Files
QUptime/install.sh
T

336 lines
12 KiB
Bash

#!/bin/bash
# QUptime installer.
#
# Downloads the latest released `qu` binary, verifies it against the
# published SHA256SUMS, installs it to /usr/local/bin, and (on systemd
# hosts) drops in a hardened quptime.service that matches the unit
# documented in docs/deployment/systemd.md.
#
# Release sources, tried in order:
# 1. Gitea: git.cer.sh/axodouble/quptime/releases (primary — canonical home)
# 2. GitHub: github.com/Axodouble/QUptime/releases (push-mirror fallback)
#
# Idempotent — re-running upgrades the binary and refreshes the unit
# without touching the data directory.
set -euo pipefail
INSTALL_BIN="/usr/local/bin/qu"
SERVICE_FILE="/etc/systemd/system/quptime.service"
SERVICE_NAME="$(basename "$SERVICE_FILE")"
SERVICE_USER="quptime"
SERVICE_GROUP="quptime"
DATA_DIR="/etc/quptime"
# Release sources, in preference order. Each row is:
# <name>|<latest-release API endpoint>|<release-asset base URL>
# The asset URL is concatenated with `/<tag>/<filename>`. Adjust here
# if the project moves hosts.
SOURCES=(
"gitea|https://git.cer.sh/api/v1/repos/axodouble/quptime/releases/latest|https://git.cer.sh/axodouble/quptime/releases/download"
"github|https://api.github.com/repos/Axodouble/QUptime/releases/latest|https://github.com/Axodouble/QUptime/releases/download"
)
fail() {
echo "Error: $*" >&2
exit 1
}
require_command() {
command -v "$1" >/dev/null 2>&1 || fail "$1 is not installed. Please install $1 and try again."
}
write_completion() {
local shell=$1 path=$2
[ -d "$(dirname "$path")" ] || return 1
if "$INSTALL_BIN" completion "$shell" > "$path" 2>/dev/null; then
echo "> installed $shell completion -> $path"
return 0
fi
rm -f "$path"
return 1
}
# fetch_from_source tries one release source end-to-end: pulls the
# latest tag from its API, downloads the per-arch binary and the
# accompanying SHA256SUMS, and verifies the checksum. Returns 0 on
# success (with RELEASE and BINARY_NAME set as globals) or 1 if any
# step fails — callers can then try the next source. Stderr is kept
# quiet so a failed primary doesn't spam the operator before the
# fallback is attempted.
fetch_from_source() {
local api_url=$1
local release_base=$2
local tmpdir=$3
local release
release=$(curl -fsSL --proto '=https' --tlsv1.2 "$api_url" 2>/dev/null | jq -r '.tag_name' 2>/dev/null) \
|| return 1
[ -n "$release" ] && [ "$release" != "null" ] || return 1
local binary_name="qu-${release}-linux-${ARCH}"
local binary_url="${release_base}/${release}/${binary_name}"
local sums_url="${release_base}/${release}/SHA256SUMS"
curl -fsSL --proto '=https' --tlsv1.2 -o "$tmpdir/$binary_name" "$binary_url" 2>/dev/null \
|| return 1
curl -fsSL --proto '=https' --tlsv1.2 -o "$tmpdir/SHA256SUMS" "$sums_url" 2>/dev/null \
|| return 1
# Verify against the SHA256SUMS that came from the same source as
# the binary. Never mix sources here — verifying a GitHub-hosted
# binary against a Gitea-hosted SHA256SUMS would defeat the
# tamper check.
(
cd "$tmpdir"
if ! grep -E "[[:space:]]\\*?${binary_name}\$" SHA256SUMS > expected.sum; then
exit 1
fi
if ! sha256sum -c expected.sum >/dev/null 2>&1; then
exit 1
fi
) || return 1
RELEASE="$release"
BINARY_NAME="$binary_name"
return 0
}
require_command curl
require_command jq
require_command sha256sum
require_command install
require_command mktemp
# --- target architecture ------------------------------------------------
case "$(uname -m)" in
x86_64) ARCH=amd64 ;;
aarch64|arm64) ARCH=arm64 ;;
*) fail "unsupported architecture: $(uname -m). Pre-built binaries are published for amd64 and arm64 only — build from source for other platforms." ;;
esac
if [ ! -w "$(dirname "$INSTALL_BIN")" ]; then
fail "Cannot write to $(dirname "$INSTALL_BIN"). Run this script with sudo, or set INSTALL_BIN to a writable location."
