From ac677baaee8aab1d05c083bfa88db6c539e962a1 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Niklas=20Hamb=C3=BCchen?= Date: Sun, 11 Oct 2020 16:21:12 +0200 Subject: [PATCH] Initial import. From: * https://gist.github.com/nh2/78d1c65e33806e7728622dbe748c2b6a * https://gist.github.com/nh2/ebc27311731f53ee623ae781ca25103f * https://gist.github.com/nh2/1d2c533085b514dc1a7719f6ff35640b * https://gist.github.com/nh2/c02612e05d1a0f5dc9fd50dda04b3e48 See https://discourse.nixos.org/t/howto-install-nixos-on-an-ovh-dedicated-server/3089. --- README.md | 14 + .../nixos-install-hetzner-cloud.sh | 65 ++++ ...etzner-dedicated-wipe-and-install-nixos.sh | 288 ++++++++++++++ ...aseweb-dedicated-wipe-and-install-nixos.sh | 364 ++++++++++++++++++ .../ovh-dedicated-wipe-and-install-nixos.sh | 322 ++++++++++++++++ 5 files changed, 1053 insertions(+) create mode 100644 README.md create mode 100644 hosters/hetzner-cloud/nixos-install-hetzner-cloud.sh create mode 100644 hosters/hetzner-dedicated/hetzner-dedicated-wipe-and-install-nixos.sh create mode 100644 hosters/leaseweb-dedicated/leaseweb-dedicated-wipe-and-install-nixos.sh create mode 100644 hosters/ovh-dedicated/ovh-dedicated-wipe-and-install-nixos.sh diff --git a/README.md b/README.md new file mode 100644 index 0000000..f7eb7b9 --- /dev/null +++ b/README.md @@ -0,0 +1,14 @@ +# nixos-install-scripts + +A collection of one-shot scripts to install NixOS on various server hosters and other hardware. + +They are designed to get NixOS onto your machine with minimal effort, usually requiring only a single command and waiting a few minutes. + +See the `hosters` directory for available hosters. + + +## Usage + +Each script contains instructions at the top. + +You must slighly modify the script, most importantly, to put your login credentials (SSH key) into it. diff --git a/hosters/hetzner-cloud/nixos-install-hetzner-cloud.sh b/hosters/hetzner-cloud/nixos-install-hetzner-cloud.sh new file mode 100644 index 0000000..a4a76d5 --- /dev/null +++ b/hosters/hetzner-cloud/nixos-install-hetzner-cloud.sh @@ -0,0 +1,65 @@ +#! /usr/bin/env bash + +# Script to install NixOS from the Hetzner Cloud NixOS bootable ISO image. +# Wipes the disk! +# Tested with Hetzner's `NixOS 20.03 (amd64/minimal)` ISO image. +# +# Run like: +# +# curl https://nh2.me/nixos-install-hetzner-cloud.sh | sudo bash +# +# To run it from the Hetzner Cloud web terminal without typing it down, +# use `xdotoool` (you have e.g. 3 seconds to focus the window): +# +# sleep 3 && xdotool type --delay 50 'curl https://nh2.me/nixos-install-hetzner-cloud.sh | sudo bash' +# +# (In the xdotool invocation you may have to replace chars so that +# the right chars appear on the US-English keyboard.) +# +# If you want to be able to SSH straight in, +# do not forget to replace the SSH key below by yours +# (in the section labelled "Replace this by your SSH pubkey"), +# and host script modified this way under and URL of your choosing. +# Otherwise you'l be running with my pubkey, but you can change it +# afterwards by logging in via the Hetzner Cloud web terminal as `root` +# with empty password. + +set -e + +# Hetzner Cloud OS images grow the root partition to the size of the local +# disk on first book. In case the NixOS live ISO is booted immediately on +# first powerup, that does not happen. Thus we need to grow the partition +# by deleting and re-creating it. +sgdisk -d 1 /dev/sda +sgdisk -N 1 /dev/sda +partprobe /dev/sda + +mkfs.ext4 -F /dev/sda1 # wipes all data! + +mount /dev/sda1 /mnt + +nixos-generate-config --root /mnt + +# Delete trailing `}` from `configuration.nix` so that we can append more to it. +sed -i -E 's:^\}\s*$::g' /mnt/etc/nixos/configuration.nix + +# Extend/override default `configuration.nix`: +echo ' + boot.loader.grub.devices = [ "/dev/sda" ]; + + # Initial empty root password for easy login: + users.users.root.initialHashedPassword = ""; + services.openssh.permitRootLogin = "prohibit-password"; + + services.openssh.enable = true; + + # Replace this by your SSH pubkey + users.users.root.openssh.authorizedKeys.keys = [ + "ssh-rsa AAAAB3NzaC1yc2EAAAABIwAAAQEAtwCIGPYJlD2eeUtxngmT+4yR7BMlK0F5kzj+84uHsxxsy+PXFrP/tScCpwmuoiEYNv/9WKnPJJfCA9XlIDr6cla1MLpaW6eg672TRYMmKzH6SLlkg+kyDmPxSIJw+KdKfnPYyva+Y/VocACYJo0voabUeLAVgtSKGz/AFzccjfOR0GmFO911zjAaR+jFb9M7t7dveNVKm9KbuBfu3giMgGg3/mKz1TKY8yk2ZOxpT5CllBb+B5BcEf+7IGNvNxr1Z0zz5cFXQ3LyBIZklnC/OaQCnD78BSiyPTkIXcmBFal2TaFwTDvki6PuCRpJy+dU1fDdgWLql97D0SVnjmmomw==" + ]; +} +' >> /mnt/etc/nixos/configuration.nix + +nixos-install --no-root-passwd + +reboot diff --git a/hosters/hetzner-dedicated/hetzner-dedicated-wipe-and-install-nixos.sh b/hosters/hetzner-dedicated/hetzner-dedicated-wipe-and-install-nixos.sh new file mode 100644 index 0000000..4ac953f --- /dev/null +++ b/hosters/hetzner-dedicated/hetzner-dedicated-wipe-and-install-nixos.sh @@ -0,0 +1,288 @@ +#!