X-Git-Url: https://gerrit.opnfv.org/gerrit/gitweb?a=blobdiff_plain;f=mcp%2Fscripts%2Flib.sh;h=02c23f365219da0ec1e17191bb44db2f4b3eebf8;hb=ec234a839dbd5eb8c615e1f6a3e34db64ef1c988;hp=8bdb5dd57110d663bac24e3697395935598b33b1;hpb=fa62907da62d0b8ae9e733a9729941065ceedf4d;p=fuel.git diff --git a/mcp/scripts/lib.sh b/mcp/scripts/lib.sh index 8bdb5dd57..02c23f365 100644 --- a/mcp/scripts/lib.sh +++ b/mcp/scripts/lib.sh @@ -1,4 +1,5 @@ #!/bin/bash -e +# shellcheck disable=SC2155,SC1001 ############################################################################## # Copyright (c) 2017 Mirantis Inc., Enea AB and others. # All rights reserved. This program and the accompanying materials @@ -11,7 +12,6 @@ # function generate_ssh_key { - # shellcheck disable=SC2155 local mcp_ssh_key=$(basename "${SSH_KEY}") local user=${USER} if [ -n "${SUDO_USER}" ] && [ "${SUDO_USER}" != 'root' ]; then @@ -32,7 +32,205 @@ function get_base_image { local image_dir=$2 mkdir -p "${image_dir}" - wget -P "${image_dir}" -N "${base_image}" + wget --progress=dot:giga -P "${image_dir}" -N "${base_image}" +} + +function __kernel_modules { + # Load mandatory kernel modules: loop, nbd + local image_dir=$1 + sudo modprobe loop + if sudo modprobe nbd max_part=8 || sudo modprobe -f nbd max_part=8; then + return 0 + fi + # CentOS (or RHEL family in general) do not provide 'nbd' out of the box + echo "[WARN] 'nbd' kernel module cannot be loaded!" + if [ ! -e /etc/redhat-release ]; then + echo "[ERROR] Non-RHEL system detected, aborting!" + echo "[ERROR] Try building 'nbd' manually or install it from a 3rd party." + exit 1 + fi + + # Best-effort attempt at building a non-maintaned kernel module + local __baseurl + local __subdir + local __uname_r=$(uname -r) + local __uname_m=$(uname -m) + if [ "${__uname_m}" = 'x86_64' ]; then + __baseurl='http://vault.centos.org/centos' + __subdir='Source/SPackages' + __srpm="kernel-${__uname_r%.${__uname_m}}.src.rpm" + else + __baseurl='http://vault.centos.org/altarch' + __subdir="Source/${__uname_m}/Source/SPackages" + # NOTE: fmt varies across releases (e.g. kernel-alt-4.11.0-44.el7a.src.rpm) + __srpm="kernel-alt-${__uname_r%.${__uname_m}}.src.rpm" + fi + + local __found='n' + local __versions=$(curl -s "${__baseurl}/" | grep -Po 'href="\K7\.[\d\.]+') + for ver in ${__versions}; do + for comp in os updates; do + local url="${__baseurl}/${ver}/${comp}/${__subdir}/${__srpm}" + if wget "${url}" -O "${image_dir}/${__srpm}" > /dev/null 2>&1; then + __found='y'; break 2 + fi + done + done + + if [ "${__found}" = 'n' ]; then + echo "[ERROR] Can't find the linux kernel SRPM for: ${__uname_r}" + echo "[ERROR] 'nbd' module cannot be built, aborting!" + echo "[ERROR] Try 'yum upgrade' or building 'nbd' krn module manually ..." + exit 1 + fi + + rpm -ivh "${image_dir}/${__srpm}" 2> /dev/null + mkdir -p ~/rpmbuild/{BUILD,BUILDROOT,RPMS,SOURCES,SPECS,SRPMS} + # shellcheck disable=SC2016 + echo '%_topdir %(echo $HOME)/rpmbuild' > ~/.rpmmacros + ( + cd ~/rpmbuild/SPECS + rpmbuild -bp --nodeps --target="${__uname_m}" kernel*.spec + cd ~/rpmbuild/BUILD/"${__srpm%.src.rpm}"/linux-* + sed -i 's/^.