J6B泊车日志分析

J6B泊车系统日志读取和分析。支持：日志文件定位、日志读取、常见错误识别、日志下载、系统资源性能分析。使用场景：查看日志、分析报错、排查问题、性能诊断、下载日志到本地。触发词：日志、log、slog2info、tail、grep、报错、错误、异常、失败、故障、性能、CPU、占用、下载、scp。

Security Scan

Skill’s stated purpose (on‑device log reading, analysis, download and CPU monitoring) matches most of its instructions, but the included shell scripts contain multiple syntax/truncation bugs and the skill requires/assumes root SSH access to a board (192.168.1.10) — both of which merit caution before use.

Purpose & Capability

Name/description (J6B parking log analysis: locate/read/download logs, detect errors, performance analysis) match the content: docs, QNX command references, and three shell scripts for analysis/download/monitoring. The use of SSH/SCP to access a board at 192.168.1.10 as root is consistent with the stated goal of downloading logs from a device.

Instruction Scope

SKILL.md and the scripts instruct the agent/user to run many system commands (ssh, scp, tar, pidin, hogs, top, slog2info) and to read device files (/app/apa/log/*, /log/coredump, /log/reset_reason.txt, /userdata/...). That is coherent for on‑device troubleshooting, but it implicitly requires privileged access to the device and will collect/send device logs to the local machine. No instructions attempt to exfiltrate data to third‑party internet hosts.

Install Mechanism

No install spec / no external downloads. The skill is instruction + bundled shell scripts only, so nothing is pulled from arbitrary URLs. Risk from install mechanism is low.

Credentials

The skill does not declare required env vars, but all download/monitor workflows assume SSH access as root@192.168.1.10 (and provide that username/IP in docs/scripts). Requiring root SSH access to a board is proportionate to collecting logs/performance data, but it is high privilege: anyone who runs the scripts or follows the instructions must supply credentials/keys with full system access. The skill also assumes writable locations on the board (/userdata, /tmp). Users should not provide credentials to untrusted skills and should prefer ephemeral keys or limited accounts where possible.

Persistence & Privilege

Skill is not always-enabled and does not request persistent platform privileges. It does not attempt to modify other skills or system-wide agent settings. The scripts do create and delete files on the target board (/tmp and /userdata), which is expected for log collection but should be noted.

What to consider before installing

What to check before installing/using this skill: - Review and fix the scripts before running them. Multiple shell files (especially analyze_log.sh) contain obvious syntax/truncation issues and will likely fail or behave unpredictably if executed as-is. - The workflows require SSH/SCP access as root to IPs 192.168.1.10 / 192.168.2.10. That is expected for board maintenance but is high privilege — avoid giving permanent credentials to an untrusted skill. Use temporary SSH keys, limit source IPs, or a restricted account if possible. - Test all scripts in a safe, isolated environment (or on a non-production board) and run them with dry-run / non-destructive options first (e.g., inspect what files would be copied/tarred/removed). Pay attention to remote rm commands in archive-mode cleanup. - Because the skill will collect logs/coredumps from the device, confirm that you are allowed to transfer those files (sensitive data may be present). Verify where downloaded data is stored locally and who can read it. - If you plan to use the skill operationally, fix the script bugs, run a code review, and prefer SSH key auth (no passwords) and least-privilege accounts. If you are not comfortable auditing shell scripts, do not run the bundled scripts — instead use the SKILL.md commands manually after reviewing them.

J6B泊车日志分析Skill

快速导航

参考文档	说明
log-locations.md	各模块日志路径详细说明
error-codes.md	常见错误识别和解决建议
performance.md	QNX性能分析方法详解（含monitor_cpu.sh脚本）
qnx-commands.md	QNX命令参考
troubleshooting.md	故障排查案例库
report-template-system-resource.md	⭐ 系统资源分析报告模板（必读）

自动化脚本

脚本	说明	用法
monitor_cpu.sh	综合CPU监控（彩色+日志）	./monitor_cpu.sh
download_log.sh	批量下载日志（打包/逐模块）	./download_log.sh [-a] [-c IP] [模块名]
analyze_log.sh	日志错误分析（模式识别+报告）	./analyze_log.sh [-l] [-r] [模块名]

download_log.sh 使用

# 下载所有模块日志
./download_log.sh

# 下载单个模块
./download_log.sh planning

# 打包模式（大数据量推荐）
./download_log.sh -a

# 指定百兆网口
./download_log.sh -c 192.168.2.10 planning

# 指定保存目录
./download_log.sh -o /tmp/logs planning od rd

analyze_log.sh 使用

# 分析板端所有模块
./analyze_log.sh

# 分析指定模块
./analyze_log.sh planning od rd

# 分析本地已下载的日志
./analyze_log.sh -l planning

# 生成分析报告
./analyze_log.sh -r

---

⚠️ 报告输出规范

当用户要求分析系统资源时，必须严格按照 report-template-system-resource.md 模板输出。

报告结构（7章）

章节	内容
第1章 系统健康度	CPU/内存/进程数/线程数总览 + CPU统计(Min/Max/Avg/Idle)
第2章 CPU占用 TOP 15	排名、进程名、CPU%、SYS%、判定
第3章 内存占用 TOP 15	排名、进程名、内存、判定
第4章 线程状态异常	Mtx/Sem异常线程列表
第5章 泊车业务进程状态	13个业务进程的运行状态、PID、在线时长
第6章 TOP完整输出	top命令原始输出
第7章 重点关注项	异常汇总 + 可能原因 + 排查建议

