records on your behalf, remove the privateZone and publicZone sections from the DNS configuration:
python -c '
import yaml;
path = "manifests/cluster-dns-02-config.yml";
data = yaml.load(open(path));
del data["spec"]["publicZone"];
del data["spec"]["privateZone"];
open(path, "w").write(yaml.dump(data, default_flow_style=False))'If you do so, you'll need to add ingress DNS records manually later on.
For disconnected clusters, Openshift has to be configured not to manage DNS, otherwise the ingress operator will try to contact the STS endpoint "sts.amazon.com" directly as opposed to the configured VPC endpoint for the cluster.
Now we can create the bootstrap Ignition configs:
$ openshift-install create ignition-configsAfter running the command, several files will be available in the directory.
$ tree
.
├── auth
│ └── kubeconfig
├── bootstrap.ign
├── master.ign
├── metadata.json
└── worker.ignMany of the operators and functions within OpenShift rely on tagging AWS resources. By default, Ignition
generates a unique cluster identifier comprised of the cluster name specified during the invocation of the installer
and a short string known internally as the infrastructure name. These values are seeded in the initial manifests within
the Ignition configuration. To use the output of the default, generated
ignition-configs extracting the internal infrastructure name is necessary.
An example of a way to get this is below:
$ jq -r .infraID metadata.json
openshift-vw9j6
You may create a VPC with various desirable characteristics for your situation (VPN, route tables, etc.). The VPC configuration and a CloudFormation template is provided here.
A created VPC via the template or manually should approximate a setup similar to this:
The DNS and load balancer configuration within a CloudFormation template is provided here. It uses a public hosted zone and creates a private hosted zone similar to the IPI installation method. It also creates load balancers, listeners, as well as hosted zone and subnet tags the same way as the IPI installation method. This template can be run multiple times within a single VPC and in combination with the VPC template provided above.
It is needed to create a TCP load balancer for ports 6443 (the Kubernetes API and its extensions) and 22623 (Ignition configurations for new machines). The targets will be the master nodes. Port 6443 must be accessible to both clients external to the cluster and nodes within the cluster. Port 22623 must be accessible to nodes within the cluster.
For the cluster name identified earlier in Create Ignition Configs, you must create a DNS entry which resolves to your created load balancer.
The entry api.$clustername.$domain should point to the external load balancer and api-int.$clustername.$domain should point to the internal load balancer.
The security group and IAM configuration within a CloudFormation template is provided here. Run this template to get the minimal and permanent set of security groups and IAM roles needed for an operational cluster. It can also be inspected for the current set of required rules to facilitate manual creation.
The bootstrap launch and other necessary, temporary security group plus IAM configuration and a CloudFormation
template is provided here. Upload your generated bootstrap.ign
file to an S3 bucket in your account and run this template to get a bootstrap node along with a predictable clean up of
the resources when complete. It can also be inspected for the set of required attributes via manual creation.
The master launch and other necessary DNS entries for etcd are provided within a CloudFormation template here. Run this template to get three master nodes. It can also be inspected for the set of required attributes needed for manual creation of the nodes, DNS entries and load balancer configuration.
$ bin/openshift-install wait-for bootstrap-complete
INFO Waiting up to 30m0s for the Kubernetes API at https://api.test.example.com:6443...
INFO API v1.12.4+c53f462 up
INFO Waiting up to 30m0s for the bootstrap-complete event...At this point, you should delete the bootstrap resources. If using the CloudFormation template, you would delete the stack created for the bootstrap to clean up all the temporary resources.
You may create compute nodes by launching individual EC2 instances discretely or by automated processes outside the cluster (e.g. Auto Scaling Groups). You can also take advantage of the built in cluster scaling mechanisms and the machine API in OpenShift, as mentioned above. In this example, we'll manually launch instances via the CloudFormatio template here. You can launch a CloudFormation stack to manage each individual compute node (you should launch at least two for a high-availability ingress router). A similar launch configuration could be used by outside automation or AWS auto scaling groups.
The CSR requests for client and server certificates for nodes joining the cluster will need to be approved by the administrator. You can view them with:
$ oc get csr
NAME AGE REQUESTOR CONDITION
csr-8b2br 15m system:serviceaccount:openshift-machine-config-operator:node-bootstrapper Approved,Issued
csr-8vnps 15m system:serviceaccount:openshift-machine-config-operator:node-bootstrapper Approved,Issued
csr-b96j4 25s system:node:ip-10-0-52-215.us-east-2.compute.internal Approved,Issued
csr-bfd72 5m26s system:node:ip-10-0-50-126.us-east-2.compute.internal Pending
csr-c57lv 5m26s system:node:ip-10-0-95-157.us-east-2.compute.internal Pending
...Administrators should carefully examine each CSR request and approve only the ones that belong to the nodes created by them. CSRs can be approved by name, for example:
oc adm certificate approve csr-bfd72If you removed the DNS Zone configuration earlier, you'll need to manually create some DNS records pointing at the ingress load balancer.
