fix/feat: update luci configuration pages, init scripts, and env setup tools

This commit is contained in:
10000ge10000
2026-03-17 14:29:45 +08:00
parent aba80fad5f
commit ee10bb0bd5
5 changed files with 260 additions and 78 deletions

View File

@@ -258,24 +258,33 @@ all) gw_bind="custom" ;; # custom = 0.0.0.0
esac
# 确保网关端口未被残留进程占用 (防止 restart 时 crash loop)
# v2026.3.14 优化: 快速轮询 + 批量 kill
_ensure_port_free() {
local p="$1" max_wait="${2:-5}" i=0
local p="$1"
local i=0 max_wait=10 # 10 * 0.2 = 2 秒
# 先检查端口是否已被占用
if command -v ss >/dev/null 2>&1; then
ss -tulnp 2>/dev/null | grep -q ":${p} " || return 0
else
netstat -tulnp 2>/dev/null | grep -q ":${p} " || return 0
fi
# 端口被占用,尝试清理
for occ_pid in $(pgrep -f "openclaw-gateway" 2>/dev/null); do
kill "$occ_pid" 2>/dev/null
done
while [ $i -lt $max_wait ]; do
if command -v ss >/dev/null 2>&1; then
ss -tulnp 2>/dev/null | grep -q ":${p} " || return 0
else
netstat -tulnp 2>/dev/null | grep -q ":${p} " || return 0
fi
# 尝试杀掉占用端口的 gateway 进程
if [ $i -eq 0 ]; then
local occ_pid
for occ_pid in $(pgrep -f "openclaw-gateway" 2>/dev/null); do
kill "$occ_pid" 2>/dev/null
done
fi
sleep 1
usleep 200000 2>/dev/null || sleep 0.2
i=$((i + 1))
done
# 最后手段: SIGKILL
local port_pid=""
if command -v ss >/dev/null 2>&1; then
@@ -283,7 +292,7 @@ _ensure_port_free() {
else
port_pid=$(netstat -tulnp 2>/dev/null | grep ":${p} " | sed -n 's|.* \([0-9]*\)/.*|\1|p' | head -1)
fi
[ -n "$port_pid" ] && kill -9 "$port_pid" 2>/dev/null && sleep 1
[ -n "$port_pid" ] && kill -9 "$port_pid" 2>/dev/null && usleep 300000 2>/dev/null
return 0
}
_ensure_port_free "$port"
@@ -350,36 +359,43 @@ stop_service() {
local port
port=$(uci -q get openclaw.main.port || echo "18789")
# 杀掉所有 openclaw / openclaw-gateway 残留进程
# (排除 web-pty.js 和 oc-config.sh它们由 pty 实例管理)
local pid
for pid in $(pgrep -f "openclaw-gateway" 2>/dev/null) \
$(pgrep -f "openclaw.*gateway.*run" 2>/dev/null); do
kill "$pid" 2>/dev/null
# v2026.3.14 优化: 快速终止进程树,减少等待时间
# 1) 先获取所有相关 PID
local gw_pids=""
gw_pids=$(pgrep -f "openclaw-gateway" 2>/dev/null)
gw_pids="$gw_pids $(pgrep -f "openclaw.*gateway.*run" 2>/dev/null)"
# 2) 同时发送 SIGTERM 给所有进程
for pid in $gw_pids; do
[ -n "$pid" ] && kill "$pid" 2>/dev/null
done
# 等待端口真正释放 (最多 8 秒)
# 3) 快速轮询等待端口释放 (200ms 间隔,最多 3 秒)
local wait_count=0
while [ $wait_count -lt 8 ]; do
local max_wait=15 # 15 * 0.2 = 3 秒
while [ $wait_count -lt $max_wait ]; do
if command -v ss >/dev/null 2>&1; then
ss -tulnp 2>/dev/null | grep -q ":${port} " || break
else
netstat -tulnp 2>/dev/null | grep -q ":${port} " || break
fi
sleep 1
usleep 200000 2>/dev/null || sleep 0.2
wait_count=$((wait_count + 1))
done
# 如果端口仍被占用,强制杀掉占用者
if [ $wait_count -ge 8 ]; then
# 4) 如果端口仍被占用,强制 SIGKILL
if [ $wait_count -ge $max_wait ]; then