*\(CONFIG_BLK_DEV_NBD\).*$/\1=m/g' .config + # http://centosfaq.org/centos/nbd-does-not-compile-for-3100-514262el7x86_64 + if grep -Rq 'REQ_TYPE_DRV_PRIV' drivers/block; then + sed -i 's/REQ_TYPE_SPECIAL/REQ_TYPE_DRV_PRIV/g' drivers/block/nbd.c + fi + gunzip -c "/boot/symvers-${__uname_r}.gz" > Module.symvers + make prepare modules_prepare + make M=drivers/block -j + modinfo drivers/block/nbd.ko + sudo mkdir -p "/lib/modules/${__uname_r}/extra/" + sudo cp drivers/block/nbd.ko "/lib/modules/${__uname_r}/extra/" + ) + sudo depmod -a + sudo modprobe nbd max_part=8 || sudo modprobe -f nbd max_part=8 +} + +function mount_image { + local image=$1 + local image_dir=$2 + OPNFV_MNT_DIR="${image_dir}/ubuntu" + + # Find free nbd, loop devices + for dev in '/sys/class/block/nbd'*; do + if [ "$(cat "${dev}/size")" = '0' ]; then + OPNFV_NBD_DEV=/dev/$(basename "${dev}") + break + fi + done + OPNFV_LOOP_DEV=$(losetup -f) + OPNFV_MAP_DEV=/dev/mapper/$(basename "${OPNFV_NBD_DEV}")p1 + export OPNFV_MNT_DIR OPNFV_LOOP_DEV + [ -n "${OPNFV_NBD_DEV}" ] && [ -n "${OPNFV_LOOP_DEV}" ] || exit 1 + qemu-img resize "${image_dir}/${image}" 3G + sudo qemu-nbd --connect="${OPNFV_NBD_DEV}" --aio=native --cache=none \ + "${image_dir}/${image}" + sudo kpartx -av "${OPNFV_NBD_DEV}" + sleep 5 # /dev/nbdNp1 takes some time to come up + # Hardcode partition index to 1, unlikely to change for Ubuntu UCA image + if sudo growpart "${OPNFV_NBD_DEV}" 1; then + sudo kpartx -u "${OPNFV_NBD_DEV}" + sudo e2fsck -pf "${OPNFV_MAP_DEV}" + sudo resize2fs "${OPNFV_MAP_DEV}" + fi + # grub-update does not like /dev/nbd*, so use a loop device to work around it + sudo losetup "${OPNFV_LOOP_DEV}" "${OPNFV_MAP_DEV}" + mkdir -p "${OPNFV_MNT_DIR}" + sudo mount "${OPNFV_LOOP_DEV}" "${OPNFV_MNT_DIR}" + sudo mount -t proc proc "${OPNFV_MNT_DIR}/proc" + sudo mount -t sysfs sys "${OPNFV_MNT_DIR}/sys" + sudo mount -o bind /dev "${OPNFV_MNT_DIR}/dev" + sudo mkdir -p "${OPNFV_MNT_DIR}/run/resolvconf" + sudo cp /etc/resolv.conf "${OPNFV_MNT_DIR}/run/resolvconf" + echo "GRUB_DISABLE_OS_PROBER=true" | \ + sudo tee -a "${OPNFV_MNT_DIR}/etc/default/grub" + sudo sed -i -e 's/^\(GRUB_TIMEOUT\)=.*$/\1=1/g' -e 's/^GRUB_HIDDEN.*$//g' \ + "${OPNFV_MNT_DIR}/etc/default/grub" +} + +function apt_repos_pkgs_image { + local apt_key_urls=(${1//,/ }) + local all_repos=(${2//,/ }) + local pkgs_i=(${3//,/ }) + local pkgs_r=(${4//,/ }) + [ -n "${OPNFV_MNT_DIR}" ] || exit 1 + + # APT keys + if [ "${#apt_key_urls[@]}" -gt 0 ]; then + for apt_key in "${apt_key_urls[@]}"; do + sudo chroot "${OPNFV_MNT_DIR}" /bin/bash -c \ + "wget -qO - '${apt_key}' | apt-key add -" + done + fi + # Additional repositories + for repo_line in "${all_repos[@]}"; do + # ||deb|[arch=]||| + local repo=(${repo_line//|/ }) + [ "${#repo[@]}" -gt 5 ] || continue + # NOTE: Names and formatting are compatible with Salt linux.system.repo + cat <<-EOF | sudo tee "${OPNFV_MNT_DIR}/etc/apt/preferences.d/${repo[0]}" + + Package: * + Pin: release a=${repo[-2]} + Pin-Priority: ${repo[1]} + + EOF + echo "${repo[@]:2}" | sudo tee \ + "${OPNFV_MNT_DIR}/etc/apt/sources.list.d/${repo[0]}.list" + done + # Install packages + if [ "${#pkgs_i[@]}" -gt 0 ]; then + sudo DEBIAN_FRONTEND="noninteractive" \ + chroot "${OPNFV_MNT_DIR}" apt-get update + sudo DEBIAN_FRONTEND="noninteractive" FLASH_KERNEL_SKIP="true" \ + chroot "${OPNFV_MNT_DIR}" apt-get install -y "${pkgs_i[@]}" + fi + # Remove packages + if [ "${#pkgs_r[@]}" -gt 0 ]; then + sudo DEBIAN_FRONTEND="noninteractive" FLASH_KERNEL_SKIP="true" \ + chroot "${OPNFV_MNT_DIR}" apt-get purge -y "${pkgs_r[@]}" + fi + # Disable cloud-init metadata service datasource + sudo mkdir -p "${OPNFV_MNT_DIR}/etc/cloud/cloud.cfg.d" + echo "datasource_list: [ NoCloud, None ]" | sudo tee \ + "${OPNFV_MNT_DIR}/etc/cloud/cloud.cfg.d/95_real_datasources.cfg" +} + +function cleanup_mounts { + # Remove any mounts, loop and/or nbd devs created while patching base image + if [ -n "${OPNFV_MNT_DIR}" ] && [ -d "${OPNFV_MNT_DIR}" ]; then + if [ -f "${OPNFV_MNT_DIR}/boot/grub/grub.cfg" ]; then + # Grub thinks it's running from a live CD + sudo sed -i -e 's/^\s*set root=.*$//g' -e 's/^\s*loopback.*$//g' \ + "${OPNFV_MNT_DIR}/boot/grub/grub.cfg" + fi + sudo rm -f "${OPNFV_MNT_DIR}/run/resolvconf/resolv.conf" + sync + if mountpoint -q "${OPNFV_MNT_DIR}"; then + sudo umount -l "${OPNFV_MNT_DIR}" || true + fi + fi + if [ -n "${OPNFV_LOOP_DEV}" ] && \ + losetup "${OPNFV_LOOP_DEV}" 1>&2 > /dev/null; then + sudo losetup -d "${OPNFV_LOOP_DEV}" + fi + if [ -n "${OPNFV_NBD_DEV}" ]; then + sudo kpartx -d "${OPNFV_NBD_DEV}" || true + sudo qemu-nbd -d "${OPNFV_NBD_DEV}" || true + fi +} + +function cleanup_uefi { + # Clean up Ubuntu boot entry if cfg01, kvm nodes online from previous deploy + local cmd_str="ssh ${SSH_OPTS} ${SSH_SALT}" + [ ! "$(hostname)" = 'cfg01' ] || cmd_str='eval' + ${cmd_str} "sudo salt -C 'kvm* or cmp*' cmd.run \ + \"which efibootmgr > /dev/null 2>&1 && \ + efibootmgr | grep -oP '(?<=Boot)[0-9]+(?=.*ubuntu)' | \ + xargs -I{} efibootmgr --delete-bootnum --bootnum {}; \ + rm -rf /boot/efi/*\"" || true } function cleanup_vms { @@ -50,21 +248,69 @@ function cleanup_vms { function prepare_vms { local base_image=$1; shift local image_dir=$1; shift + local repos_pkgs_str=$1; shift # ^-sep list of repos, pkgs to install/rm local vnodes=("$@") + local image=base_image_opnfv_fuel.img + local vcp_image=${image%.*}_vcp.img + local _o=${base_image/*\/} + local _h=$(echo "${repos_pkgs_str}.$(md5sum "${image_dir}/${_o}")" | \ + md5sum | cut -c -8) + local _tmp + cleanup_uefi cleanup_vms get_base_image "${base_image}" "${image_dir}" - # shellcheck disable=SC2016 - envsubst '${SALT_MASTER},${CLUSTER_DOMAIN}' < \ - user-data.template > user-data.sh + IFS='^' read -r -a repos_pkgs <<< "${repos_pkgs_str}" + echo "[INFO] Lookup cache / build patched base image for fingerprint: ${_h}" + _tmp="${image%.