gitea-runner: detect existing LXC on re-run, refresh OS on update

Ports openbao's host-side update path to gitea-runner: main() now uses
find_existing_lxc (hostname/tag match) to switch into a new update_lxc()
instead of always recreating the container, with an explicit --update
dispatch case reaching update_runner() (previously inferred only from
/usr/local/bin/act_runner presence). require_root is ported from openbao
for the host-side pct branch.

Extracted a small exec_in_lxc() helper (option a from the two offered)
rather than duplicating the curl-pipe invocation, since main() now needs
to drive both the create and update paths through the same piping logic
with only the trailing --install/--update flag differing — matching
openbao's own exec_in_lxc for consistency.

update_runner() now calls refresh_os_packages before touching the binary,
mirroring openbao's update_inside_lxc(). Also switched create_lxc's LXC
tag from the hardcoded "cicd" to the LXC_TAG variable (added in the prior
commit but unused until now) so find_existing_lxc's tag match actually
works.

Documented the OS-refresh-on-update behavior in both READMEs.

Closes #15

gitea-runner/README.md

@@ -20,7 +20,7 @@ Single script, three automatic modes:
 bash -c "$(curl -fsSL https://gitea.arnodo.fr/Damien/infra-scripts/raw/branch/main/gitea-runner/install.sh)"
 ```
 
-The script automatically creates an Alpine 3.23 LXC with Docker and act_runner.
+The script automatically creates an Alpine LXC (template auto-detected from `pveam available`) with Docker and act_runner.
 
 #### Customization
 
@@ -34,6 +34,7 @@ CTID=120 HOSTNAME=runner-02 CORES=4 RAM=4096 bash -c "$(curl -fsSL https://gitea
 |----------|---------|-------------|
 | `CTID` | auto | Container ID |
 | `RUNNER_HOSTNAME` | `gitea-runner` | LXC Hostname |
+| `TEMPLATE` | auto-detected | Alpine template; auto-detected from `pveam available` |
 | `CORES` | `2` | CPU cores |
 | `RAM` | `2048` | RAM in MiB |
 | `DISK` | `8` | Disk in GB |
@@ -59,7 +60,7 @@ From inside the LXC:
 curl -fsSL https://gitea.arnodo.fr/Damien/infra-scripts/raw/branch/main/gitea-runner/install.sh | bash
 ```
 
-The script detects that act_runner is already installed and switches to update mode automatically.
+The script detects that act_runner is already installed and switches to update mode automatically. Re-running from the Proxmox host does the same, plus refreshes the LXC's Alpine packages first (`apk update && apk upgrade`).
 
 ### Architecture
 

gitea-runner/install.sh

@@ -8,14 +8,15 @@ set -euo pipefail
 
 # --- Config (override via environment) ---
 CTID="${CTID:-}"
-HOSTNAME="${RUNNER_HOSTNAME:-gitea-runner}"
+HOSTNAME_LXC="${RUNNER_HOSTNAME:-gitea-runner}"
-TEMPLATE="${TEMPLATE:-alpine-3.23-default_20260116_amd64.tar.xz}"
+TEMPLATE="${TEMPLATE:-}"                          # auto-detected when empty
 STORAGE="${STORAGE:-local-lvm}"
 TEMPLATE_STORAGE="${TEMPLATE_STORAGE:-local}"
 CORES="${CORES:-2}"
 RAM="${RAM:-2048}"
 DISK="${DISK:-8}"
 BRIDGE="${BRIDGE:-vmbr0}"
+LXC_TAG="${LXC_TAG:-gitea-runner}"                # stable identifier for the container
 SCRIPT_URL="https://gitea.arnodo.fr/Damien/infra-scripts/raw/branch/main/gitea-runner/install.sh"
 GITEA_HOSTNAME="${GITEA_HOSTNAME:-gitea.taila5ad8.ts.net}"
 GITEA_API="https://gitea.com/api/v1/repos/gitea/act_runner/releases"
@@ -31,6 +32,33 @@ log_info()  { echo -e "${GREEN}[INFO]${NC} $1"; }
 log_warn()  { echo -e "${YELLOW}[WARN]${NC} $1"; }
 log_error() { echo -e "${RED}[ERROR]${NC} $1"; }
 
