CentOS 7 上稳定OpenShift-ansible（Origin）版本及多节点安装咨询

Hey there! Let me break this down for you clearly from my hands-on experience:

适用于CentOS 7的最新稳定版OpenShift Origin（OKD）及对应openshift-ansible版本 #

OpenShift Origin was officially renamed to OKD later on. The latest stable branch that fully supports CentOS 7 is OKD 3.11 — this is also the last major stable release with official CentOS 7 support (subsequent 4.x versions dropped CentOS 7 support and shifted to RHEL 8/Fedora-based distros). The matching latest stable openshift-ansible toolkit belongs to the 3.11.x series; you can grab its final release by cloning the official repo's release-3.11 branch:

git clone https://github.com/openshift/openshift-ansible.git
cd openshift-ansible
git checkout release-3.11

多节点安装实战经验 & 踩坑技巧 #

I've set up OKD 3.11 multi-node clusters multiple times, so here are some hard-earned tips to save you headaches:

Pre-install Must-Dos #

• Hardware & System Checks: Each node needs at least 2 CPU cores and 8GB RAM (master nodes should have 4 cores + 16GB RAM for stability), plus a minimum 50GB disk space (container images and logs eat up space fast). Make sure your CentOS 7 is version 7.6 or newer; update first and reboot:

  yum update -y && reboot
  

• Disable Troublesome Services:
  • Turn off firewalld (it often causes compatibility issues; use iptables instead or configure rules manually):

    systemctl stop firewalld && systemctl disable firewalld
    

  • Disable SELinux (both temporarily and permanently — permission issues here will ruin your day):

    setenforce 0
    sed -i 's/^SELINUX=.*/SELINUX=disabled/' /etc/selinux/config
    

  • Turn off swap (Kubernetes/OKD requires this strictly):

    swapoff -a
    sed -i '/swap/d' /etc/fstab
    

• Network & Hostname Setup: All nodes must resolve each other's hostnames reliably — editing /etc/hosts is the most foolproof way (don't rely on flaky DNS). Also, sync system time across all nodes with chronyd (a time difference over 5 minutes will break the cluster):

  yum install chronyd -y
  systemctl start chronyd && systemctl enable chronyd
  

Ansible Configuration & Installation Phase #

• Control Node Prep: Pick one node (can be a master or a dedicated control node) to install Ansible 2.7.x — OKD 3.11 only works with this version, don't install a newer one:

  yum install ansible-2.7.* -y
  

• Inventory File Tuning: This is the core of the setup! Double-check every entry. For small clusters, it's fine to reuse master nodes as etcd nodes. Here's a sample inventory snippet:

  [OSEv3:children]
  masters
  nodes
  etcd
  
  [masters]
  master01.example.com ansible_user=root
  
  [nodes]
  master01.example.com openshift_node_group_name='node-config-master'
  node01.example.com openshift_node_group_name='node-config-compute'
  node02.example.com openshift_node_group_name='node-config-compute'
  
  [etcd]
  master01.example.com
  

  Key notes: Specify ansible_user as root or a user with full sudo access, and make sure openshift_node_group_name matches the correct node role.
• Run Pre-check Playbook: Always run this before the actual installation to catch issues early:

  ansible-playbook playbooks/prerequisites.yml
  

  It checks dependencies, open ports, and system configs — fix any errors before proceeding, don't force the installation.
• Start Installation: Run the cluster deployment playbook with -v to get detailed output, which helps with troubleshooting:

  ansible-playbook playbooks/deploy_cluster.yml -v
  

  If it fails, don't just retry blindly. Check logs under /var/log/openshift-ansible/ first, then dig into specific node issues (e.g., use docker logs <pod-name> on the problematic node if a container won't start).

Post-install Validation & Optimization #

• Cluster Health Check: On the master node, run these commands to confirm everything is running:

  oc get nodes
  oc get pods --all-namespaces
  

  Ensure all nodes show Ready status and all core pods are Running.
• Common Issue Fixes:
  • Image pull failures: Due to network restrictions, configure Docker image acceleration on all nodes beforehand or pull required images manually.
  • Nodes can't join the cluster: Check if core ports (6443, 2379, 8443, etc.) are open between nodes, and verify hostname resolution works correctly.
  • Permission errors: Make sure directories like /var/lib/docker and /var/lib/openshift have the right permissions (owned by root and docker group).

Hope these tips help you get your cluster up and running smoothly! If you hit specific errors, feel free to share log snippets for further troubleshooting.

内容的提问来源于stack exchange，提问作者Adigun Gbenga