/usr/bin/env bash + +# Installs NixOS on a Hetzner server, wiping the server. +# +# This is for a specific server configuration; adjust where needed. +# +# Prerequisites: +# * Update the script to adjust SSH pubkeys, hostname, NixOS version etc. +# +# Usage: +# ssh root@YOUR_SERVERS_IP bash -s < hetzner-dedicated-wipe-and-install-nixos.sh +# +# When the script is done, make sure to boot the server from HD, not rescue mode again. + +# Explanations: +# +# * Adapted from https://gist.github.com/nh2/78d1c65e33806e7728622dbe748c2b6a +# * Following largely https://nixos.org/nixos/manual/index.html#sec-installing-from-other-distro. +# * **Important:** We boot in legacy-BIOS mode, not UEFI, because that's what Hetzner uses. +# * NVMe devices aren't supported for booting (those require EFI boot) +# * We set a custom `configuration.nix` so that we can connect to the machine afterwards, +# inspired by https://nixos.wiki/wiki/Install_NixOS_on_Hetzner_Online +# * This server has 2 HDDs. +# We put everything on RAID1. +# Storage scheme: `partitions -> RAID -> LVM -> ext4`. +# * A root user with empty password is created, so that you can just login +# as root and press enter when using the Hetzner spider KVM. +# Of course that empty-password login isn't exposed to the Internet. +# Change the password afterwards to avoid anyone with physical access +# being able to login without any authentication. +# * The script reboots at the end. + +set -eu +set -o pipefail + +set -x + +# Inspect existing disks +lsblk + +# Undo existing setups to allow running the script multiple times to iterate on it. +# We allow these operations to fail for the case the script runs the first time. +set +e +umount /mnt +vgchange -an +set -e + +# Stop all mdadm arrays that the boot may have activated. +mdadm --stop --scan + +# Prevent mdadm from auto-assembling arrays. +# Otherwise, as soon as we create the partition tables below, it will try to +# re-assemple a previous RAID if any remaining RAID signatures are present, +# before we even get the chance to wipe them. +# From: +# https://unix.stackexchange.com/questions/166688/prevent-debian-from-auto-assembling-raid-at-boot/504035#504035 +# We use `>` because the file may already contain some detected RAID arrays, +# which would take precedence over our ``. +echo 'AUTO -all +ARRAY UUID=00000000:00000000:00000000:00000000' > /etc/mdadm/mdadm.conf + +# Create partition tables (--script to not ask) +parted --script /dev/sda mklabel gpt +parted --script /dev/sdb mklabel gpt + +# Create partitions (--script to not ask) +# +# We create the 1MB BIOS boot partition at the front. +# +# Note we use "MB" instead of "MiB" because otherwise `--align optimal` has no effect; +# as per documentation https://www.gnu.org/software/parted/manual/html_node/unit.html#unit: +# > Note that as of parted-2.4, when you specify start and/or end values using IEC +# > binary units like "MiB", "GiB", "TiB", etc., parted treats those values as exact +# +# Note: When using `mkpart` on GPT, as per +# https://www.gnu.org/software/parted/manual/html_node/mkpart.html#mkpart +# the first argument to `mkpart` is not a `part-type`, but the GPT partition name: +# ... part-type is one of 'primary', 'extended' or 'logical', and may be specified only with 'msdos' or 'dvh' partition tables. +# A name must be specified for a 'gpt' partition table. +# GPT partition names are limited to 36 UTF-16 chars, see https://en.wikipedia.org/wiki/GUID_Partition_Table#Partition_entries_(LBA_2-33). +parted --script --align optimal /dev/sda -- mklabel gpt mkpart 'BIOS-boot-partition' 1MB 2MB set 1 bios_grub on mkpart 'data-partition' 2MB '100%' +parted --script --align optimal /dev/sdb -- mklabel gpt mkpart 'BIOS-boot-partition' 1MB 2MB set 1 bios_grub on mkpart 'data-partition' 2MB '100%' + +# Relaod partitions +partprobe + +# Wait for all devices to exist +udevadm settle --timeout=5 --exit-if-exists=/dev/sda1 +udevadm settle --timeout=5 --exit-if-exists=/dev/sda2 +udevadm settle --timeout=5 --exit-if-exists=/dev/sdb1 +udevadm settle --timeout=5 --exit-if-exists=/dev/sdb2 + +# Wipe any previous RAID signatures +mdadm --zero-superblock --force /dev/sda2 +mdadm --zero-superblock --force /dev/sdb2 + +# Create RAIDs +# Note that during creating and boot-time assembly, mdadm cares about the +# host name, and the existence and contents of `mdadm.conf`! +# This also affects the names appearing in /dev/md/ being different +# before and after reboot in general (but we take extra care here +# to pass explicit names, and set HOMEHOST for the rebooting system further +# down, so that the names appear the same). +# Almost all details of this are explained in +# https://bugzilla.redhat.com/show_bug.cgi?id=606481#c14 +# and the followup comments by Doug Ledford. +mdadm --create --run --verbose /dev/md0 --level=1 --raid-devices=2 --homehost=hetzner --name=root0 /dev/sda2 /dev/sdb2 + +# Assembling the RAID can result in auto-activation of previously-existing LVM +# groups, preventing the RAID block device wiping below with +# `Device or resource busy`. So disable all VGs first. +vgchange -an + +# Wipe filesystem signatures that might be on the RAID from some +# possibly existing older use of the disks (RAID creation does not do that). +# See https://serverfault.com/questions/911370/why-does-mdadm-zero-superblock-preserve-file-system-information +wipefs -a /dev/md0 + +# Disable RAID recovery. We don't want this to slow down machine provisioning +# in the rescue mode. It can run in normal operation after reboot. +echo 0 > /proc/sys/dev/raid/speed_limit_max + +# LVM +# PVs +pvcreate /dev/md0 +# VGs +vgcreate vg0 /dev/md0 +# LVs (--yes to automatically wipe detected file system signatures) +lvcreate --yes --extents 95%FREE -n root0 vg0 # 5% slack space + +# Filesystems (-F to not ask on preexisting FS) +mkfs.ext4 -F -L root /dev/mapper/vg0-root0 + +# Creating file systems changes their UUIDs. +# Trigger udev so that the entries in /dev/disk/by-uuid get