You can create either a wildcard *.apps.{baseDomain}. or specific records (more on the specific records below).
You can use A, CNAME, alias, etc. records, as you see fit.
For example, you can create wildcard alias records by retrieving the ingress load balancer status:
$ oc -n openshift-ingress get service router-default
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
router-default LoadBalancer 172.30.62.215 ab37f072ec51d11e98a7a02ae97362dd-240922428.us-east-2.elb.amazonaws.com 80:31499/TCP,443:30693/TCP 5mThen find the hosted zone ID for the load balancer (or use this table):
$ aws elb describe-load-balancers | jq -r '.LoadBalancerDescriptions[] | select(.DNSName == "ab37f072ec51d11e98a7a02ae97362dd-240922428.us-east-2.elb.amazonaws.com").CanonicalHostedZoneNameID'
Z3AADJGX6KTTL2And finally, add the alias records to your private and public zones:
$ aws route53 change-resource-record-sets --hosted-zone-id "${YOUR_PRIVATE_ZONE}" --change-batch '{
> "Changes": [
> {
> "Action": "CREATE",
> "ResourceRecordSet": {
> "Name": "\\052.apps.your.cluster.domain.example.com",
> "Type": "A",
> "AliasTarget":{
> "HostedZoneId": "Z3AADJGX6KTTL2",
> "DNSName": "ab37f072ec51d11e98a7a02ae97362dd-240922428.us-east-2.elb.amazonaws.com.",
> "EvaluateTargetHealth": false
> }
> }
> }
> ]
> }'
$ aws route53 change-resource-record-sets --hosted-zone-id "${YOUR_PUBLIC_ZONE}" --change-batch '{
> "Changes": [
> {
> "Action": "CREATE",
> "ResourceRecordSet": {
> "Name": "\\052.apps.your.cluster.domain.example.com",
> "Type": "A",
> "AliasTarget":{
> "HostedZoneId": "Z3AADJGX6KTTL2",
> "DNSName": "ab37f072ec51d11e98a7a02ae97362dd-240922428.us-east-2.elb.amazonaws.com.",
> "EvaluateTargetHealth": false
> }
> }
> }
> ]
> }'If you prefer to add explicit domains instead of using a wildcard, you can create entries for each of the cluster's current routes:
$ oc get --all-namespaces -o jsonpath='{range .items[*]}{range .status.ingress[*]}{.host}{"\n"}{end}{end}' routes
oauth-openshift.apps.your.cluster.domain.example.com
console-openshift-console.apps.your.cluster.domain.example.com
downloads-openshift-console.apps.your.cluster.domain.example.com
alertmanager-main-openshift-monitoring.apps.your.cluster.domain.example.com
grafana-openshift-monitoring.apps.your.cluster.domain.example.com
prometheus-k8s-openshift-monitoring.apps.your.cluster.domain.example.com$ bin/openshift-install wait-for install-complete
INFO Waiting up to 30m0s for the cluster to initialize...Also, you can observe the running state of your cluster pods:
$ oc get pods --all-namespaces
NAMESPACE NAME READY STATUS RESTARTS AGE
kube-system etcd-member-ip-10-0-3-111.us-east-2.compute.internal 1/1 Running 0 35m
kube-system etcd-member-ip-10-0-3-239.us-east-2.compute.internal 1/1 Running 0 37m
kube-system etcd-member-ip-10-0-3-24.us-east-2.compute.internal 1/1 Running 0 35m
openshift-apiserver-operator openshift-apiserver-operator-6d6674f4f4-h7t2t 1/1 Running 1 37m
openshift-apiserver apiserver-fm48r 1/1 Running 0 30m
openshift-apiserver apiserver-fxkvv 1/1 Running 0 29m
openshift-apiserver apiserver-q85nm 1/1 Running 0 29m
...
openshift-service-ca-operator openshift-service-ca-operator-66ff6dc6cd-9r257 1/1 Running 0 37m
openshift-service-ca apiservice-cabundle-injector-695b6bcbc-cl5hm 1/1 Running 0 35m
openshift-service-ca configmap-cabundle-injector-8498544d7-25qn6 1/1 Running 0 35m
openshift-service-ca service-serving-cert-signer-6445fc9c6-wqdqn 1/1 Running 0 35m
openshift-service-catalog-apiserver-operator openshift-service-catalog-apiserver-operator-549f44668b-b5q2w 1/1 Running 0 32m
openshift-service-catalog-controller-manager-operator openshift-service-catalog-controller-manager-operator-b78cr2lnm 1/1 Running 0 31m