\r \r

Linux System Health Diagnostic Skill\r

\r You are a Linux OS diagnostic expert. When a user reports any of the following problems, use this skill:\r

• Performance: server slow, high load, lag, unresponsive\r
• Memory: OOM killed, out of memory, memory leak, swap thrashing\r
• Disk: disk full, read-only filesystem, inode exhaustion, log files too large\r
• CPU: high CPU, IO wait, process stuck, load average spike\r
• Network: DNS failure, connection timeout, port exhaustion, CLOSE_WAIT accumulation, firewall blocking\r
• Process: crash, zombie processes, too many open files, file descriptor limit\r
• Browser automation: missing shared libraries, Chromium sandbox error, headless browser failures\r
• Locale/Encoding: garbled text, character encoding issues, locale not configured\r \r Use the judgment rules below to systematically diagnose OS-level root causes.\r \r When NOT to use this skill: For application-level issues specific to OpenClaw (gateway config, API keys, model configuration, service management, systemd units), use the openclaw-diagnostic skill instead. This skill only covers OS-level diagnostics.\r \r Diagnostic workflow:\r

1. Always start with Section 1 (System Environment Baseline) to establish context\r
2. Then run the sections relevant to the user's reported symptoms\r
3. If the root cause is unclear, run all sections in order for a comprehensive check\r \r

Commands: Run the corresponding section in scripts/diagnostics.sh. Run as root with export LANG=C.\r \r Issue Registry: See reference.md for severity level definitions and the complete issue name table.\r \r Data access scope — this skill collects OS-level diagnostic data. Review before running in sensitive environments:\r \r | Category | What is accessed | Sections |\r |----------|-----------------|----------|\r | System config files | /etc/os-release, /etc/resolv.conf, /etc/security/limits.conf, /etc/default/locale, /etc/locale.conf, /etc/systemd/journald.conf | 1, 6, 8, 11, 17 |\r | Kernel interfaces | /proc/meminfo, /proc/stat, /proc/loadavg, /proc/sys/fs/*, /proc/sys/net/*, /sys/kernel/mm/* | 2, 3, 5, 6, 7, 14 |\r | Kernel ring buffer | dmesg — may contain process names and OOM kill details | 2, 7, 12 |\r | Systemd journal | journalctl -k — kernel messages only | 2 |\r | Log directory | /var/log/ size enumeration only (does not read log content) | 11 |\r | Process & socket table | ps, ss -p — exposes PIDs, command names, socket owners | 2, 3, 10, 15 |\r | User home directories | /root/.cache/ms-playwright, /home/*/.cache/ms-playwright — Chromium binary search only | 16 |\r | Outbound network probes | DNS resolution tests (nslookup/dig/getent to github.com), nameserver TCP/53 reachability, Chrome headless launch test (about:blank) | 8, 16 |\r | Write operation | Creates and immediately removes /tmp/.oc_write_test to verify filesystem writability — the only write in the entire script | 12 |\r \r Output format: After running diagnostics, report findings as a severity-sorted list (FATAL > CRITICAL > ERROR > WARNING > INFO). For each issue found, include:\r

• Issue name (e.g., OpenClaw.Memory.SystemMemoryCritical)\r
• Severity level\r
• Observed value vs threshold\r
• Recommended remediation\r \r ---\r \r

1. System Environment Baseline\r

\r Collect OS context for subsequent analysis.\r \r Judgment rules:\r

• Record output as OpenClaw.System.EnvironmentBaseline (INFO) — no issues, context only.\r \r ---\r \r

2. Memory & OOM\r

\r Detect low memory and past OOM kills that affect any workload on this server.\r \r Judgment rules:\r

• MemAvailable / MemTotal \x3C 5% → OpenClaw.Memory.SystemMemoryCritical (CRITICAL)\r
  • Remediation: Kill unnecessary processes, add swap, or increase instance RAM\r
• MemAvailable / MemTotal \x3C 10% → OpenClaw.Memory.SystemMemoryLow (WARNING)\r
  • Remediation: Monitor closely; consider scaling up\r
• MemTotal \x3C 2 GB → OpenClaw.Memory.InsufficientTotalMemory (ERROR)\r
  • Remediation: 4 GB+ RAM recommended for production workloads\r
• dmesg contains "oom-killer" → OpenClaw.Memory.OOMKillerEvent (WARNING)\r
  • Remediation: Identify which processes were killed; review memory allocation\r \r ---\r \r

3. CPU & Performance\r

\r Resource contention causes slow responses; high iowait indicates disk bottlenecks.\r \r Judgment rules:\r

• Load average (1 min) > 2x nproc → OpenClaw.CPU.SystemLoadHigh (WARNING)\r
  • Remediation: Identify top CPU consumers; check for runaway processes\r
• CPU idle \x3C 10% (i.e., total utilization > 90%) → OpenClaw.CPU.SystemCPUExhausted (CRITICAL)\r
  • Remediation: Identify top process; check for log flooding or computation storms\r
• iowait > 30% (from /proc/stat) → OpenClaw.CPU.HighIOWait (WARNING)\r
  • Remediation: Check disk I/O — likely excessive logging or disk-bound workload\r \r ---\r \r

4. Network Infrastructure\r

\r Basic network configuration, DNS, IPv6, and firewall state.\r \r Judgment rules:\r

• IPv6 enabled and services bind :: but upstream resolves to IPv4 only → OpenClaw.Network.IPv6Mismatch (WARNING)\r
  • Remediation: Set NODE_OPTIONS='--dns-result-order=ipv4first' or sysctl -w net.ipv6.conf.all.disable_ipv6=1\r \r ---\r \r

5. Disk & inotify\r

\r Disk space exhaustion and inotify limits cause "ENOSPC" errors.\r \r Judgment rules:\r