+require_root() {
+  if [[ "$(id -u)" -ne 0 ]]; then
+    log_error "This script must be run as root (current uid: $(id -u))."
+    log_error "On Proxmox, launch it from the host shell or via the Web UI shell, both of which run as root."
+    exit 1
+  fi
+}
+
+# ============================================================
+# Load shared helpers (lib/common.sh: detect_latest_alpine_template,
+# enable_tty1_autologin, find_existing_lxc, refresh_os_packages).
+#
+# Same reasoning as openbao/install.sh: a local checkout has the file
+# on disk right next to us, but the documented curl one-liner (host or
+# piped into `pct exec` inside the LXC) has no BASH_SOURCE path worth
+# trusting, so fall back to fetching lib/common.sh over HTTP next to
+# SCRIPT_URL. The LXC already needs outbound network to curl this very
+# script and to download the act_runner binary, so this adds no new
+# failure mode.
+# ============================================================
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]:-.}")" 2>/dev/null && pwd || true)"
+if [[ -n "$SCRIPT_DIR" && -f "${SCRIPT_DIR}/../lib/common.sh" ]]; then
+  source "${SCRIPT_DIR}/../lib/common.sh"
+else
+  source <(curl -fsSL "$(dirname "$(dirname "$SCRIPT_URL")")/lib/common.sh")
+fi
+
 # --- Helpers ---
 get_latest_release() {
   local release
@@ -67,12 +95,34 @@ download_runner() {
   echo "$release" > "$VERSION_FILE"
 }
 
+# Inject the script into the container and execute it in the requested mode.
+# Forwards the runtime configuration the inner invocation needs to reproduce
+# what the user requested on the host (mirrors openbao/install.sh's helper
+# of the same name).
+exec_in_lxc() {
+  local ctid="$1"
+  local mode="$2"   # --install or --update
+
+  pct exec "$ctid" -- sh -c "apk add --no-cache bash curl jq > /dev/null 2>&1"
+  curl -fsSL "$SCRIPT_URL" \
+    | pct exec "$ctid" -- env \
+        SCRIPT_URL="$SCRIPT_URL" \
+        GITEA_HOSTNAME="$GITEA_HOSTNAME" \
+        bash -s -- "$mode"
+}
+
 # ============================================================
 # MODE 1: Proxmox host — create LXC container
 # ============================================================
 create_lxc() {
   log_info "=== Gitea Act Runner — LXC Creation ==="
 
+  if [[ -z "$TEMPLATE" ]]; then
+    TEMPLATE=$(detect_latest_alpine_template)
+  else
+    log_info "Using user-provided template: $TEMPLATE"
+  fi
+
   # Auto-select next CTID if not specified
   if [[ -z "$CTID" ]]; then
     CTID=$(pvesh get /cluster/resources --type vm --output-format json 2>/dev/null \
@@ -86,16 +136,16 @@ create_lxc() {
     pveam download "$TEMPLATE_STORAGE" "$TEMPLATE"
   fi
 
-  log_info "Creating LXC $CTID ($HOSTNAME)..."
+  log_info "Creating LXC $CTID ($HOSTNAME_LXC)..."
   pct create "$CTID" "${TEMPLATE_STORAGE}:vztmpl/${TEMPLATE}" \
-    --hostname "$HOSTNAME" \
+    --hostname "$HOSTNAME_LXC" \
     --cores "$CORES" \
     --memory "$RAM" \
     --rootfs "${STORAGE}:${DISK}" \
     --net0 "name=eth0,bridge=${BRIDGE},ip=dhcp" \
     --unprivileged 1 \
     --features nesting=1,keyctl=1 \
-    --tags "infra-script,cicd" \
+    --tags "infra-script,${LXC_TAG}" \
     --start 0
 
   log_info "Configuring LXC for Docker and Tailscale..."
@@ -111,8 +161,7 @@ EOF
   sleep 5
 
   log_info "Injecting install script into container..."
-  pct exec "$CTID" -- sh -c "apk add --no-cache bash curl jq > /dev/null 2>&1"
+  exec_in_lxc "$CTID" "--install"
-  curl -fsSL "$SCRIPT_URL" | pct exec "$CTID" -- bash -s -- --install
 
   local ip
   ip=$(pct exec "$CTID" -- ip -4 addr show eth0 2>/dev/null | awk '/inet /{print $2}' | cut -d/ -f1)
@@ -122,7 +171,7 @@ EOF
   log_info "LXC $CTID created successfully!"
   log_info "========================================="
   echo ""
-  echo "  Hostname : $HOSTNAME"
+  echo "  Hostname : $HOSTNAME_LXC"
   echo "  IP       : ${ip:-pending}"
   echo ""
   echo "Next steps:"
@@ -133,6 +182,28 @@ EOF
   echo ""
 }
 
