X-Git-Url: https://gerrit.opnfv.org/gerrit/gitweb?a=blobdiff_plain;f=mcp%2Fscripts%2Flib.sh;h=5f1275dc5fd7474a326922cd1d96eac74d08576f;hb=f870328e38c59375b2b19b1667a4baba0ce31bdb;hp=982cc37fd41927ef596170ad1cf8f32327ec9870;hpb=a4abff17c7b71408762353bb1ef52d9ffbec2fb7;p=fuel.git diff --git a/mcp/scripts/lib.sh b/mcp/scripts/lib.sh index 982cc37fd..5f1275dc5 100644 --- a/mcp/scripts/lib.sh +++ b/mcp/scripts/lib.sh @@ -1,7 +1,7 @@ #!/bin/bash -e -# shellcheck disable=SC2155,SC1001 +# shellcheck disable=SC2155,SC1001,SC2015,SC2128 ############################################################################## -# Copyright (c) 2017 Mirantis Inc., Enea AB and others. +# Copyright (c) 2018 Mirantis Inc., Enea AB and others. # All rights reserved. This program and the accompanying materials # are made available under the terms of the Apache License, Version 2.0 # which accompanies this distribution, and is available at @@ -32,7 +32,81 @@ 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=$1 + test -e /dev/loop-control || sudo modprobe loop + if sudo modprobe nbd max_part=8 || sudo modprobe -f nbd max_part=8; then + return 0 + fi + if [ -e /dev/nbd0 ]; then return 0; fi # nbd might be inbuilt + # CentOS (or RHEL family in general) do not provide 'nbd' out of the box + echo "[WARN] 'nbd' kernel module cannot be loaded!" + if [ ! -e /etc/redhat-release ]; then + echo "[ERROR] Non-RHEL system detected, aborting!" + echo "[ERROR] Try building 'nbd' manually or install it from a 3rd party." + exit 1 + fi + + # Best-effort attempt at building a non-maintaned kernel module + local __baseurl + local __subdir + local __uname_r=$(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 { @@ -40,7 +114,6 @@ function mount_image { local image_dir=$2 OPNFV_MNT_DIR="${image_dir}/ubuntu" - sudo modprobe nbd loop # Find free nbd, loop devices for dev in '/sys/class/block/nbd'*; do if [ "$(cat "${dev}/size")" = '0' ]; then @@ -48,7 +121,7 @@ function mount_image { break fi done - OPNFV_LOOP_DEV=$(losetup -f) + OPNFV_LOOP_DEV=$(sudo 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 @@ -56,10 +129,11 @@ function mount_image { 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 -yf "${OPNFV_MAP_DEV}" + sudo e2fsck -pf "${OPNFV_MAP_DEV}" sudo resize2fs "${OPNFV_MAP_DEV}" fi # grub-update does not like /dev/nbd*, so use a loop device to work around it @@ -140,7 +214,7 @@ function cleanup_mounts { fi fi if [ -n "${OPNFV_LOOP_DEV}" ] && \ - losetup "${OPNFV_LOOP_DEV}" 1>&2 > /dev/null; then + sudo losetup "${OPNFV_LOOP_DEV}" 1>&2 > /dev/null; then sudo losetup -d "${OPNFV_LOOP_DEV}" fi if [ -n "${OPNFV_NBD_DEV}" ]; then @@ -152,6 +226,7 @@ function cleanup_mounts { function cleanup_uefi { # Clean up Ubuntu boot entry if cfg01, kvm nodes online from previous deploy local cmd_str="ssh ${SSH_OPTS} ${SSH_SALT}" + ping -c 1 -w 1 "${SALT_MASTER}" || return 0 [ ! "$(hostname)" = 'cfg01' ] || cmd_str='eval' ${cmd_str} "sudo salt -C 'kvm* or cmp*' cmd.run \ \"which efibootmgr > /dev/null 2>&1 && \ @@ -168,7 +243,8 @@ function cleanup_vms { for node in $(virsh list --name --all | grep -P '\w{3}\d{2}'); do virsh domblklist "${node}" | awk '/^.da/ {print $2}' | \ xargs --no-run-if-empty -I{} sudo rm -f {} - virsh undefine "${node}" --remove-all-storage --nvram + virsh undefine "${node}" --remove-all-storage --nvram || \ + virsh undefine "${node}" --remove-all-storage done } @@ -178,57 +254,139 @@ function prepare_vms { 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}" "${image_dir}" + IFS='^' read -r -a repos_pkgs <<< "${repos_pkgs_str}" - rm -f "${image_dir}/${image%.