talos_upgrade.sh

#!/bin/bash

set -uo pipefail

version=""

CEPH_HEALTH_TIMEOUT="15m"
NODE_HEALTH_TIMEOUT="15m"
NODE_UPGRADE_TIMEOUT="15m"

function usage() {
    echo "${0}"
    echo "     -h             - Show Usage"
    echo "     -v [version]   - (Required) Specify version for upgrade"
    exit 1
}

if [[ ${#} -eq 0 ]]; then
   usage
fi

optstring=":hv:"

while getopts ${optstring} arg; do
  case ${arg} in
    v)
      version="${OPTARG}"
      ;;
    h)
      usage
      exit 1
      ;;
    \?)
      echo "Invalid option: -${OPTARG}."
      echo
      usage
      exit 1
      ;;
  esac
done

if [ -z "${version}" ]; then
    echo 'No version specified'
    echo
    usage
    exit 1
fi

for machine in $(kubectl get nodes -o jsonpath='{.items[*].metadata.name}'); do
    machine_ip=$(kubectl get node ${machine} -o jsonpath='{.status.addresses[?(@.type=="InternalIP")].address}')
    echo "Upgrading ${machine}"

    echo -n "  Checking version... "
    machine_version=$(kubectl get node ${machine} -o jsonpath='{.metadata.labels.feature\.node\.kubernetes\.io\/system-os_release\.VERSION_ID}')
    if [ "${machine_version}" = "${version}" ]; then
        echo -e "already upgraded\n\n---"
        continue
    else
        echo "needs upgrade"
    fi

    echo -n "  Waiting for all nodes to be Ready...            "
    kubectl wait --for=condition=Ready nodes --all --timeout=${NODE_HEALTH_TIMEOUT} &>/dev/null
    if [ $? -eq 0 ]; then
        echo "ok"
    else
        echo "failed"
        exit 1
    fi

    echo -n "  Waiting for Ceph cluster to be healthy...       "
    kubectl --namespace rook-ceph wait --timeout=${CEPH_HEALTH_TIMEOUT} --for=jsonpath='{.status.ceph.health}=HEALTH_OK' cephcluster/rook-ceph &>/dev/null
    if [ $? -eq 0 ]; then
        echo "ok"
    else
        echo "failed"
        exit 1
    fi

    # Trigger Upgrade
    echo -n "  Triggering upgrade...       "
    talosctl upgrade --image ghcr.io/siderolabs/installer:${version} --nodes ${machine_ip} &>/dev/null
    if [ $? -eq 0 ]; then
        echo "ok"
    else
        echo "failed"
        exit 1
    fi

    echo -n "  Waiting for node upgrade to complete... "
    kubectl wait node ${machine} --timeout=${NODE_UPGRADE_TIMEOUT} --for=jsonpath="{.metadata.labels.feature\.node\.kubernetes\.io\/system-os_release\.VERSION_ID}=${version}" &>/dev/null
    if [ $? -eq 0 ]; then
        echo "ok"
    else
        echo "failed"
        exit 1
    fi

    echo -e "\n---"
done