refreshed. +# `nixos-generate-config` depends on those being up-to-date. +# See https://github.com/NixOS/nixpkgs/issues/62444 +udevadm trigger + +# Wait for FS labels to appear +udevadm settle --timeout=5 --exit-if-exists=/dev/disk/by-label/root + +# NixOS pre-installation mounts + +# Mount target root partition +mount /dev/disk/by-label/root /mnt + +# Installing nix + +# Installing nix requires `sudo`; the Hetzner rescue mode doesn't have it. +apt-get install -y sudo + +# Allow installing nix as root, see +# https://github.com/NixOS/nix/issues/936#issuecomment-475795730 +mkdir -p /etc/nix +echo "build-users-group =" > /etc/nix/nix.conf + +curl -L https://nixos.org/nix/install | sh +set +u +x # sourcing this may refer to unset variables that we have no control over +. $HOME/.nix-profile/etc/profile.d/nix.sh +set -u -x + +# Keep in sync with `system.stateVersion` set below! +# nix-channel --add https://nixos.org/channels/nixos-20.03 nixpkgs +nix-channel --add https://nixos.org/channels/nixos-20.03 nixpkgs +nix-channel --update + +# Getting NixOS installation tools +nix-env -iE "_: with import { configuration = {}; }; with config.system.build; [ nixos-generate-config nixos-install nixos-enter manual.manpages ]" + +nixos-generate-config --root /mnt + +# Find the name of the network interface that connects us to the Internet. +# Inspired by https://unix.stackexchange.com/questions/14961/how-to-find-out-which-interface-am-i-using-for-connecting-to-the-internet/302613#302613 +RESCUE_INTERFACE=$(ip route get 8.8.8.8 | grep -Po '(?<=dev )(\S+)') + +# Find what its name will be under NixOS, which uses stable interface names. +# See https://major.io/2015/08/21/understanding-systemds-predictable-network-device-names/#comment-545626 +# NICs for most Hetzner servers are not onboard, which is why we use +# `ID_NET_NAME_PATH`otherwise it would be `ID_NET_NAME_ONBOARD`. +INTERFACE_DEVICE_PATH=$(udevadm info -e | grep -Po "(?<=^P: )(.*${RESCUE_INTERFACE})") +UDEVADM_PROPERTIES_FOR_INTERFACE=$(udevadm info --query=property "--path=$INTERFACE_DEVICE_PATH") +NIXOS_INTERFACE=$(echo "$UDEVADM_PROPERTIES_FOR_INTERFACE" | grep -o -E 'ID_NET_NAME_PATH=\w+' | cut -d= -f2) +echo "Determined NIXOS_INTERFACE as '$NIXOS_INTERFACE'" + +IP_V4=$(ip route get 8.8.8.8 | grep -Po '(?<=src )(\S+)') +echo "Determined IP_V4 as $IP_V4" + +# Determine Internet IPv6 by checking route, and using ::1 +# (because Hetzner rescue mode uses ::2 by default). +# The `ip -6 route get` output on Hetzner looks like: +# # ip -6 route get 2001:4860:4860:0:0:0:0:8888 +# 2001:4860:4860::8888 via fe80::1 dev eth0 src 2a01:4f8:151:62aa::2 metric 1024 pref medium +IP_V6="$(ip route get 2001:4860:4860:0:0:0:0:8888 | head -1 | cut -d' ' -f7 | cut -d: -f1-4)::1" +echo "Determined IP_V6 as $IP_V6" + + +# From https://stackoverflow.com/questions/1204629/how-do-i-get-the-default-gateway-in-linux-given-the-destination/15973156#15973156 +read _ _ DEFAULT_GATEWAY _ < <(ip route list match 0/0); echo "$DEFAULT_GATEWAY" +echo "Determined DEFAULT_GATEWAY as $DEFAULT_GATEWAY" + + +# Generate `configuration.nix`. Note that we splice in shell variables. +cat > /mnt/etc/nixos/configuration.nix <' (using the system hostname). + # This results mdadm considering such disks as "foreign" as opposed to + # "local", and showing them as e.g. '/dev/md/hetzner:root0' + # instead of '/dev/md/root0'. + # This is mdadm's protection against accidentally putting a RAID disk + # into the wrong machine and corrupting data by accidental sync, see + # https://bugzilla.redhat.com/show_bug.cgi?id=606481#c14 and onward. + # We set the HOMEHOST manually go get the short '/dev/md' names, + # and so that things look and are configured the same on all such + # machines irrespective of host names. + # We do not worry about plugging disks into the wrong machine because + # we will never exchange disks between machines. + environment.etc."mdadm.conf".text = '' + HOMEHOST hetzner + ''; + # The RAIDs are assembled in stage1, so we need to make the config + # available there. + boot.initrd.mdadmConf = config.environment.etc."mdadm.conf".text; + + # Network (Hetzner uses static IP assignments, and we don't use DHCP here) + networking.useDHCP = false; + networking.interfaces."$NIXOS_INTERFACE".ipv4.addresses = [ + { + address = "$IP_V4"; + prefixLength = 24; + } + ]; + networking.interfaces."$NIXOS_INTERFACE".ipv6.addresses = [ + { + address = "$IP_V6"; + prefixLength = 64; + } + ]; + networking.defaultGateway = "$DEFAULT_GATEWAY"; + networking.defaultGateway6 = { address = "fe80::1"; interface = "$NIXOS_INTERFACE"; }; + networking.nameservers = [ "8.8.8.8" ]; + + # Initial empty root password for easy login: + users.users.root.initialHashedPassword = ""; + services.openssh.permitRootLogin = "prohibit-password"; + + users.users.root.openssh.authorizedKeys.keys = [ + "ssh-rsa AAAAB3NzaC1yc2EAAAABIwAAAQEAtwCIGPYJlD2eeUtxngmT+4yR7BMlK0F5kzj+84uHsxxsy+PXFrP/tScCpwmuoiEYNv/9WKnPJJfCA9XlIDr6cla1MLpaW6eg672TRYMmKzH6SLlkg+kyDmPxSIJw+KdKfnPYyva+Y/VocACYJo0voabUeLAVgtSKGz/AFzccjfOR0GmFO911zjAaR+jFb9M7t7dveNVKm9KbuBfu3giMgGg3/mKz1TKY8yk2ZOxpT5CllBb+B5BcEf+7IGNvNxr1Z0zz5cFXQ3LyBIZklnC/OaQCnD78BSiyPTkIXcmBFal2TaFwTDvki6PuCRpJy+dU1fDdgWLql97D0SVnjmmomw== nh2@deditus.de" + ]; + + services.openssh.enable = true; + + # This value determines the NixOS release with which your system is to be + # compatible, in order to avoid breaking some software such as database + # servers. You should change this only after NixOS release notes say you + # should. + system.stateVersion = "20.03"; # Did you read the comment? + +} +EOF + +# Install NixOS +PATH="$PATH" NIX_PATH="$NIX_PATH" `which nixos-install` --no-root-passwd --root /mnt --max-jobs 40 + +umount /mnt + +reboot diff --git a/hosters/leaseweb-dedicated/leaseweb-dedicated-wipe-and-install-nixos.sh b/hosters/leaseweb-dedicated/leaseweb-dedicated-wipe-and-install-nixos.sh new file mode 100644 index 0000000..abba023 --- /dev/null +++ b/hosters/leaseweb-dedicated/leaseweb-dedicated-wipe-and-install-nixos.sh @@ -0,0 +1,364 @@ +#!