*}.${_h}.img" + if [ "${image_dir}/${_tmp}" -ef "${image_dir}/${image}" ]; then + echo "[INFO] Patched base image found" + else + rm -f "${image_dir}/${image%.*}"* + if [[ ! "${repos_pkgs_str}" =~ ^\^+$ ]]; then + echo "[INFO] Patching base image ..." + cp "${image_dir}/${_o}" "${image_dir}/${_tmp}" + __kernel_modules "${image_dir}" + mount_image "${_tmp}" "${image_dir}" + apt_repos_pkgs_image "${repos_pkgs[@]:0:4}" + cleanup_mounts + else + echo "[INFO] No patching required, using vanilla base image" + ln -sf "${image_dir}/${_o}" "${image_dir}/${_tmp}" + fi + ln -sf "${image_dir}/${_tmp}" "${image_dir}/${image}" + fi + + # Create config ISO and resize OS disk image for each foundation node VM for node in "${vnodes[@]}"; do - # create/prepare images - ./create-config-drive.sh -k "$(basename "${SSH_KEY}").pub" -u user-data.sh \ - -h "${node}" "${image_dir}/mcp_${node}.iso" - cp "${image_dir}/${base_image/*\/}" "${image_dir}/mcp_${node}.qcow2" + if [[ "${node}" =~ ^(cfg01|mas01) ]]; then + user_data='user-data.mcp.sh' + else + user_data='user-data.admin.sh' + fi + ./create-config-drive.sh -k "$(basename "${SSH_KEY}").pub" \ + -u "${user_data}" -h "${node}" "${image_dir}/mcp_${node}.iso" + cp "${image_dir}/${image}" "${image_dir}/mcp_${node}.qcow2" qemu-img resize "${image_dir}/mcp_${node}.qcow2" 100G done + + # VCP VMs base image specific changes + if [[ ! "${repos_pkgs_str}" =~ \^{3}$ ]] && [ -n "${repos_pkgs[*]:4}" ]; then + echo "[INFO] Lookup cache / build patched VCP image for md5sum: ${_h}" + _tmp="${vcp_image%.*}.${_h}.img" + if [ "${image_dir}/${_tmp}" -ef "${image_dir}/${vcp_image}" ]; then + echo "[INFO] Patched VCP image found" + else + echo "[INFO] Patching VCP image ..." + cp "${image_dir}/${image}" "${image_dir}/${_tmp}" + __kernel_modules "${image_dir}" + mount_image "${_tmp}" "${image_dir}" + apt_repos_pkgs_image "${repos_pkgs[@]:4:4}" + cleanup_mounts + ln -sf "${image_dir}/${_tmp}" "${image_dir}/${vcp_image}" + fi + fi } function create_networks { @@ -72,15 +318,15 @@ function create_networks { # create required networks, including constant "mcpcontrol" # FIXME(alav): since we renamed "pxe" to "mcpcontrol", we need to make sure # we delete the old "pxe" virtual network, or it would cause IP conflicts. - # FIXME(alav): The same applies for "fuel1" virsh network. - for net in "fuel1" "pxe" "mcpcontrol" "${vnode_networks[@]}"; do + for net in "pxe" "mcpcontrol" "${vnode_networks[@]}"; do if virsh net-info "${net}" >/dev/null 2>&1; then virsh net-destroy "${net}" || true virsh net-undefine "${net}" fi # in case of custom network, host should already have the bridge in place - if [ -f "net_${net}.xml" ] && [ ! -d "/sys/class/net/${net}/bridge" ]; then - virsh net-define "net_${net}.xml" + if [ -f "virsh_net/net_${net}.xml" ] && \ + [ ! -d "/sys/class/net/${net}/bridge" ]; then + virsh net-define "virsh_net/net_${net}.xml" virsh net-autostart "${net}" virsh net-start "${net}" fi @@ -89,14 +335,15 @@ function