Skip to content

howto_test_offline.md

Latest commit

 

History

History
137 lines (95 loc) · 3.19 KB

howto_test_offline.md

File metadata and controls

137 lines (95 loc) · 3.19 KB

Running MicroShift Fully Offline

There are several steps needed to run MicroShift in a fully offline setting, with no network access at all. This document describes a process for setting up a host to test MicroShift with no network.

Image Management

Refer to Embedding MicroShift Container Images for Offline Deployments for instructions for preparing the ostree image with the necessary container images embedded when they cannot be pulled over the network.

Setup

The commands below are meant to be exeuted as the root user. Please login as root. Make sure the appropriate kubeconfig file is available in your environment so the oc client commands can run as expected.

Warning: These instructions involve disabling the network interface on the host. You will need to access the host directly, not via ssh. For VMs, virsh console is useful.

MicroShift Configuration

Kubernetes expects to have some basic network settings on the host for node identification and DNS. When real settings are not going to be available, alternatives need to be provided for robust disconnected operation.

  1. Start by preparing a VM or host in the normal way and installing the version of MicroShift you want to test.

  2. Stop MicroShift and clean up the OVN data.

/usr/bin/microshift-cleanup-data --ovn
  1. Determine the interface on the host that has the default route.
IFACE=$(ip route | grep default | awk 'NR==1 {print $5}')

(This is usually enp1s0 in a VM)

  1. Stop external network access by bringing the NIC down.
nmcli connection down $IFACE
nmcli connection modify $IFACE connection.autoconnect no
  1. Ensure the host has a persistent name
NAME=hostname.example.com
hostnamectl hostname $NAME
  1. Add a fake IP on the loopback interface.
IP=10.44.0.1
nmcli con add type loopback con-name stable-microshift ifname lo ip4 $IP/32
  1. Configure DNS to look at the local name server (note this should not be the same IP, it's on a different subnet).
nmcli conn modify stable-microshift ipv4.ignore-auto-dns yes
nmcli conn modify stable-microshift ipv4.dns "10.44.1.1"
  1. Add an entry for the node's hostname from step 5 in /etc/hosts using the IP from step 6.
echo “$IP $NAME” >> /etc/hosts

  1. Add the name from step 5 to the MicroShift config file as the node.hostnameOverride setting.

  2. Add the IP from step 6 to the MicroShift config file as the node.nodeIP setting.

  3. Restart microshift

systemctl enable --now microshift
  1. Verify it works
oc get pods -A

  1. Reboot the host

  2. Verify it works (watch for a while, because the initial pod information is cached from before the reboot and is not accurate)

oc get pods -A
  1. Re-enable the NIC
IFACE=$(ip route | grep default | awk 'NR==1 {print $5}')
nmcli connection up $IFACE
  1. Verify that MicroShift is still ok
oc get pods -A
  1. Set the NIC to autoconnect on startup
nmcli connection modify $IFACE connection.autoconnect yes

  1. Reboot the host

  2. Verify it works (watch for a while, because the initial pod information is cached from before the reboot and is not accurate)

watch oc get pods -A