/usr/bin/env bash + +# Installs NixOS on a Leaseweb server, wiping the server. +# +# This is for a specific server configuration; adjust where needed. +# Originally written for a Leaseweb HP DL120 G7 server. +# +# Prerequisites: +# * Update the script to adjust SSH pubkeys, hostname, NixOS version etc. +# +# Usage: +# ssh root@YOUR_SERVERS_IP bash -s < leaseweb-dedicated-wipe-and-install-nixos.sh +# +# When the script is done, make sure to boot the server from HD, not rescue mode again. + +# Explanations: +# +# * Following largely https://nixos.org/nixos/manual/index.html#sec-installing-from-other-distro. +# * Adapted from https://gist.github.com/nh2/78d1c65e33806e7728622dbe748c2b6a +# * Following largely https://nixos.org/nixos/manual/index.html#sec-installing-from-other-distro. +# * **Important:** We boot in legacy-BIOS mode, not UEFI, because that's what the HP DL120 G7 supports, +# see https://lists.freebsd.org/pipermail/freebsd-proliant/2014-June/000666.html. +# * NVMe devices aren't supported for booting (those require EFI boot) +# * We set a custom `configuration.nix` so that we can connect to the machine afterwards. +# * This server has 2 HDDs. +# We put everything on RAID1. +# Storage scheme: `partitions -> RAID -> LVM -> ext4`. +# * A root user with empty password is created, so that you can just login +# as root and press enter when using a KVM. +# Of course that empty-password login isn't exposed to the Internet. +# Change the password afterwards to avoid anyone with physical access +# being able to login without any authentication. +# * The script reboots at the end. + +set -eu +set -o pipefail + +set -x + +# Inspect existing disks +lsblk + +# Undo existing setups to allow running the script multiple times to iterate on it. +# We allow these operations to fail for the case the script runs the first time. +set +e +umount /mnt +vgchange -an +set -e + +# Stop all mdadm arrays that the boot may have activated. +mdadm --stop --scan + +# Prevent mdadm from auto-assembling arrays. +# Otherwise, as soon as we create the partition tables below, it will try to +# re-assemple a previous RAID if any remaining RAID signatures are present, +# before we even get the chance to wipe them. +# From: +# https://unix.stackexchange.com/questions/166688/prevent-debian-from-auto-assembling-raid-at-boot/504035#504035 +# We use `>` because the file may already contain some detected RAID arrays, +# which would take precedence over our ``. +echo 'AUTO -all +ARRAY UUID=00000000:00000000:00000000:00000000' > /etc/mdadm/mdadm.conf + +# Create partition tables (--script to not ask) +parted --script /dev/sda mklabel gpt +parted --script /dev/sdb mklabel gpt + +# Create partitions (--script to not ask) +# +# We create the 1MB BIOS boot partition at the front. +# +# Note we use "MB" instead of "MiB" because otherwise `--align optimal` has no effect; +# as per documentation https://www.gnu.org/software/parted/manual/html_node/unit.html#unit: +# > Note that as of parted-2.4, when you specify start and/or end values using IEC +# > binary units like "MiB", "GiB", "TiB", etc., parted treats those values as exact +# +# Note: When using `mkpart` on GPT, as per +# https://www.gnu.org/software/parted/manual/html_node/mkpart.html#mkpart +# the first argument to `mkpart` is not a `part-type`, but the GPT partition name: +# ... part-type is one of 'primary', 'extended' or 'logical', and may be specified only with 'msdos' or 'dvh' partition tables. +# A name must be specified for a 'gpt' partition table. +# GPT partition names are limited to 36 UTF-16 chars, see https://en.wikipedia.org/wiki/GUID_Partition_Table#Partition_entries_(LBA_2-33). +parted --script --align optimal /dev/sda -- mklabel gpt mkpart 'BIOS-boot-partition' 1MB 2MB set 1 bios_grub on mkpart 'data-partition' 2MB '100%' +parted --script --align optimal /dev/sdb -- mklabel gpt mkpart 'BIOS-boot-partition' 1MB 2MB set 1 bios_grub on mkpart 'data-partition' 2MB '100%' + +# Relaod partitions +partprobe + +# Wait for all devices to exist +udevadm settle --timeout=5 --exit-if-exists=/dev/sda1 +udevadm settle --timeout=5 --exit-if-exists=/dev/sda2 +udevadm settle --timeout=5 --exit-if-exists=/dev/sdb1 +udevadm settle --timeout=5 --exit-if-exists=/dev/sdb2 + +# Wipe any previous RAID signatures +mdadm --zero-superblock --force /dev/sda2 +mdadm --zero-superblock --force /dev/sdb2 + +# Create RAIDs +# Note that during creating and boot-time assembly, mdadm cares about the +# host name, and the existence and contents of `mdadm.conf`! +# This also affects the names appearing in /dev/md/ being different +# before and after reboot in general (but we take extra care here +# to pass explicit names, and set HOMEHOST for the rebooting system further +# down, so that the names appear the same). +# Almost all details of