*}"* - if [[ ! "${repos_pkgs_str}" =~ ^\^+$ ]]; then - IFS='^' read -r -a repos_pkgs <<< "${repos_pkgs_str}" - cp "${image_dir}/${base_image/*\/}" "${image_dir}/${image}" - mount_image "${image}" "${image_dir}" - apt_repos_pkgs_image "${repos_pkgs[@]:0:4}" - cleanup_mounts + echo "[INFO] Lookup cache / build patched base image for fingerprint: ${_h}" + _tmp="${image%.*}.${_h}.img" + if [ "${image_dir}/${_tmp}" -ef "${image_dir}/${image}" ]; then + echo "[INFO] Patched base image found" else - ln -sf "${image_dir}/${base_image/*\/}" "${image_dir}/${image}" + rm -f "${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 - # CWD should be - # shellcheck disable=SC2016 - envsubst '${SALT_MASTER},${CLUSTER_DOMAIN}' < \ - user-data.template > user-data.sh - # Create config ISO and resize OS disk image for each foundation node VM for node in "${vnodes[@]}"; do - ./create-config-drive.sh -k "$(basename "${SSH_KEY}").pub" -u user-data.sh \ - -h "${node}" "${image_dir}/mcp_${node}.iso" + 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_${node}.qcow2" 100G + # Prepare dedicated drive for cinder on cmp nodes + if [[ "${node}" =~ ^(cmp) ]]; then + qemu-img create "${image_dir}/mcp_${node}_storage.qcow2" 100G + fi done # VCP VMs base image specific changes if [[ ! "${repos_pkgs_str}" =~ \^{3}$ ]] && [ -n "${repos_pkgs[*]:4}" ]; then - mount_image "${image}" "${image_dir}" - apt_repos_pkgs_image "${repos_pkgs[@]:4:4}" - cleanup_mounts - ln -sf "${image_dir}/${image}" "${image_dir}/${image%.*}_vcp.img" + 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 jumpserver_pkg_install { + local req_type=$1 + if [ -n "$(command -v apt-get)" ]; then + pkg_type='deb'; pkg_cmd='sudo apt-get install -y' + else + pkg_type='rpm'; pkg_cmd='sudo yum install -y --skip-broken' + fi + eval "$(parse_yaml "./requirements_${pkg_type}.yaml")" + for section in 'common' "$(uname -i)"; do + section_var="${req_type}_${section}[*]" + pkg_list+=" ${!section_var}" + done + # shellcheck disable=SC2086 + ${pkg_cmd} ${pkg_list} +} + +function jumpserver_check_requirements { + # shellcheck disable=SC2178 + local vnodes=$1; shift + local br=("$@") + local err_br_not_found='Linux bridge not found!' + local err_br_virsh_net='is a virtual network, Linux bridge expected!' + local warn_br_endpoint="Endpoints might be inaccessible from external hosts!" + # MaaS requires a Linux bridge for PXE/admin + if [[ "${vnodes}" =~ mas01 ]]; then + if ! brctl showmacs "${br[0]}" >/dev/null 2>&1; then + notify_e "[ERROR] PXE/admin (${br[0]}) ${err_br_not_found}" + fi + # Assume virsh network name matches bridge name (true if created by us) + if virsh net-info "${br[0]}" >/dev/null 2>&1; then + notify_e "[ERROR] ${br[0]} ${err_br_virsh_net}" + fi + fi + # If virtual nodes are present, public should be a Linux bridge + if [ "$(echo "${vnodes}" | wc -w)" -gt 2 ]; then + if ! brctl showmacs "${br[3]}" >/dev/null 2>&1; then + if [[ "${vnodes}" =~ mas01 ]]; then + # Baremetal nodes *require* a proper public network + notify_e "[ERROR] Public (${br[3]}) ${err_br_not_found}" + else + notify_n "[WARN] Public (${br[3]}) ${err_br_not_found}" 3 + notify_n "[WARN] ${warn_br_endpoint}" 3 + fi + fi + if virsh net-info "${br[3]}" >/dev/null 2>&1; then + if [[ "${vnodes}" =~ mas01 ]]; then + notify_e "[ERROR] ${br[3]} ${err_br_virsh_net}" + else + notify_n "[WARN] ${br[3]} ${err_br_virsh_net}" 3 + notify_n "[WARN] ${warn_br_endpoint}" 3 + fi + fi fi } function create_networks { local vnode_networks=("$@") # 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. - for net in "pxe" "mcpcontrol" "${vnode_networks[@]}"; do + for net in "mcpcontrol" "${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 @@ -240,7 +398,6 @@ function create_vms { # 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="" @@ -251,23 +408,34 @@ function create_vms { # create vms with specified options for serialized_vnode_data in "${vnodes[@]}"; do + if [ -z "${serialized_vnode_data}" ]; then continue; fi IFS=',' read -r -a vnode_data <<< "${serialized_vnode_data}" # prepare network args - net_args=" --network network=mcpcontrol,model=virtio" - if [ "${DEPLOY_TYPE:-}" = 'baremetal' ]; then + 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=${net},model=virtio" done + # dedicated storage drive for cinder on cmp nodes + virt_extra_storage= + if [[ "${vnode_data[0]}" =~ ^(cmp) ]]; then + virt_extra_storage="--disk path=${image_dir}/mcp_${vnode_data[0]}_storage.qcow2,format=qcow2,bus=virtio,cache=none,io=native" + fi + # shellcheck disable=SC2086 virt-install --name "${vnode_data[0]}" \ --ram "${vnode_data[1]}" --vcpus "${vnode_data[2]}" \ --cpu host-passthrough --accelerate ${net_args} \ --disk path="${image_dir}/mcp_${vnode_data[0]}.qcow2",format=qcow2,bus=virtio,cache=none,io=native \ + ${virt_extra_storage} \ --os-type linux --os-variant none \ --boot hd --vnc --console pty --autostart --noreboot \ --disk path="${image_dir}/mcp_${vnode_data[0]}.iso",device=cdrom \ @@ -286,6 +454,23 @@ function update_mcpcontrol_network { "" --live --config } +function reset_vms { + local vnodes=("$@") + local cmd_str="ssh ${SSH_OPTS} ${SSH_SALT}" + + # reset non-infrastructure vms, wait for them to come back online + for node in "${vnodes[@]}"; do + if [[ ! "${node}" =~ (cfg01|mas01) ]]; then + virsh reset "${node}" + fi + done + for node in "${vnodes[@]}"; do + if [[ ! "${node}" =~ (cfg01|mas01) ]]; then + wait_for 20.0 "${cmd_str} sudo salt -C '${node}*' saltutil.sync_all" + fi + done +} + function start_vms { local vnodes=("$@") @@ -344,13 +529,57 @@ 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%.*}" = "${total_attempts}" ]; then + eval "${cmdstr}" && echo "[wait_for] OK: ${cmdstr}" && return 0 || true + else + ! (eval "${cmdstr}" || echo 'No response') |& tee /dev/stderr | \ + grep -Eq '(Not connected|No response)' && \ + echo "[wait_for] OK: ${cmdstr}" && return 0 || true + fi + sleep "${sleep_time}" done + echo "[wait_for] ERROR: Failed after max attempts: ${cmdstr}" return 1 ) } + +function do_udev_cfg { + local _conf='/etc/udev/rules.d/99-opnfv-fuel-vnet-mtu.rules' + # http://linuxaleph.blogspot.com/2013/01/how-to-network-jumbo-frames-to-kvm-guest.html + echo 'SUBSYSTEM=="net", ACTION=="add", KERNEL=="vnet*", RUN+="/sbin/ip link set mtu 9000 dev '"'"%k"'"'"' |& sudo tee "${_conf}" + sudo udevadm control --reload || true +} + +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 docker_install { + # Mininum effort attempt at installing Docker if missing + if ! which docker; then + curl -fsSL https://get.docker.com -o get-docker.sh + sudo sh get-docker.sh + rm get-docker.sh + # On RHEL distros, the Docker service should be explicitly started + sudo systemctl start docker + fi +}