Compact Cluster Support about ocp4-helpernode

[mransonwang]

@christianh814 I'm considering install ocp 4.5 with helper node, would you give some advice on how to modify DNS, HAproxy, DHCP manually, so the ocp 4.5 three nodes cluster could be installed smoothly, many thanks.

[christianh814]

@mransonwang

For now I would use "dummy" values for the workers. For example in your vars.yaml file, it can look like this.

---
disk: vda
helper:
  name: "helper"
  ipaddr: "192.168.7.77"
dns:
  domain: "example.com"
  clusterid: "ocp4"
  forwarder1: "8.8.8.8"
  forwarder2: "8.8.4.4"
dhcp:
  router: "192.168.7.1"
  bcast: "192.168.7.255"
  netmask: "255.255.255.0"
  poolstart: "192.168.7.10"
  poolend: "192.168.7.30"
  ipid: "192.168.7.0"
  netmaskid: "255.255.255.0"
bootstrap:
  name: "bootstrap"
  ipaddr: "192.168.7.20"
  macaddr: "52:54:00:60:72:67"
masters:
  - name: "master0"
    ipaddr: "192.168.7.21"
    macaddr: "52:54:00:e7:9d:67"
  - name: "master1"
    ipaddr: "192.168.7.22"
    macaddr: "52:54:00:80:16:23"
  - name: "master2"
    ipaddr: "192.168.7.23"
    macaddr: "52:54:00:d5:1c:39"
workers:
  - name: "worker0-dummy"
    ipaddr: "192.168.7.11"
    macaddr: "52:54:00:f4:26:a1"
  - name: "worker1-dummy"
    ipaddr: "192.168.7.12"
    macaddr: "52:54:00:82:90:00"

Run the playbook with these "dummy" entries.

Then, after the playbook is ran, edit the following files

• /var/named/zonefile.db Edit this file to remove all the worker entries
• /var/named/reverse.db Edit this file to remove all the worker entries
• /etc/dhcp/dhcpd.conf Edit this file to remove all the worker entries
• /var/lib/tftpboot/pxelinux.cfg/ You need to cd into this directory and remove the files that has the "dummy" MAC addresses
• /etc/haproxy/haproxy.cfg Edit this file and change the IP for the worker addresses to point to the master address

Then restart all services

systemctl restart named dhcpd tftp haproxy

Check the services

helpernodecheck services

⚠️ Note that tftp might appear "dead" (this is okay)

After creating the manifests, SKIP this step

$ sed -i 's/mastersSchedulable: true/mastersSchedulable: false/g' manifests/cluster-scheduler-02-config.yml # DON'T DO THIS STEP

Can you test this out? If it works for you I'll add it to the documentation as a stopgap until we get a 3 node "compact" cluster supported.

[mransonwang]

@christianh814 Yes, will try it

[mransonwang]

@christianh814 Just verified the 3 masters OCP installation, the installation looked fine. How ever, due to my environment resources limitation, some pods status were Preempting, I think it shoud not related with helper node

[christianh814]

Closed by #127