fix: harden stack validation and lifecycle execution

manage.conf

@@ -1,7 +1,12 @@
 # Maximum number of stacks executed concurrently within each stage.
-# Use 0 for unlimited concurrency.
+# Use 0 for unlimited concurrency. Range: 0-65535; no leading zeros.
 max_parallel 3
 
+# Wait for running/healthy services on up, start, and update.
+# Healthchecks are needed to verify application readiness.
+# Range: 0-86400 seconds; 0 disables waiting. Default: 120 seconds.
+wait_timeout 120
+
 # Each declaration is an ordered stage; stacks on the same line run in parallel.
 # Add --parallel N before the stacks to override max_parallel for that stage.
 # Example: before --parallel 2 stack-a stack-b

manage.sh

@@ -9,6 +9,7 @@ CONFIG_FILE="${STACKS_CONFIG:-$STACKS_DIR/manage.conf}"
 LOCK_FILE="${STACKS_LOCK_FILE:-/run/lock/stacks-manage.lock}"
 
 DEFAULT_MAX_PARALLEL=2
+WAIT_TIMEOUT=120
 
 BEFORE_STAGES=()
 BEFORE_LIMITS=()
@@ -40,19 +41,32 @@ fi
 
 handle_signal() {
 	local status="$1"
+	local pid attempt alive
 	local -a pids=()
 
-	trap - INT TERM
+	trap '' INT TERM
-	mapfile -t pids < <(jobs -pr)
+	mapfile -t pids < <(jobs -p)
-	((${#pids[@]} == 0)) || kill -TERM "${pids[@]}" 2>/dev/null
+	for pid in "${pids[@]}"; do
-	wait 2>/dev/null
+		kill -TERM -- "-$pid" 2>/dev/null || :
+	done
+
+	# Allow five seconds for cleanup, then stop every remaining descendant.
+	for ((attempt = 0; attempt < 50; attempt++)); do
+		alive=0
+		for pid in "${pids[@]}"; do
+			kill -0 -- "-$pid" 2>/dev/null && alive=1
+		done
+		((alive)) || break
+		sleep 0.1
+	done
+	for pid in "${pids[@]}"; do
+		kill -KILL -- "-$pid" 2>/dev/null || :
+	done
+	((${#pids[@]} == 0)) || wait "${pids[@]}" 2>/dev/null
 	printf '\n'
 	exit "$status"
 }
 
-trap 'handle_signal 130' INT
-trap 'handle_signal 143' TERM
-
 usage() {
 	cat <<EOF
 Usage: $(basename "$0") COMMAND [STACK]
@@ -75,7 +89,9 @@ Inspect:
   help, -h, --help Show this help
 
 Stage configuration:
-  max_parallel N                    Set default concurrency (0 = unlimited)
+  max_parallel N                   Set concurrency, 0-65535 (0 = unlimited)
+  wait_timeout N                   Wait 0-86400 seconds for startup (default 120;
+                                   0 disables waiting)
   before [--parallel N] STACK...    Add an early stage
   after [--parallel N] STACK...     Add a late stage
   exclude STACK...                  Exclude stacks from bulk operations
@@ -83,6 +99,10 @@ Stage configuration:
   Each before/after declaration is one ordered stage; unlisted stacks form
   the default stage. Lifecycle commands run before -> default -> after,
   reversed for stop/down. A failed stage skips all remaining stages.
+  up/start/update wait for running/healthy services when waiting is enabled.
+  Services need healthchecks to verify readiness beyond the running state.
+  up/update validate every selected Compose configuration before making changes.
+  Numeric settings do not accept leading zeros.
 
 Config: $CONFIG_FILE
 EOF
@@ -100,11 +120,24 @@ contains() {
 	return 1
 }
 
-validate_parallel_limit() {
+validate_integer_range() {
-	local value="$1"
+	local kind="$1"
+	local value="$2"
+	local maximum="$3"
 
-	[[ "$value" =~ ^[0-9]+$ ]] ||
+	[[ "$value" =~ ^(0|[1-9][0-9]*)$ ]] &&
-		die "parallel limit must be a non-negative integer: $value"
+		((${#value} <= ${#maximum})) && ((value <= maximum)) ||
+		die "$kind must be an integer between 0 and $maximum without leading zeros: $value"
+}
+
+validate_parallel_limit() {
+	validate_integer_range "parallel limit" "$1" 65535
+}
+
+wait_timeout() {
+	[[ $# -eq 1 ]] || die "wait_timeout expects exactly one value"
+	validate_integer_range "wait timeout" "$1" 86400
+	WAIT_TIMEOUT="$1"
 }
 
 validate_name() {
@@ -175,7 +208,7 @@ exclude() {
 load_config() {
 	[[ -f "$CONFIG_FILE" ]] || return 0
 
-	# The trusted config calls max_parallel, before, after, and exclude above.
+	# The trusted config calls the configuration functions above.
 	source "$CONFIG_FILE" || die "cannot load config file: $CONFIG_FILE"
 }
 
@@ -186,7 +219,9 @@ discover_stacks() {
 
 	for compose_file in "$STACKS_DIR"/*/compose.yaml; do
 		stack="${compose_file%/compose.yaml}"
-		DISCOVERED_STACKS+=("${stack##*/}")
+		stack="${stack##*/}"
+		validate_name "stack directory" "$stack"
+		DISCOVERED_STACKS+=("$stack")
 	done
 }
 
