release: v1.0.7 — 依赖包名修正、Gateway crash loop 修复

This commit is contained in:
10000ge10000
2026-03-06 00:33:17 +08:00
parent 20a8c016a4
commit e844b4754b
9 changed files with 130 additions and 9 deletions

View File

@@ -190,6 +190,37 @@ all) gw_bind="custom" ;; # custom = 0.0.0.0
*) gw_bind="$bind" ;;
esac
# 确保网关端口未被残留进程占用 (防止 restart 时 crash loop)
_ensure_port_free() {
local p="$1" max_wait="${2:-5}" i=0
while [ $i -lt $max_wait ]; do
if command -v ss >/dev/null 2>&1; then
ss -tlnp 2>/dev/null | grep -q ":${p} " || return 0
else
netstat -tlnp 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
i=$((i + 1))
done
# 最后手段: SIGKILL
local port_pid=""
if command -v ss >/dev/null 2>&1; then
port_pid=$(ss -tlnp 2>/dev/null | grep ":${p} " | sed -n 's/.*pid=\([0-9]*\).*/\1/p' | head -1)
else
port_pid=$(netstat -tlnp 2>/dev/null | grep ":${p} " | sed -n 's|.* \([0-9]*\)/.*|\1|p' | head -1)
fi
[ -n "$port_pid" ] && kill -9 "$port_pid" 2>/dev/null && sleep 1
return 0
}
_ensure_port_free "$port"
# 启动 OpenClaw Gateway (主服务, 前台运行)
procd_open_instance "gateway"
procd_set_param command "$NODE_BIN" "$oc_entry" gateway run \
@@ -205,7 +236,7 @@ OC_GLOBAL="$OC_GLOBAL" \
OC_DATA="$OC_DATA" \
PATH="${NODE_BASE}/bin:${OC_GLOBAL}/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin"
procd_set_param user openclaw
procd_set_param respawn 3600 5 5
procd_set_param respawn 3600 10 5
procd_set_param stdout 1
procd_set_param stderr 1
procd_set_param pidfile /var/run/openclaw.pid
@@ -245,8 +276,44 @@ procd_close_instance
}
stop_service() {
# procd 会自动处理进程停止
return 0
# procd 会自动停止它管理的主进程 (openclaw)
# 但 openclaw 会 fork 出 openclaw-gateway 子进程实际监听端口
# procd 不一定能正确追踪并杀掉子进程树,需要手动清理
local port
port=$(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
done
# 等待端口真正释放 (最多 8 秒)
local wait_count=0
while [ $wait_count -lt 8 ]; do
if command -v ss >/dev/null 2>&1; then
ss -tlnp 2>/dev/null | grep -q ":${port} " || break
else
netstat -tlnp 2>/dev/null | grep -q ":${port} " || break
fi
sleep 1
wait_count=$((wait_count + 1))
done
# 如果端口仍被占用,强制杀掉占用者
if [ $wait_count -ge 8 ]; then
local port_pid=""
if command -v ss >/dev/null 2>&1; then
port_pid=$(ss -tlnp 2>/dev/null | grep ":${port} " | sed -n 's/.*pid=\([0-9]*\).*/\1/p' | head -1)
else
port_pid=$(netstat -tlnp 2>/dev/null | grep ":${port} " | sed -n 's|.* \([0-9]*\)/.*|\1|p' | head -1)
fi
[ -n "$port_pid" ] && kill -9 "$port_pid" 2>/dev/null
sleep 1
fi
}
service_triggers() {
@@ -255,6 +322,8 @@ 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
start
}
@@ -384,6 +453,8 @@ local mins=$(( (uptime % 3600) / 60 ))
echo "运行时间: ${hours}小时 ${mins}分钟"
fi
fi
elif pgrep -f "openclaw.*gateway" >/dev/null 2>&1; then
echo "网关: 正在启动 (首次启动可能需要 2~5 分钟)"
else
echo "网关: 已停止"
fi