this are explained in +# https://bugzilla.redhat.com/show_bug.cgi?id=606481#c14 +# and the followup comments by Doug Ledford. +mdadm --create --run --verbose /dev/md0 --level=1 --raid-devices=2 --homehost=leaseweb --name=root0 /dev/sda2 /dev/sdb2 + +# Assembling the RAID can result in auto-activation of previously-existing LVM +# groups, preventing the RAID block device wiping below with +# `Device or resource busy`. So disable all VGs first. +vgchange -an + +# Wipe filesystem signatures that might be on the RAID from some +# possibly existing older use of the disks (RAID creation does not do that). +# See https://serverfault.com/questions/911370/why-does-mdadm-zero-superblock-preserve-file-system-information +wipefs -a /dev/md0 + +# Disable RAID recovery. We don't want this to slow down machine provisioning +# in the rescue mode. It can run in normal operation after reboot. +echo 0 > /proc/sys/dev/raid/speed_limit_max + +# LVM +# PVs +pvcreate /dev/md0 +# VGs +vgcreate vg0 /dev/md0 +# LVs (--yes to automatically wipe detected file system signatures) +lvcreate --yes --extents 95%FREE -n root0 vg0 # 5% slack space + +# Filesystems (-F to not ask on preexisting FS) +mkfs.ext4 -F -L root /dev/mapper/vg0-root0 + +# Creating file systems changes their UUIDs. +# Trigger udev so that the entries in /dev/disk/by-uuid get refreshed. +# `nixos-generate-config` depends on those being up-to-date. +# See https://github.com/NixOS/nixpkgs/issues/62444 +udevadm trigger + +# Wait for FS labels to appear +udevadm settle --timeout=5 --exit-if-exists=/dev/disk/by-label/root + +# NixOS pre-installation mounts + +# Mount target root partition +mount /dev/disk/by-label/root /mnt + +# Installing nix + +# Allow installing nix as root, see +# https://github.com/NixOS/nix/issues/936#issuecomment-475795730 +mkdir -p /etc/nix +echo "build-users-group =" > /etc/nix/nix.conf + +curl -L https://nixos.org/nix/install | sh +set +u +x # sourcing this may refer to unset variables that we have no control over +. $HOME/.nix-profile/etc/profile.d/nix.sh +set -u -x + +# Keep in sync with `system.stateVersion` set below! +# nix-channel --add https://nixos.org/channels/nixos-20.03 nixpkgs +nix-channel --add https://nixos.org/channels/nixos-20.03 nixpkgs +nix-channel --update + +# Getting NixOS installation tools +nix-env -iE "_: with import { configuration = {}; }; with config.system.build; [ nixos-generate-config nixos-install nixos-enter manual.manpages ]" + +nixos-generate-config --root /mnt + +# Find the name of the network interface that connects us to the Internet. +# Inspired by https://unix.stackexchange.com/questions/14961/how-to-find-out-which-interface-am-i-using-for-connecting-to-the-internet/302613#302613 +RESCUE_INTERFACE=$(ip route get 8.8.8.8 | grep -Po '(?<=dev )(\S+)') + +# Find what its name will be under NixOS, which uses stable interface names. +# See https://major.io/2015/08/21/understanding-systemds-predictable-network-device-names/#comment-545626 +# +# IMPORTANT: +# There is a known complication in that Linux somewhere between 4.19 and 5.4.27 +# switched from classifying only 1 of the 2 network interfaces of the server as +# "onboard" to classifying both as "onboard", thus "enp2s0" shows up as "eno0" +# instead in newer kernels. +# See: +# https://gist.github.com/nh2/71854c40a1a1a7c15bc8a8105e854f88#file-analysis-md +# So once the Leaseweb GRML rescue mode upgrades to a newer kernel, the value of +# `NIXOS_INTERFACE` should be successfully found from `RESCUE_INTERFACE` using +# the `ID_NET_NAME_ONBOARD` grep below; but until then (when the grep is empty) +# we have to detect this situation, turning `enp2s0` into `eno0` ourselves, +# because we want to boot a NixOS that uses the new kernel (>= 5.4.27) of which +# we know that it will detect the card as "onboard" and thus call it "eno". +INTERFACE_DEVICE_PATH=$(udevadm info -e | grep -Po "(?<=^P: )(.*${RESCUE_INTERFACE})") +UDEVADM_PROPERTIES_FOR_INTERFACE=$(udevadm info --query=property "--path=$INTERFACE_DEVICE_PATH") +set +o pipefail # allow the grep to fail, see comment above +NIXOS_INTERFACE=$(echo "$UDEVADM_PROPERTIES_FOR_INTERFACE" | grep -o -E 'ID_NET_NAME_ONBOARD=\w+' | cut -d= -f2) +set -o pipefail +# The following `if` logic can be deleted once versions < 20.03 are no longer relevant. +if [ -z "$NIXOS_INTERFACE" ]; then + echo "Could not determine NIXOS_INTERFACE from udevadm, RESCUE_INTERFACE is '$RESCUE_INTERFACE'" + # Set this to 1 iff you are installing a newer kernel as described in the comment above: + INSTALLING_NEWER_KERNEL=1 + if [ "$INSTALLING_NEWER_KERNEL" == "1" ]; then + echo "INSTALLING_NEWER_KERNEL=1 is active, setting NIXOS_INTERFACE=eno0" + NIXOS_INTERFACE="eno0" + else + echo "INSTALLING_NEWER_KERNEL=1 is NOT active, setting NIXOS_INTERFACE=$RESCUE_INTERFACE" + NIXOS_INTERFACE="$RESCUE_INTERFACE" + fi +else + echo "Determined NIXOS_INTERFACE as '$NIXOS_INTERFACE'" +fi + +IP_V4=$(ip route get 8.8.8.8 | grep -Po '(?