fi
# --- download + verify (with fallback) ----------------------------------
TMPDIR=$(mktemp -d)
trap 'rm -rf "$TMPDIR"' EXIT
# Globals filled in by fetch_from_source on success.
RELEASE=""
BINARY_NAME=""
INSTALLED_FROM=""
INSTALLED_TMP=""
for source_spec in "${SOURCES[@]}"; do
IFS='|' read -r src_name src_api src_base <<<"$source_spec"
src_tmp="$TMPDIR/$src_name"
mkdir -p "$src_tmp"
echo "> trying release source: $src_name"
# `set -e` would abort the whole script the moment fetch_from_source
# returns nonzero; we want the loop to fall through to the next
# source instead. Wrap the call so a failure is just data.
if fetch_from_source "$src_api" "$src_base" "$src_tmp"; then
INSTALLED_FROM="$src_name"
INSTALLED_TMP="$src_tmp"
echo "> $src_name: ${RELEASE} ✓ checksum OK"
break
fi
echo "> $src_name: unavailable"
done
if [ -z "$INSTALLED_FROM" ]; then
fail "no release source reachable — tried: $(printf '%s ' "${SOURCES[@]%%|*}"). Check network access to git.cer.sh and github.com."
fi
install -m 0755 "$INSTALLED_TMP/$BINARY_NAME" "$INSTALL_BIN"
echo "> qu ${RELEASE} installed to $INSTALL_BIN (source: $INSTALLED_FROM)"
# --- shell completions --------------------------------------------------
if "$INSTALL_BIN" --help 2>/dev/null | grep -q "completion"; then
write_completion bash /usr/share/bash-completion/completions/qu \
|| write_completion bash /etc/bash_completion.d/qu \
|| true
write_completion zsh /usr/share/zsh/site-functions/_qu || true
write_completion fish /usr/share/fish/vendor_completions.d/qu.fish || true
else
echo "> qu does not expose completion support; skipping shell completion installation."
fi
# --- systemd unit -------------------------------------------------------
if ! command -v systemctl >/dev/null 2>&1; then
echo
echo "> systemd is not available on this system. Installation stops here."
echo "> Run \`qu serve\` manually (or wire it into the supervisor of your choice)."
exit 0
fi
# Dedicated service user. Hardened unit drops all capabilities and
# locks the daemon down with ProtectSystem=strict, so it must run as
# its own unprivileged account rather than the invoking sudo user.
if ! id "$SERVICE_USER" >/dev/null 2>&1; then
echo "> creating system user $SERVICE_USER"
useradd --system --no-create-home --shell /usr/sbin/nologin "$SERVICE_USER"
fi
install -d -o "$SERVICE_USER" -g "$SERVICE_GROUP" -m 0750 "$DATA_DIR"
# Repair ownership and permissions on the data dir's contents. Catches:
# - re-running the installer over a previous install where the
# service user/group changed.
# - the operator ran `qu init` or `qu serve` as root once (easy
# mistake: `sudo qu init` is shorter than the documented
# `sudo -u quptime qu init`). When the daemon runs as root its
# DataDir() resolves to /etc/quptime, so any files it writes land
# owned by root:root — the systemd service then fails with
# `open node.yaml: permission denied`.
# - someone or something (a stray `chmod -R`, a misguided backup
# restore) tightened or loosened modes. Re-running the installer
# should be enough to get back to a working baseline.
# The canonical layout (mirrors the modes the daemon writes itself
# in internal/config and internal/crypto):
# /etc/quptime/ quptime:quptime 0750
# /etc/quptime/keys/ quptime:quptime 0700
# /etc/quptime/node.yaml quptime:quptime 0600
# /etc/quptime/cluster.yaml quptime:quptime 0600
# /etc/quptime/trust.yaml quptime:quptime 0600
# /etc/quptime/keys/private.pem quptime:quptime 0600
# /etc/quptime/keys/public.pem quptime:quptime 0644
# /etc/quptime/keys/cert.pem quptime:quptime 0644
# The runtime dir /var/run/quptime is owned by systemd via
# RuntimeDirectory= and rebuilt at each service start, so we leave it
# alone.
repair_perms() {
# Always reset the top-level dir mode — `install -d` only sets it
# on creation, not on re-run.
chown "$SERVICE_USER:$SERVICE_GROUP" "$DATA_DIR"
chmod 0750 "$DATA_DIR"