local port_pid=""
if command -v ss >/dev/null 2>&1; then
port_pid=$(ss -tulnp 2>/dev/null | grep ":${port} " | sed -n 's/.*pid=\([0-9]*\).*/\1/p' | head -1)
else
port_pid=$(netstat -tulnp 2>/dev/null | grep ":${port} " | sed -n 's|.* \([0-9]*\)/.*|\1|p' | head -1)
fi
[ -n "$port_pid" ] && kill -9 "$port_pid" 2>/dev/null
sleep 1
if [ -n "$port_pid" ]; then
kill -9 "$port_pid" 2>/dev/null
# 等待内核回收 (缩短到 0.5 秒)
usleep 500000 2>/dev/null || sleep 0.5
fi
fi
}
@@ -390,8 +406,8 @@ procd_add_reload_trigger "openclaw"
reload_service() {
stop
# stop_service 已确保端口释放,但额外等待 1 秒让内核回收
sleep 1
# v2026.3.14: stop_service 已优化等待逻辑,缩短额外等待时间
usleep 300000 2>/dev/null || sleep 0.3
start
}
@@ -407,47 +423,45 @@ restart_gateway() {
local port
port=$(uci -q get openclaw.main.port || echo "18789")
# ── 第一步: kill 监听端口的 gateway 子进程 (openclaw-gateway) ──
# openclaw 启动后会 fork 出 openclaw-gateway 子进程实际监听端口
# 必须先杀子进程释放端口,否则 procd respawn 的新实例会因端口冲突而崩溃
local port_pid=""
# v2026.3.14 优化: 快速终止并重启
# 1) 同时获取端口 PID 和 procd 管理的 PID
local port_pid="" gw_pid=""
if command -v ss >/dev/null 2>&1; then
port_pid=$(ss -tulnp 2>/dev/null | grep ":${port} " | sed -n 's/.*pid=\([0-9]*\).*/\1/p' | head -1)
else
port_pid=$(netstat -tulnp 2>/dev/null | grep ":${port} " | sed -n 's|.* \([0-9]*\)/.*|\1|p' | head -1)
fi
[ -n "$port_pid" ] && kill "$port_pid" 2>/dev/null
# ── 第二步: kill procd 管理的 gateway 主进程 (openclaw) ──
# procd 追踪的是主进程 PIDkill 它才能触发 respawn
local gw_pid=""
gw_pid=$(ubus call service list '{"name":"openclaw"}' 2>/dev/null | \
jsonfilter -e '$.openclaw.instances.gateway.pid' 2>/dev/null) || true
if [ -n "$gw_pid" ] && kill -0 "$gw_pid" 2>/dev/null; then
kill "$gw_pid" 2>/dev/null
fi
# 2) 同时发送 SIGTERM 给所有相关进程
[ -n "$port_pid" ] && kill "$port_pid" 2>/dev/null
[ -n "$gw_pid" ] && kill -0 "$gw_pid" 2>/dev/null && kill "$gw_pid" 2>/dev/null
# ── 第三步: 兜底 kill 所有 openclaw gateway 相关残留进程 ──
# 避免任何残留进程占据端口
sleep 1
# 3) 兜底: kill 所有残留进程
for pid in $(pgrep -f "openclaw-gateway" 2>/dev/null) $(pgrep -f "openclaw.*gateway.*run" 2>/dev/null); do
kill "$pid" 2>/dev/null
done
# ── 第四步: 等待端口真正释放 (最多 5 秒) ──
# 4) 快速轮询等待端口释放 (200ms 间隔,最多 2 秒)
local wait_count=0
while [ $wait_count -lt 5 ]; do
while [ $wait_count -lt 10 ]; do
if command -v ss >/dev/null 2>&1; then
ss -tulnp 2>/dev/null | grep -q ":${port} " || break
else
netstat -tulnp 2>/dev/null | grep -q ":${port} " || break
fi
sleep 1
usleep 200000 2>/dev/null || sleep 0.2
wait_count=$((wait_count + 1))
done
# ── 第五步: 如果 procd 中没有 gateway 服务注册 (首次/崩溃),调用 start ──
# 5) 如果端口仍被占用,强制 SIGKILL
if [ $wait_count -ge 10 ]; then
[ -n "$port_pid" ] && kill -9 "$port_pid" 2>/dev/null
usleep 300000 2>/dev/null || sleep 0.3
fi
# 6) 如果 procd 中没有 gateway 服务注册,调用 start
if [ -z "$gw_pid" ]; then
/etc/init.d/openclaw start >/dev/null 2>&1
fi