<=src )(\S+)') +echo "Determined IP_V4 as $IP_V4" + +# From https://stackoverflow.com/questions/1204629/how-do-i-get-the-default-gateway-in-linux-given-the-destination/15973156#15973156 +read _ _ DEFAULT_GATEWAY _ < <(ip route list match 0/0); echo "$DEFAULT_GATEWAY" +echo "Determined DEFAULT_GATEWAY as $DEFAULT_GATEWAY" + +# The Leaseweb GRML Rescue mode as of writing has no IPv6 connectivity, +# so we cannot get the IPv6 address here. + + +# Generate `configuration.nix`. Note that we splice in shell variables. +cat > /mnt/etc/nixos/configuration.nix <' (using the system hostname). + # This results mdadm considering such disks as "foreign" as opposed to + # "local", and showing them as e.g. '/dev/md/leaseweb:root0' + # instead of '/dev/md/root0'. + # This is mdadm's protection against accidentally putting a RAID disk + # into the wrong machine and corrupting data by accidental sync, see + # https://bugzilla.redhat.com/show_bug.cgi?id=606481#c14 and onward. + # We set the HOMEHOST manually go get the short '/dev/md' names, + # and so that things look and are configured the same on all such + # machines irrespective of host names. + # We do not worry about plugging disks into the wrong machine because + # we will never exchange disks between machines. + environment.etc."mdadm.conf".text = '' + HOMEHOST leaseweb + ''; + # The RAIDs are assembled in stage1, so we need to make the config + # available there. + boot.initrd.mdadmConf = config.environment.etc."mdadm.conf".text; + + # Network + # Leaseweb uses static IP assignments only, see: + # https://kb.leaseweb.com/network/ipv4-address-assignment-and-usage-guidelines#IPv4addressassignmentandusageguidelines-DHCP + networking.useDHCP = false; + networking.interfaces."$NIXOS_INTERFACE".ipv4.addresses = [ + { + address = "$IP_V4"; + prefixLength = 24; + } + ]; + networking.defaultGateway = "$DEFAULT_GATEWAY"; + networking.nameservers = [ "8.8.8.8" ]; + + # Initial empty root password for easy login: + users.users.root.initialHashedPassword = ""; + services.openssh.permitRootLogin = "prohibit-password"; + + users.users.root.openssh.authorizedKeys.keys = [ + "ssh-rsa AAAAB3NzaC1yc2EAAAABIwAAAQEAtwCIGPYJlD2eeUtxngmT+4yR7BMlK0F5kzj+84uHsxxsy+PXFrP/tScCpwmuoiEYNv/9WKnPJJfCA9XlIDr6cla1MLpaW6eg672TRYMmKzH6SLlkg+kyDmPxSIJw+KdKfnPYyva+Y/VocACYJo0voabUeLAVgtSKGz/AFzccjfOR0GmFO911zjAaR+jFb9M7t7dveNVKm9KbuBfu3giMgGg3/mKz1TKY8yk2ZOxpT5CllBb+B5BcEf+7IGNvNxr1Z0zz5cFXQ3LyBIZklnC/OaQCnD78BSiyPTkIXcmBFal2TaFwTDvki6PuCRpJy+dU1fDdgWLql97D0SVnjmmomw== nh2@deditus.de" + ]; + + services.openssh.enable = true; + + # This value determines the NixOS release with which your system is to be + # compatible, in order to avoid breaking some software such as database + # servers. You should change this only after NixOS release notes say you + # should. + system.stateVersion = "20.03"; # Did you read the comment? + +} +EOF + +# TODO Remove once https://github.com/NixOS/nixpkgs/pull/85895 is merged and +# backported to 20.03, or this script installs a newer version that has it. +rm -f extra-grub-install-flags-20.03.tar.gz +wget 'https://github.com/nh2/nixpkgs/archive/extra-grub-install-flags-20.03.tar.gz' +rm -rf nixpkgs-extra-grub-install-flags-20.03 +tar xf extra-grub-install-flags-20.03.tar.gz +NIX_PATH=nixpkgs=$PWD/nixpkgs-extra-grub-install-flags-20.03 + + +# Install NixOS +PATH="$PATH" NIX_PATH="$NIX_PATH" `which nixos-install` --no-root-passwd --root /mnt --max-jobs 40 + +umount /mnt + +reboot diff --git a/hosters/ovh-dedicated/ovh-dedicated-wipe-and-install-nixos.sh b/hosters/ovh-dedicated/ovh-dedicated-wipe-and-install-nixos.sh new file mode 100644 index 0000000..496bd3c --- /dev/null +++ b/hosters/ovh-dedicated/ovh-dedicated-wipe-and-install-nixos.sh @@ -0,0 +1,322 @@ +#!/usr/bin/env bash + +# Installs NixOS on an OVH server, wiping the server. +# +# This is for a specific server configuration; adjust where needed. +# Originally written for an OVH STOR-1 server. +# +# Prerequisites: +# * Create a LUKS key file at /root/benacofs-luks-key +# e.g. by copying it up. +# * Update the script to adjust SSH pubkeys, hostname NixOS version etc. +# +# Usage: +# ssh root@YOUR_SERVERS_IP bash -s < ovh-dedicated-wipe-and-install-nixos.sh +# +# When the script is done, make sure to boot the server from HD, not rescue mode again. + +# Explanations: +# +# * Following largely https://nixos.org/nixos/manual/index.html#sec-installing-from-other-distro. +# * **Important:** We boot in UEFI mode, thus requiring an ESP. +# Booting in LEGACY mode (non-UEFI boot, without ESP) would require that: +# * `/boot` is on the same device as GRUB +# * NVMe devices aren not used for booting (those require EFI boot) +# We also did not manage to boot our OVH server in LEGACY mode on our SuperMicro mainboard, even when we installed `/` (including `/boot`) directly to a simple RAID1ed GPT partition. The screen just stayed black. +# * We set a custom `configuration.nix` so that we can connect to the machine afterwards. +# * This server has 1 SSD and 4 HDDs. +# We'll ignore the SSD, putting the OS on the HDDs as well, so that everything is on RAID1. +# We wipe the SSD though, so that if it had some boot partitions on it, they don't interfere. +# Storage scheme: `partitions -> RAID -> LUKS -> LVM -> ext4`. +# * A root user with empty password is created, so that you can just login +# as root and press enter when using the OVH KVM. +# Of course that empty-password login isn't exposed to the Internet. +# Change the password afterwards to avoid anyone with physical access +# being able to login without any authentication. +# * The script reboots at the end. + + +set -eu +set -o pipefail + +set -x + +# Inspect existing disks +lsblk + +# Undo existing setups to allow running the script multiple times to iterate on it. +# We allow these operations to fail for the case the script runs the first time. +set +e +umount /mnt/boot/ESP* +umount /mnt +vgchange -an +cryptsetup luksClose data0-unencrypted +cryptsetup luksClose data1-unencrypted +set -e + +# Stop all mdadm arrays that the boot may have activated. +mdadm --stop --scan + +# Create partition tables (--script to not ask) +parted --script /dev/sda mklabel gpt +parted --script /dev/sdb mklabel gpt +parted --script /dev/sdc mklabel gpt +parted --script /dev/sdd mklabel gpt +parted --script /dev/nvme0n1 mklabel gpt + +# Create partitions (--script to not ask) +# +# Create EFI system partition (ESP) and main partition for each boot device. +# We make it 550 M as recommended by the author of gdisk (https://www.rodsbooks.com/linux-uefi/); +# using 550 ensures it's greater than 512 MiB, no matter if Mi or M were used. +# For the non-boot devices, we still make space for an ESP partition +# (in case the disks get repurposed for that at some point) but mark +# it as `off` and label it `*-unused` to avoid confusion. +# +# Note we use "MB" instead of "MiB" because otherwise `--align optimal` has no effect; +# as per documentation https://www.gnu.org/software/parted/manual/html_node/unit.html#unit: +# > Note that as of parted-2.4, when you specify start and/or end values using IEC +# > binary units like "MiB", "GiB", "TiB", etc., parted treats those values as exact +# +# Note: When using `mkpart` on GPT, as per +# https://www.gnu.org/software/parted/manual/html_node/mkpart.html#mkpart +# the first argument to `mkpart` is not a `part-type`, but the GPT partition name: +# ... part-type is one of 'primary', 'extended' or 'logical', and may be specified only with 'msdos' or 'dvh' partition tables. +# A name must be specified for a 'gpt' partition table. +# GPT partition names are limited to 36 UTF-16 chars, see https://en.wikipedia.org/wiki/GUID_Partition_Table#Partition_entries_(LBA_2-33). +parted --script --align optimal /dev/sda -- mklabel gpt mkpart 'ESP-partition0' fat32 1MB 551MB set 1 esp on mkpart 'OS-partition0' 551MB 500GB mkpart 'data-partition0' 500GB '100%' +parted --script --align optimal /dev/sdb -- mklabel gpt mkpart 'ESP-partition1' fat32 1MB 551MB set 1 esp on mkpart 'OS-partition1' 551MB 500GB mkpart 'data-partition1' 500GB '100%' +parted --script --align optimal /dev/sdc -- mklabel gpt mkpart 'ESP-partition2-unused' fat32 1MB 551MB set 1 esp off mkpart 'data-partition2' 551MB '100%' +parted --script --align optimal /dev/sdd -- mklabel gpt mkpart 'ESP-partition3-unused' fat32 1MB 551MB set 1 esp off mkpart 'data-partition3' 551MB '100%' + +# Relaod partitions +partprobe + +# Wait for all devices to exist +udevadm settle --timeout=5 --exit-if-exists=/dev/sda1 +udevadm settle --timeout=5 --exit-if-exists=/dev/sda2 +udevadm settle --timeout=5 --exit-if-exists=/dev/sda3 +udevadm settle --timeout=5 --exit-if-exists=/dev/sdb1 +udevadm settle --timeout=5 --exit-if-exists=/dev/sdb2 +udevadm settle --timeout=5 --exit-if-exists=/dev/sdb3 +udevadm settle --timeout=5 --exit-if-exists=/dev/sdc1 +udevadm settle --timeout=5 --exit-if-exists=/dev/sdc2 +udevadm settle --timeout=5 --exit-if-exists=/dev/sdd1 +udevadm settle --timeout=5 --exit-if-exists=/dev/sdd2 + +# Wipe any previous RAID signatures +mdadm --zero-superblock /dev/sda2 +mdadm --zero-superblock /dev/sda3 +mdadm --zero-superblock /dev/sdb2 +mdadm --zero-superblock /dev/sdb3 +mdadm --zero-superblock /dev/sdc2 +mdadm --zero-superblock /dev/sdd2 + +# Create RAIDs +# Note that during creating and boot-time assembly, mdadm cares about the +# host name, and the existence and contents of `mdadm.conf`! +# This also affects the names appearing in /dev/md/ being different +# before and after reboot in general (but we take extra care here +# to pass explicit names, and set HOMEHOST for the rebooting system further +# down, so that the names appear the same). +# Almost all details of this are explained in +# https://bugzilla.redhat.com/show_bug.cgi?id=606481#c14 +# and the followup comments by Doug Ledford. +mdadm --create --run --verbose /dev/md/root0 --level=1 --raid-devices=2 --homehost=benaco-cdn --name=root0 /dev/sda2 /dev/sdb2 +mdadm --create --run --verbose /dev/md/data0-encrypted --level=1 --raid-devices=2 --homehost=benaco-cdn --name=data0-encrypted /dev/sda3 /dev/sdb3 +mdadm --create --run --verbose /dev/md/data1-encrypted --level=1 --raid-devices=2 --homehost=benaco-cdn --name=data1-encrypted /dev/sdc2 /dev/sdd2 + +# Assembling the RAID can result in auto-activation of previously-existing LVM +# groups, preventing the RAID block device wiping below with +# `Device or resource busy`. So disable all VGs first. +vgchange -an + +# Wipe filesystem signatures that might be on the RAID from some +# possibly existing older use of the disks (RAID creation does not do that). +# See https://serverfault.com/questions/911370/why-does-mdadm-zero-superblock-preserve-file-system-information +wipefs -a /dev/md/root0 +wipefs -a /dev/md/data0-encrypted +wipefs -a /dev/md/data1-encrypted + +# Disable RAID recovery. We don't want this to slow down machine provisioning +# in the rescue mode. It can run in normal operation after reboot. +echo 0 > /proc/sys/dev/raid/speed_limit_max + +# LUKS encryption (--batch-mode to not ask) +cryptsetup --batch-mode luksFormat /dev/md/data0-encrypted /root/benacofs-luks-key +cryptsetup --batch-mode luksFormat /dev/md/data1-encrypted /root/benacofs-luks-key + +# Decrypt +cryptsetup luksOpen /dev/md/data0-encrypted data0-unencrypted --key-file /root/benacofs-luks-key +cryptsetup luksOpen /dev/md/data1-encrypted data1-unencrypted --key-file /root/benacofs-luks-key + +# LVM +# PVs +pvcreate /dev/mapper/data0-unencrypted +pvcreate /dev/mapper/data1-unencrypted +# VGs +vgcreate vg0 /dev/mapper/data0-unencrypted /dev/mapper/data1-unencrypted +# LVs +lvcreate --extents 95%FREE -n benacofs vg0 # 5% slack space + +# Filesystems (-F to not ask on preexisting FS) +mkfs.fat -F 32 -n esp0 /dev/disk/by-partlabel/ESP-partition0 +mkfs.fat -F 32 -n esp1 /dev/disk/by-partlabel/ESP-partition1 +mkfs.ext4 -F -L root /dev/md/root0 +mkfs.ext4 -F -L benacofs /dev/mapper/vg0-benacofs + +# Creating file systems changes their UUIDs. +# Trigger udev so that the entries in /dev/disk/by-uuid get refreshed. +# `nixos-generate-config` depends on those being up-to-date. +# See https://github.com/NixOS/nixpkgs/issues/62444 +udevadm trigger + +# Wait for FS labels to appear +udevadm settle --timeout=5 --exit-if-exists=/dev/disk/by-label/root +udevadm settle --timeout=5 --exit-if-exists=/dev/disk/by-label/benacofs + +# NixOS pre-installation mounts + +# Mount target root partition +mount /dev/disk/by-label/root /mnt +# Mount efivars unless already mounted +# (OVH rescue doesn't have them by default and the NixOS installer needs this) +mount | grep efivars || mount -t efivarfs efivarfs /sys/firmware/efi/efivars +# Mount our ESP partitions +mkdir -p /mnt/boot/ESP0 +mkdir -p /mnt/boot/ESP1 +mount /dev/disk/by-label/esp0 /mnt/boot/ESP0 +mount /dev/disk/by-label/esp1 /mnt/boot/ESP1 + +# Installing nix + +# Allow installing nix as root, see +# https://github.com/NixOS/nix/issues/936#issuecomment-475795730 +mkdir -p /etc/nix +echo "build-users-group =" > /etc/nix/nix.conf + +curl -L https://nixos.org/nix/install | sh +set +u +x # sourcing this may refer to unset variables that we have no control over +. $HOME/.nix-profile/etc/profile.d/nix.sh +set -u -x + +nix-channel --add https://nixos.org/channels/nixos-19.03 nixpkgs +nix-channel --update + +# Getting NixOS installation tools +nix-env -iE "_: with import { configuration = {}; }; with config.system.build; [ nixos-generate-config nixos-install nixos-enter manual.manpages ]" + +nixos-generate-config --root /mnt + +# On the OVH rescue mode, the default Internet interface is called `eth0`. +# Find what its name will be under NixOS, which uses stable interface names. +# See https://major.io/2015/08/21/understanding-systemds-predictable-network-device-names/#comment-545626 +INTERFACE=$(udevadm info -e | grep -A 11 ^P.*eth0 | grep -o -E 'ID_NET_NAME_ONBOARD=\w+' | cut -d= -f2) +echo "Determined INTERFACE as $INTERFACE" + +IP_V4=$(ip route get 8.8.8.8 | head -1 | cut -d' ' -f8) +echo "Determined IP_V4 as $IP_V4" + +# From https://stackoverflow.com/questions/1204629/how-do-i-get-the-default-gateway-in-linux-given-the-destination/15973156#15973156 +read _ _ DEFAULT_GATEWAY _ < <(ip route list match 0/0); echo "$DEFAULT_GATEWAY" +echo "Determined DEFAULT_GATEWAY as $DEFAULT_GATEWAY" + + +# Generate `configuration.nix`. Note that we splice in shell variables. +cat > /mnt/etc/nixos/configuration.nix <' (using the system hostname). + # This results mdadm considering such disks as "foreign" as opposed to + # "local", and showing them as e.g. '/dev/md/benaco-cdn:data0' + # instead of '/dev/md/data0'. + # This is mdadm's protection against accidentally putting a RAID disk + # into the wrong machine and corrupting data by accidental sync, see + # https://bugzilla.redhat.com/show_bug.cgi?id=606481#c14 and onward. + # We set the HOMEHOST manually go get the short '/dev/md' names, + # and so that things look and are configured the same on all such CDN + # machines irrespective of host names. + # We do not worry about plugging disks into the wrong machine because + # we will never exchange disks between CDN machines. + environment.etc."mdadm.conf".text = '' + HOMEHOST benaco-cdn + ''; + # The RAIDs are assembled in stage1, so we need to make the config + # available there. + boot.initrd.mdadmConf = config.environment.etc."mdadm.conf".text; + + # Network (OVH uses static IP assignments, no DHCP) + networking.useDHCP = false; + networking.interfaces."$INTERFACE".ipv4.addresses = [ + { + address = "$IP_V4"; + prefixLength = 24; + } + ]; + networking.defaultGateway = "$DEFAULT_GATEWAY"; + networking.nameservers = [ "8.8.8.8" ]; + + # Initial empty root password for easy login: + users.users.root.initialHashedPassword = ""; + services.openssh.permitRootLogin = "prohibit-password"; + + users.users.root.openssh.authorizedKeys.keys = [ + "ssh-rsa AAAAB3NzaC1yc2EAAAABIwAAAQEAtwCIGPYJlD2eeUtxngmT+4yR7BMlK0F5kzj+84uHsxxsy+PXFrP/tScCpwmuoiEYNv/9WKnPJJfCA9XlIDr6cla1MLpaW6eg672TRYMmKzH6SLlkg+kyDmPxSIJw+KdKfnPYyva+Y/VocACYJo0voabUeLAVgtSKGz/AFzccjfOR0GmFO911zjAaR+jFb9M7t7dveNVKm9KbuBfu3giMgGg3/mKz1TKY8yk2ZOxpT5CllBb+B5BcEf+7IGNvNxr1Z0zz5cFXQ3LyBIZklnC/OaQCnD78BSiyPTkIXcmBFal2TaFwTDvki6PuCRpJy+dU1fDdgWLql97D0SVnjmmomw== nh2@deditus.de" + ]; + + services.openssh.enable = true; + + # This value determines the NixOS release with which your system is to be + # compatible, in order to avoid breaking some software such as database + # servers. You should change this only after NixOS release notes say you + # should. + system.stateVersion = "19.03"; # Did you read the comment? + +} +EOF + +# Install NixOS +PATH="$PATH" NIX_PATH="$NIX_PATH" `which nixos-install` --no-root-passwd --root /mnt --max-jobs 40 + +reboot