判定阈值

指标	🟢 正常	🟡 注意	🔴 告警
单进程 CPU	< 10%	10% ~ 50%	> 50%
系统总 CPU	< 50%	50% ~ 80%	> 80%
单进程内存	< 200M	200M ~ 500M	> 500M
可用内存	> 2000M	500M ~ 2000M	< 500M
线程 Mtx/Sem	0 个	1 ~ 3 个	> 3 个

泊车业务进程清单（第5章必须覆盖）

sensorcenter, image_preprocess, perception_rd, perception_od, perception_fusion, dr, loc, pad, gridmap, planning, ui_control, adaptercom, nos_adasSimtool

---

常用操作速查

F1: 日志文件定位

主日志路径: /app/apa/log/ Coredump路径: /log/coredump

各模块日志快速查找：

模块	命令
planning	ls -la /app/apa/log/planning/
perception (od/rd)	ls -la /app/apa/log/{od,rd}/
loc	ls -la /app/apa/log/loc/
sensorcenter	ls -la /app/apa/log/sensorcenter/
所有模块	ls -la /app/apa/log/

F2: 基础日志读取

# 查看最新日志
tail -100 /app/apa/log/planning/*.log

# 实时跟踪
tail -f /app/apa/log/planning/*.log

# 搜索错误
grep -i "error" /app/apa/log/planning/*.log

# 查看系统日志
slog2info | grep -i planning

F3: 日志下载

网络信息：

• 千兆网口：192.168.1.10
• 百兆网口：192.168.2.10
• 用户名：root

# 单个模块日志
scp root@192.168.1.10:/app/apa/log/planning/*.log ./

# 批量下载所有日志
scp root@192.168.1.10:/app/apa/log/*.log ./

# 打包下载
ssh root@192.168.1.10 "tar czf /tmp/logs.tar.gz -C /app/apa/log ."
scp root@192.168.1.10:/tmp/logs.tar.gz ./

F4: 进程状态查看

# 查看进程详情（QNX特有）
pidin -p planning

# 查看CPU占用
hogs | grep -E "planning|od|rd|loc|psd|sensorcenter"

# 查看系统状态
top

F5: 系统资源性能分析

方式一：快速诊断（推荐新手）

# 综合监控脚本（彩色输出+自动日志）
/userdata/monitor_cpu.sh

特性：

• 🎨 彩色输出：高CPU红色、警告黄色
• 📊 TOP 15排名：CPU占用、内存占用
• 📝 自动日志：保存到 /userdata/log/cpu_monitor/
• 🔄 自动清理：7天后自动删除旧日志
• ⏱️ 实时刷新：每3秒更新一次

方式二：手动三步法（推荐熟练用户）

# 第一步：找出占用CPU的进程
hogs -i -s 1

# 第二步：查看该进程详情
top -p <PID>

# 第三步：分析线程状态
# - Run: 正常计算
# - Mtx/Sem: 锁竞争，需检查
# - NSlp: 正常休眠

诊断结论参考

PIDS	SYS	STATE	可能原因
>50%	<30%	多Run	算法过重，需优化
>50%	>70%	多Run	死循环或系统调用频繁
正常	正常	多Mtx/Sem	锁竞争严重，检查代码
正常	正常	多NSlp	正常休眠

后台监控脚本使用

# 上传脚本到板端
scp scripts/monitor_cpu.sh root@192.168.1.10:/userdata/

# SSH登录板端
ssh root@192.168.1.10

# 添加执行权限
chmod +x /userdata/monitor_cpu.sh

# 前台运行（实时查看）
/userdata/monitor_cpu.sh

# 后台运行（适合长期监控）
nohup /userdata/monitor_cpu.sh > /dev/null 2>&1 &

# 查看监控日志
tail -f /userdata/log/cpu_monitor/cpu_monitor_*.log

# 停止后台进程
kill $(pidof monitor_cpu.sh)

---

工作流程

场景1：泊车失败，需要查看日志

1. 定位日志路径

   ls -la /app/apa/log/
   

2. 查看最新错误

   grep -i "error\|fail" /app/apa/log/planning/*.log | tail -50
   

3. 实时跟踪

   tail -f /app/apa/log/planning/*.log
   

4. 如需下载分析

   scp root@192.168.1.10:/app/apa/log/planning/*.log ./
   

场景2：系统卡顿，需要分析性能

1. 快速启动综合监控

   /userdata/monitor_cpu.sh
   

2. 观察监控输出

   • CPU占用 > 80% 红色警告
   • 内存占用 > 10000K 黄色提示
   • 查看TOP 15排名定位问题进程

3. 深入分析

   # 查看目标进程详情
   top -p <PID>
   

4. 分析问题

   • PIDS高 + SYS高 → 可能死循环或系统调用频繁
   • 多线程卡在Mtx/Sem → 存在锁竞争

5. 查看历史监控日志

   tail -100 /userdata/log/cpu_monitor/cpu_monitor_*.log
   

场景3：进程崩溃，需要分析coredump

1. 查找coredump文件

   ls -la /log/coredump/
   

2. 查看崩溃原因

   cat /log/reset_reason.txt
   

3. 结合日志分析

   slog2info | grep -i "crash\|segmentation"
   

---

网络连接

连接到板端：

# 千兆网口
ssh root@192.168.1.10

# 百兆网口
ssh root@192.168.2.10

测试连接：

ping 192.168.1.10

---

QNX vs Linux 常用命令对比

功能	QNX	Linux
查看进程详情	pidin	ps
结束进程	slay	kill
系统日志	slog2info	journalctl
资源监控	hogs, top	top, htop