+# ============================================================
+# MODE 1b: Proxmox host — update an existing LXC
+# ============================================================
+update_lxc() {
+  local ctid="$1"
+  log_info "=== Gitea Act Runner — updating existing LXC ${ctid} ==="
+
+  if ! pct status "$ctid" | grep -q running; then
+    log_info "Starting LXC ${ctid}..."
+    pct start "$ctid"
+    sleep 3
+  fi
+
+  log_info "Refreshing Alpine packages inside LXC ${ctid}..."
+  pct exec "$ctid" -- sh -c "apk update >/dev/null && apk upgrade >/dev/null"
+
+  log_info "Upgrading act_runner binary inside LXC ${ctid}..."
+  exec_in_lxc "$ctid" "--update"
+
+  log_info "Update of LXC ${ctid} complete."
+}
+
 # ============================================================
 # MODE 2: Inside LXC — fresh install
 # ============================================================
@@ -232,24 +303,7 @@ LOGROTATE
 
   ln -sf /usr/sbin/logrotate /etc/periodic/daily/logrotate 2>/dev/null || true
 
-  log_info "Enabling console auto-login on tty1..."
+  enable_tty1_autologin
-  # Alpine ships busybox getty by default; agetty (from util-linux) is what
-  # supports --autologin.
-  apk add --no-cache agetty >/dev/null 2>&1 || apk add --no-cache util-linux >/dev/null
-
-  # Replace any existing tty1 entry, then append our autologin line. Doing it
-  # in two steps (delete + append) is more robust than an in-place sed against
-  # a pattern that may drift across Alpine releases.
-  sed -i '/^tty1::/d' /etc/inittab
-  echo 'tty1::respawn:/sbin/agetty --autologin root --noclear 38400 tty1' >> /etc/inittab
-
-  # Tell PID 1 to re-read /etc/inittab so the change takes effect without a reboot.
-  kill -HUP 1 2>/dev/null || true
-
-  # Kick any getty/agetty still attached to tty1 so init respawns it *now* with
-  # the new line — otherwise the first web-console session lands on the stale
-  # process and the operator has to type `exit` once before autologin kicks in.
-  pkill -KILL -f '(getty|agetty).*tty1' 2>/dev/null || true
 
   log_info "Cleaning up..."
   rm -rf /var/cache/apk/*
@@ -275,6 +329,8 @@ LOGROTATE
 update_runner() {
   log_info "=== Gitea Act Runner — Update ==="
 
+  refresh_os_packages
+
   local release
   release=$(get_latest_release)
 
@@ -305,12 +361,27 @@ update_runner() {
 # Main — detect context
 # ============================================================
 main() {
-  if [[ "${1:-}" == "--install" ]]; then
+  case "${1:-}" in
-    # Explicitly called in install mode (from pct exec)
+    --install)
-    install_runner
+      install_runner
-  elif command -v pct &> /dev/null; then
+      return
+      ;;
+    --update)
+      update_runner
+      return
+      ;;
+  esac
+
+  if command -v pct &> /dev/null; then
     # We're on the Proxmox host
-    create_lxc
+    require_root
+    local existing=""
+    if existing=$(find_existing_lxc); then
+      log_info "Found existing gitea-runner LXC (CTID ${existing}, hostname/tag match) — switching to update mode."
+      update_lxc "$existing"
+    else
+      create_lxc
+    fi
   elif [[ -f /usr/local/bin/act_runner ]]; then
     # act_runner exists — update mode
     update_runner

lib/common.sh

@@ -26,7 +26,18 @@ set -euo pipefail
 # template if `pveam` is unavailable or returns nothing.
 # ============================================================
 detect_latest_alpine_template() {
-  :
+  local tmpl
+  tmpl=$(pveam available --section system 2>/dev/null \
+    | awk '/^system[[:space:]]+alpine-/ {print $2}' \
+    | sort -V \
+    | tail -n1)
+
+  if [[ -z "$tmpl" ]]; then
+    log_warn "Could not query pveam; falling back to a known-good Alpine template."
+    tmpl="alpine-3.22-default_20250617_amd64.tar.xz"
+  fi
+  log_info "Selected Alpine template: $tmpl"
+  echo "$tmpl"
 }
 