create_networks { function create_vms { local image_dir=$1; shift + # vnode data should be serialized with the following format: + # ',,|,,[...]' IFS='|' read -r -a vnodes <<< "$1"; shift - local vnode_networks=("$@") # AArch64: prepare arch specific arguments local virt_extra_args="" if [ "$(uname -i)" = "aarch64" ]; then - # No Cirrus VGA on AArch64, use vga std - virt_extra_args="$virt_extra_args --video=vga" + # No Cirrus VGA on AArch64, use virtio instead + virt_extra_args="$virt_extra_args --video=virtio" fi # create vms with specified options @@ -104,10 +351,13 @@ function create_vms { IFS=',' read -r -a vnode_data <<< "${serialized_vnode_data}" # prepare network args - net_args=" --network network=mcpcontrol,model=virtio" - if [ "${vnode_data[0]}" = "mas01" ]; then - # MaaS node's 3rd interface gets connected to PXE/Admin Bridge + local vnode_networks=("$@") + if [[ "${vnode_data[0]}" =~ ^(cfg01|mas01) ]]; then + net_args=" --network network=mcpcontrol,model=virtio" + # 3rd interface gets connected to PXE/Admin Bridge (cfg01, mas01) vnode_networks[2]="${vnode_networks[0]}" + else + net_args=" --network bridge=${vnode_networks[0]},model=virtio" fi for net in "${vnode_networks[@]:1}"; do net_args="${net_args} --network bridge=${net},model=virtio" @@ -128,14 +378,12 @@ function create_vms { function update_mcpcontrol_network { # set static ip address for salt master node, MaaS node - # shellcheck disable=SC2155 local cmac=$(virsh domiflist cfg01 2>&1| awk '/mcpcontrol/ {print $5; exit}') - # shellcheck disable=SC2155 local amac=$(virsh domiflist mas01 2>&1| awk '/mcpcontrol/ {print $5; exit}') virsh net-update "mcpcontrol" add ip-dhcp-host \ - "" --live + "" --live --config [ -z "${amac}" ] || virsh net-update "mcpcontrol" add ip-dhcp-host \ - "" --live + "" --live --config } function start_vms { @@ -196,13 +444,92 @@ function wait_for { local total_attempts=$1; shift local cmdstr=$* local sleep_time=10 - echo "[NOTE] Waiting for cmd to return success: ${cmdstr}" + echo -e "\n[wait_for] Waiting for cmd to return success: ${cmdstr}" # shellcheck disable=SC2034 for attempt in $(seq "${total_attempts}"); do - # shellcheck disable=SC2015 - eval "${cmdstr}" && return 0 || true - echo -n '.'; sleep "${sleep_time}" + echo "[wait_for] Attempt ${attempt}/${total_attempts%.*} for: ${cmdstr}" + if [ "${total_attempts%.*}" = "${total_attempts}" ]; then + # shellcheck disable=SC2015 + eval "${cmdstr}" && echo "[wait_for] OK: ${cmdstr}" && return 0 || true + else + !(eval "${cmdstr}" || echo __fuel_wf_failure__) |& tee /dev/stderr | \ + grep -Eq '(Not connected|No response|__fuel_wf_failure__)' && \ + echo "[wait_for] OK: ${cmdstr}" && return 0 || true + fi + sleep "${sleep_time}" done + echo "[wait_for] ERROR: Failed after max attempts: ${cmdstr}" return 1 ) } + +function do_sysctl_cfg { + local _conf='/etc/sysctl.d/99-opnfv-fuel-bridge.conf' + # https://wiki.libvirt.org/page/Net.bridge.bridge-nf-call_and_sysctl.conf + if modprobe br_netfilter bridge; then + echo 'net.bridge.bridge-nf-call-arptables = 0' |& sudo tee "${_conf}" + echo 'net.bridge.bridge-nf-call-iptables = 0' |& sudo tee -a "${_conf}" + echo 'net.bridge.bridge-nf-call-ip6tables = 0' |& sudo tee -a "${_conf}" + # Some distros / sysadmins explicitly blacklist br_netfilter + sudo sysctl -q -p "${_conf}" || true + fi +} + +function get_nova_compute_pillar_data { + local value=$(salt -C 'I@nova:compute and *01*' pillar.get _param:"${1}" --out yaml | cut -d ' ' -f2) + if [ "${value}" != "''" ]; then + echo "${value}" + fi +} + +function do_templates() { + local git_repo_root=$1; shift + local image_dir=$1; shift + local target_lab=$1; shift + local target_pod=$1; shift + local lab_config_uri=$1; shift + local scenario_dir=${1:-} + + RECLASS_CLUSTER_DIR=$(cd "${git_repo_root}/mcp/reclass/classes/cluster"; pwd) + PHAROS_GEN_CFG="./pharos/config/utils/generate_config.py" + PHAROS_INSTALLER_ADAPTER="./pharos/config/installers/fuel/pod_config.yml.j2" + PHAROS_INSTALLER_NETMAP="$(dirname "${PHAROS_INSTALLER_ADAPTER}")/net_map.j2" + BASE_CONFIG_PDF="${lab_config_uri}/labs/${target_lab}/${target_pod}.yaml" + BASE_CONFIG_IDF="${lab_config_uri}/labs/${target_lab}/idf-${target_pod}.yaml" + LOCAL_PDF="${image_dir}/$(basename "${BASE_CONFIG_PDF}")" + LOCAL_IDF="${image_dir}/$(basename "${BASE_CONFIG_IDF}")" + LOCAL_PDF_RECLASS="${image_dir}/pod_config.yml" + + # Two-stage expansion, first stage handles pod_config and scenarios only + if [ -n "${scenario_dir}" ]; then + # Make sample PDF/IDF available via default lab-config (pharos submodule) + ln -sf "$(readlink -f "../config/labs/local")" "./pharos/labs/" + + # Expand scenario file and main reclass input (pod_config.yaml) based on PDF + if ! curl --create-dirs -o "${LOCAL_PDF}" "${BASE_CONFIG_PDF}"; then + notify_e "[ERROR] Could not retrieve PDF (Pod Descriptor File)!" + elif ! curl -o "${LOCAL_IDF}" "${BASE_CONFIG_IDF}"; then + notify_e "[ERROR] Could not retrieve IDF (Installer Descriptor File)!" + elif ! "${PHAROS_GEN_CFG}" -y "${LOCAL_PDF}" \ + -j "${PHAROS_INSTALLER_ADAPTER}" > "${LOCAL_PDF_RECLASS}"; then + notify_e "[ERROR] Could not convert PDF+IDF to reclass model input!" + fi + template_dirs="${scenario_dir}" + template_err_str='Could not convert j2 scenario definitions!' + else + # Expand reclass and virsh network templates based on PDF + IDF + printenv | \ + awk '/^(SALT|MCP|MAAS|CLUSTER).*=/ { gsub(/=/,": "); print }' >> "${LOCAL_PDF}" + template_dirs="${RECLASS_CLUSTER_DIR} virsh_net ./*j2" + template_err_str='Could not convert PDF to network definitions!' + fi + # shellcheck disable=SC2086 + find ${template_dirs} -name '*.j2' | while read -r tp; do + # Jinja2 import does not allow '..' directory traversal + ln -sf "$(readlink -f "${PHAROS_INSTALLER_NETMAP}")" "$(dirname "${tp}")" + if ! "${PHAROS_GEN_CFG}" -y "${LOCAL_PDF}" -j "${tp}" > "${tp%.j2}"; then + notify_e "[ERROR] ${template_err_str}" + fi + rm -f "$(dirname "${tp}")/$(basename "${PHAROS_INSTALLER_NETMAP}")" + done +}