# Reassert ownership across the whole tree in one pass.
if [ -n "$(ls -A "$DATA_DIR" 2>/dev/null)" ]; then
chown -R "$SERVICE_USER:$SERVICE_GROUP" "$DATA_DIR"
fi
# keys/ is a directory with its own tighter mode.
if [ -d "$DATA_DIR/keys" ]; then
chmod 0700 "$DATA_DIR/keys"
fi
# Each known file gets its canonical mode if it exists. We don't
# create anything that isn't already there — that's `qu init`'s
# job — and we don't touch unknown files an operator may have
# parked in the dir.
local f
for f in node.yaml cluster.yaml trust.yaml keys/private.pem; do
[ -f "$DATA_DIR/$f" ] && chmod 0600 "$DATA_DIR/$f"
done
for f in keys/public.pem keys/cert.pem; do
[ -f "$DATA_DIR/$f" ] && chmod 0644 "$DATA_DIR/$f"
done
}
repair_perms
echo "> reasserted ownership ($SERVICE_USER:$SERVICE_GROUP) and modes under $DATA_DIR"
echo "> writing $SERVICE_FILE"
cat > "$SERVICE_FILE" <<'EOF'
[Unit]
Description=QUptime distributed uptime monitor
Documentation=https://git.cer.sh/axodouble/quptime
Wants=network-online.target
After=network-online.target
[Service]
Type=simple
ExecStart=/usr/local/bin/qu serve
Restart=always
RestartSec=5s
User=quptime
Group=quptime
# Where state lives. RuntimeDirectory creates /var/run/quptime/ each
# boot owned by User:Group with mode 0750.
Environment=QUPTIME_DIR=/etc/quptime
RuntimeDirectory=quptime
RuntimeDirectoryMode=0750
ReadWritePaths=/etc/quptime /var/run/quptime
# Hardening. Comment out individual directives if a probe needs
# something we've revoked.
NoNewPrivileges=true
ProtectSystem=strict
ProtectHome=true
PrivateTmp=true
PrivateDevices=true
ProtectKernelTunables=true
ProtectKernelModules=true
ProtectControlGroups=true
ProtectClock=true
ProtectHostname=true
RestrictNamespaces=true
RestrictRealtime=true
RestrictSUIDSGID=true
LockPersonality=true
MemoryDenyWriteExecute=true
# Network access is required (we're a network monitor). Keep address
# families minimal — AF_NETLINK is needed for some libc lookups.
RestrictAddressFamilies=AF_UNIX AF_INET AF_INET6 AF_NETLINK
# If you need raw ICMP, *also* uncomment:
# AmbientCapabilities=CAP_NET_RAW
# CapabilityBoundingSet=CAP_NET_RAW
# Otherwise drop all capabilities:
CapabilityBoundingSet=
[Install]
WantedBy=multi-user.target
EOF
systemctl daemon-reload
systemctl enable "$SERVICE_NAME" >/dev/null
echo "> ${SERVICE_NAME} installed and enabled (not yet started)"
cat <<EOF
Installation complete.
Next steps:
1. Initialise the node identity. Either:
a) Let \`qu serve\` auto-init from environment variables.
Drop a systemd override like:
sudo systemctl edit ${SERVICE_NAME}
[Service]
Environment=QUPTIME_ADVERTISE=<this-host>:9901
# On follower nodes, also set the shared join secret:
# Environment=QUPTIME_CLUSTER_SECRET=<paste from first node>
b) Or run \`qu init\` once explicitly. IMPORTANT: run as the
${SERVICE_USER} user, not root — otherwise node.yaml lands
owned by root and the service can't read it on start.
sudo -u ${SERVICE_USER} QUPTIME_DIR=${DATA_DIR} \\
qu init --advertise <this-host>:9901
If you already ran it as root and the service is failing
with "permission denied" on node.yaml, repair with:
sudo chown -R ${SERVICE_USER}:${SERVICE_GROUP} ${DATA_DIR}
2. Start the service:
sudo systemctl start ${SERVICE_NAME}
sudo -u ${SERVICE_USER} qu status
3. For ICMP checks, the daemon defaults to unprivileged UDP-mode
pings — those need the ping_group_range sysctl widened to include
the ${SERVICE_USER} GID, or grant CAP_NET_RAW in the unit. See
docs/deployment/systemd.md for the recipes.
Full documentation: https://git.cer.sh/axodouble/quptime/src/branch/master/docs
EOF