@@ -293,6 +328,11 @@ run_stack() {
 	local stack="$1"
 	local action="$2"
 	local show_heading="${3:-1}"
+	local -a wait_args=()
+
+	if ((WAIT_TIMEOUT > 0)); then
+		wait_args=(--wait --wait-timeout "$WAIT_TIMEOUT")
+	fi
 
 	if [[ "$show_heading" == "1" ]]; then
 		echo
@@ -301,13 +341,14 @@ run_stack() {
 
 	case "$action" in
 	up)
-		compose "$stack" config -q &&
+		compose "$stack" up -d "${wait_args[@]}"
-			compose "$stack" up -d
+		;;
+	start)
+		compose "$stack" start "${wait_args[@]}"
 		;;
 	update)
-		compose "$stack" config -q &&
+		compose "$stack" pull &&
-			compose "$stack" pull &&
+			compose "$stack" up -d "${wait_args[@]}"
-			compose "$stack" up -d
 		;;
 	*)
 		compose "$stack" "$action"
@@ -315,6 +356,21 @@ run_stack() {
 	esac
 }
 
+preflight_configs() {
+	local stack
+	local -a failed_stacks=()
+
+	(($# > 0)) || return 0
+	printf '\n==> Checking Compose configurations\n'
+	for stack in "$@"; do
+		compose "$stack" config -q || failed_stacks+=("$stack")
+	done
+	if ((${#failed_stacks[@]} > 0)); then
+		printf 'ERROR: invalid Compose configuration: %s\n' "${failed_stacks[*]}" >&2
+		return 1
+	fi
+}
+
 run_stage() {
 	local stage_index="$1"
 	local action="$2"
@@ -596,59 +652,87 @@ acquire_lock() {
 		die "another manage.sh operation is already running"
 }
 
-ACTION="${1:-}"
+main() {
-TARGET="${2:-}"
+	local ACTION="${1:-}"
+	local TARGET="${2:-}"
+	local -a targets=()
 
-[[ $# -le 2 ]] || die "too many arguments"
+	[[ $# -le 2 ]] || die "too many arguments"
 
-case "$ACTION" in
+	case "$ACTION" in
-help | -h | --help)
+	help | -h | --help)
-	usage
+		usage
-	exit 0
+		exit 0
-	;;
+		;;
-esac
+	esac
 
-load_config
+	load_config
-discover_stacks
+	discover_stacks
-validate_config_stacks
+	validate_config_stacks
-build_execution_stages
+	build_execution_stages
 
-case "$ACTION" in
+	case "$ACTION" in
-up | start | stop | restart | down | pull | update)
+	up | start | stop | restart | down | pull | update)
-	acquire_lock
+		acquire_lock
 
-	if [[ -n "$TARGET" ]]; then
+		if [[ -n "$TARGET" ]]; then
+			require_target "$ACTION" "$TARGET"
+			targets=("$TARGET")
+		else
+			targets=("${MANAGED_STACKS[@]}")
+		fi
+		if [[ "$ACTION" == "up" || "$ACTION" == "update" ]]; then
+			preflight_configs "${targets[@]}" || return 1
+		fi
+
+		if [[ -n "$TARGET" ]]; then
+			run_stack "$TARGET" "$ACTION"
+		else
+			run_all_staged "$ACTION"
+		fi
+		;;
+	ps)
+		if [[ -n "$TARGET" ]]; then
+			require_target "$ACTION" "$TARGET"
+			run_stack "$TARGET" "$ACTION"
+		else
+			run_all_sequential "$ACTION"
+		fi
+		;;
+	logs)
 		require_target "$ACTION" "$TARGET"
-		run_stack "$TARGET" "$ACTION"
+		compose "$TARGET" logs -f --tail=100
-	else
+		;;
-		run_all_staged "$ACTION"
+	config)
-	fi
-	;;
-ps)
-	if [[ -n "$TARGET" ]]; then
 		require_target "$ACTION" "$TARGET"
-		run_stack "$TARGET" "$ACTION"
+		compose "$TARGET" config
-	else
+		;;
-		run_all_sequential "$ACTION"
+	images)
-	fi
+		list_images "$TARGET"
-	;;
+		;;
-logs)
+	list)
-	require_target "$ACTION" "$TARGET"
+		[[ -z "$TARGET" ]] || die "list does not accept a stack name"
-	compose "$TARGET" logs -f --tail=100
+		list_stacks
-	;;
+		;;
-config)
+	*)
-	require_target "$ACTION" "$TARGET"
+		usage
-	compose "$TARGET" config
+		exit 1
-	;;
+		;;
-images)
+	esac
-	list_images "$TARGET"
+}
-	;;
+
-list)
+if [[ "${BASH_SOURCE[0]}" == "$0" ]]; then
-	[[ -z "$TARGET" ]] || die "list does not accept a stack name"
+	trap 'handle_signal 130' INT
-	list_stacks
+	trap 'handle_signal 143' TERM
-	;;
+	# One isolated process group covers the scheduler and all Docker descendants.
-*)
+	# Disable job control inside it so parallel workers stay in the same group.
-	usage
+	set -m
-	exit 1
+	(
-	;;
+		set +m
-esac
+		trap - INT TERM
+		main "$@"
+	) <&0 &
+	main_pid=$!
+	set +m
+	wait "$main_pid"
+fi