 # ============================================================
@@ -34,7 +45,24 @@ detect_latest_alpine_template() {
 # Idempotent: safe to call on every install/update.
 # ============================================================
 enable_tty1_autologin() {
-  :
+  log_info "Enabling console auto-login on tty1..."
+  # Alpine ships busybox getty by default; agetty (from util-linux) is what
+  # supports --autologin.
+  apk add --no-cache agetty >/dev/null 2>&1 || apk add --no-cache util-linux >/dev/null
+
+  # Replace any existing tty1 entry, then append our autologin line. Doing it
+  # in two steps (delete + append) is more robust than an in-place sed against
+  # a pattern that may drift across Alpine releases.
+  sed -i '/^tty1::/d' /etc/inittab
+  echo 'tty1::respawn:/sbin/agetty --autologin root --noclear 38400 tty1' >> /etc/inittab
+
+  # Tell PID 1 to re-read /etc/inittab so the change takes effect without a reboot.
+  kill -HUP 1 2>/dev/null || true
+
+  # Kick any getty/agetty still attached to tty1 so init respawns it *now* with
+  # the new line — otherwise the first web-console session lands on the stale
+  # process and the operator has to type `exit` once before autologin kicks in.
+  pkill -KILL -f '(getty|agetty).*tty1' 2>/dev/null || true
 }
 
 # ============================================================
@@ -45,7 +73,17 @@ enable_tty1_autologin() {
 # find_existing_lxc already does).
 # ============================================================
 find_existing_lxc() {
-  :
+  local id host tags
+  while read -r id _; do
+    [[ -z "$id" || "$id" == "VMID" ]] && continue
+    host=$(pct config "$id" 2>/dev/null | awk -F': ' '/^hostname:/ {print $2}' || true)
+    tags=$(pct config "$id" 2>/dev/null | awk -F': ' '/^tags:/ {print $2}' || true)
+    if [[ "$host" == "$HOSTNAME_LXC" ]] || [[ ",${tags//;/,}," == *",${LXC_TAG},"* ]]; then
+      echo "$id"
+      return 0
+    fi
+  done < <(pct list | awk 'NR>1 {print $1}')
+  return 1
 }
 
 # ============================================================
@@ -53,5 +91,6 @@ find_existing_lxc() {
 # Callable both host-side (via pct exec) and inside the LXC.
 # ============================================================
 refresh_os_packages() {
-  :
+  log_info "Refreshing Alpine packages..."
+  apk update >/dev/null && apk upgrade >/dev/null
 }

openbao/README.md

@@ -110,7 +110,9 @@ curl -fsSL https://gitea.arnodo.fr/Damien/infra-scripts/raw/branch/feat/lxc-Open
 
 The script auto-detects the presence of `/usr/local/bin/bao` and switches to
 update mode. The OpenRC service is stopped, the binary is swapped (the old one
-is kept as `bao.bak.<ts>`), then the service is restarted.
+is kept as `bao.bak.<ts>`), then the service is restarted. Re-running from the
+Proxmox host does the same, plus refreshes the LXC's Alpine packages first
+(`apk update && apk upgrade`).
 
 ### Architecture
 

openbao/install.sh

@@ -56,6 +56,29 @@ log_info()  { echo -e "${GREEN}[INFO]${NC} $1" >&2; }
 log_warn()  { echo -e "${YELLOW}[WARN]${NC} $1" >&2; }
 log_error() { echo -e "${RED}[ERROR]${NC} $1" >&2; }
 
+# ============================================================
+# Load shared helpers (lib/common.sh: detect_latest_alpine_template,
+# enable_tty1_autologin, find_existing_lxc, refresh_os_packages).
+#
+# This script runs in three different contexts, only one of which has a
+# real file on disk next to it:
+#   - local checkout (`bash openbao/install.sh`)         -> lib/common.sh
+#     sits right there at ../lib/common.sh, source it straight from disk.
+#   - Proxmox host, documented one-liner (`bash -c "$(curl ... )"`)
+#     -> no checkout, no BASH_SOURCE path worth trusting.
+#   - inside the LXC (exec_in_lxc does `curl ... | pct exec ... bash -s --`)
+#     -> same story, script arrives on stdin.
+# For the latter two we fetch lib/common.sh over HTTP, next to SCRIPT_URL.
+# The LXC already needs outbound network to curl this very script and to
+# download the bao binary, so this adds no new failure mode.
+# ============================================================
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]:-.}")" 2>/dev/null && pwd || true)"
+if [[ -n "$SCRIPT_DIR" && -f "${SCRIPT_DIR}/../lib/common.sh" ]]; then
+  source "${SCRIPT_DIR}/../lib/common.sh"
+else
+  source <(curl -fsSL "$(dirname "$(dirname "$SCRIPT_URL")")/lib/common.sh")
+fi
+
 # ============================================================
 # Generic helpers
 # ============================================================
@@ -204,37 +227,6 @@ configure_tailscale_proxy() {
 # Proxmox-host helpers
 # ============================================================
 
-# Detect newest Alpine LXC template available from the Proxmox repos.
-detect_latest_alpine_template() {
-  local tmpl
-  tmpl=$(pveam available --section system 2>/dev/null \
-    | awk '/^system[[:space:]]+alpine-/ {print $2}' \
-    | sort -V \
-    | tail -n1)
-
-  if [[ -z "$tmpl" ]]; then
-    log_warn "Could not query pveam; falling back to a known-good Alpine template."
-    tmpl="alpine-3.22-default_20250617_amd64.tar.xz"
-  fi
-  log_info "Selected Alpine template: $tmpl"
-  echo "$tmpl"
-}
-
-# Find an existing LXC by tag or hostname. Echoes CTID, returns 1 if none.
-find_existing_lxc() {
-  local id host tags
-  while read -r id _; do
-    [[ -z "$id" || "$id" == "VMID" ]] && continue
-    host=$(pct config "$id" 2>/dev/null | awk -F': ' '/^hostname:/ {print $2}' || true)
-    tags=$(pct config "$id" 2>/dev/null | awk -F': ' '/^tags:/ {print $2}' || true)
-    if [[ "$host" == "$HOSTNAME_LXC" ]] || [[ ",${tags//;/,}," == *",${LXC_TAG},"* ]]; then
-      echo "$id"
-      return 0
-    fi
-  done < <(pct list | awk 'NR>1 {print $1}')
-  return 1
-}
-
 ensure_template_present() {
   local tmpl="$1"
   if ! pveam list "$TEMPLATE_STORAGE" 2>/dev/null | grep -q "$tmpl"; then
@@ -363,6 +355,9 @@ update_lxc() {
   fi
 
   log_info "Refreshing Alpine packages inside LXC ${ctid}..."
+  # refresh_os_packages() is a bash function local to this process; it can't
+  # run over `pct exec ... sh -c` without shipping the function definition
+  # into the container, so this call site stays inline rather than dedupe.
   pct exec "$ctid" -- sh -c "apk update >/dev/null && apk upgrade >/dev/null"
 
   log_info "Upgrading bao binary inside LXC ${ctid}..."
@@ -474,24 +469,7 @@ EOF
   log_info "Starting openbao service..."
   rc-service openbao start || log_warn "openbao failed to start — inspect /var/log/openbao.log"
 
-  log_info "Enabling console auto-login on tty1..."
+  enable_tty1_autologin
-  # Alpine ships busybox getty by default; agetty (from util-linux) is what
-  # supports --autologin.
-  apk add --no-cache agetty >/dev/null 2>&1 || apk add --no-cache util-linux >/dev/null
-
-  # Replace any existing tty1 entry, then append our autologin line. Doing it
-  # in two steps (delete + append) is more robust than an in-place sed against
-  # a pattern that may drift across Alpine releases.
-  sed -i '/^tty1::/d' /etc/inittab
-  echo 'tty1::respawn:/sbin/agetty --autologin root --noclear 38400 tty1' >> /etc/inittab
-
-  # Tell PID 1 to re-read /etc/inittab so the change takes effect without a reboot.
-  kill -HUP 1 2>/dev/null || true
-
-  # Kick any getty/agetty still attached to tty1 so init respawns it *now* with
-  # the new line — otherwise the first web-console session lands on the stale
-  # process and the operator has to type `exit` once before autologin kicks in.
-  pkill -KILL -f '(getty|agetty).*tty1' 2>/dev/null || true
 
   configure_tailscale_proxy
 
@@ -563,8 +541,7 @@ MOTD
 # ============================================================
 update_inside_lxc() {
   log_info "=== OpenBao — update ==="
-  apk update >/dev/null
+  refresh_os_packages
-  apk upgrade >/dev/null
   install_or_upgrade_bao
   configure_tailscale_proxy
   log_info "Update complete."