• Any filesystem usage >= 95% → OpenClaw.Disk.FilesystemFull (CRITICAL)\r
  • Remediation: Clean old logs and data; extend partition or add disk\r
• Any filesystem usage >= 80% → OpenClaw.Disk.FilesystemHighUsage (WARNING)\r
  • Remediation: Monitor; plan cleanup or expansion\r
• max_user_watches \x3C 65536 → OpenClaw.Disk.InotifyWatchesTooLow (ERROR)\r
  • Remediation: echo 'fs.inotify.max_user_watches=524288' >> /etc/sysctl.d/99-inotify.conf && sysctl -p /etc/sysctl.d/99-inotify.conf\r
• max_user_instances \x3C 256 → OpenClaw.Disk.InotifyInstancesTooLow (WARNING)\r
  • Remediation: echo 'fs.inotify.max_user_instances=512' >> /etc/sysctl.d/99-inotify.conf && sysctl -p /etc/sysctl.d/99-inotify.conf\r \r ---\r \r

6. File Descriptor & Process Limits\r

\r Low ulimits cause "too many open files" (EMFILE) errors under load.\r \r Judgment rules:\r

• Shell ulimit -n \x3C 4096 → OpenClaw.Limits.NofileTooLow (ERROR)\r
  • Remediation: Add * soft nofile 65536 and * hard nofile 65536 to /etc/security/limits.conf; re-login\r
• limits.conf nofile value > fs.nr_open → OpenClaw.Limits.NofileExceedsKernelMax (CRITICAL)\r
  • Remediation: Increase fs.nr_open first: sysctl -w fs.nr_open=1048576 and persist in /etc/sysctl.d/\r
• file-nr allocated / max > 80% → OpenClaw.Limits.SystemFileDescriptorsHigh (WARNING)\r
  • Remediation: Identify processes holding many FDs (ls /proc/*/fd 2>/dev/null | wc -l); increase fs.file-max if needed\r \r ---\r \r

7. Kernel & Sysctl Tuning\r

\r nf_conntrack, TCP tuning, and somaxconn affect high-concurrency workloads.\r \r Judgment rules:\r

• nf_conntrack_max \x3C 65536 → OpenClaw.Kernel.NfConntrackMaxTooLow (ERROR)\r
  • Remediation: sysctl -w net.netfilter.nf_conntrack_max=262144 and persist in /etc/sysctl.d/99-sysctl.conf\r
• dmesg contains "nf_conntrack: table full" → OpenClaw.Kernel.NfConntrackTableFull (CRITICAL)\r
  • Remediation: Increase nf_conntrack_max; check for connection leaks\r
• somaxconn \x3C 1024 → OpenClaw.Kernel.SomaxconnTooLow (WARNING)\r
  • Remediation: sysctl -w net.core.somaxconn=4096 and persist\r
• tcp_max_tw_buckets \x3C 10000 → OpenClaw.Kernel.TcpMaxTwBucketsTooLow (WARNING)\r
  • Remediation: sysctl -w net.ipv4.tcp_max_tw_buckets=262144\r
• tcp_tw_reuse = 0 → OpenClaw.Kernel.TcpTwReuseNotEnabled (WARNING)\r
  • Remediation: sysctl -w net.ipv4.tcp_tw_reuse=1\r
• TIME_WAIT count from ss -s > 10000 → OpenClaw.Kernel.TimeWaitOverflow (WARNING)\r
  • Remediation: Enable tcp_tw_reuse, increase tcp_max_tw_buckets, reduce tcp_fin_timeout\r
• ListenOverflows > 0 in /proc/net/netstat → OpenClaw.Kernel.TcpListenOverflows (WARNING)\r
  • Remediation: Increase somaxconn and application backlog setting\r
• vm.overcommit_memory = 2 and swap \x3C 1 GB → OpenClaw.Kernel.StrictOvercommitWithLowSwap (WARNING)\r
  • Remediation: Add swap space or set vm.overcommit_memory=0\r \r ---\r \r

8. DNS Resolution Health\r

\r Broken or slow DNS causes EAI_AGAIN errors, API timeouts, and silent connectivity failures.\r \r Judgment rules:\r

• /etc/resolv.conf is empty or has zero nameserver lines → OpenClaw.Network.NoDNSNameservers (ERROR)\r
  • Remediation: Add nameservers — e.g., echo 'nameserver 8.8.8.8' >> /etc/resolv.conf; for systemd-resolved check /etc/systemd/resolved.conf\r
• nslookup, dig, and getent all fail for a known-good domain → OpenClaw.Network.DNSResolutionFailed (CRITICAL)\r
  • Remediation: Verify network connectivity; check if nameservers are reachable; inspect firewall rules blocking UDP/TCP port 53\r
• Any configured nameserver fails TCP/53 reachability test → OpenClaw.Network.DNSNameserverUnreachable (WARNING)\r
  • Remediation: Replace unreachable nameserver in /etc/resolv.conf; consider adding a backup nameserver\r \r ---\r \r

9. Time Synchronization\r

\r Clock drift causes SSL/TLS certificate validation failures, API auth token rejection, and log timestamp inconsistencies.\r \r Judgment rules:\r

• None of chronyd, ntpd, or systemd-timesyncd is active → OpenClaw.Time.NTPServiceNotRunning (ERROR)\r
  • Remediation: Install and enable a time sync service — yum install chrony && systemctl enable --now chronyd (RHEL/CentOS) or apt install chrony && systemctl enable --now chronyd (Debian/Ubuntu)\r
• timedatectl shows "NTP synchronized: no" → OpenClaw.Time.ClockNotSynchronized (CRITICAL)\r
  • Remediation: Start NTP service; verify NTP server reachability (chronyc sources or ntpq -p); check firewall allows UDP port 123\r
• chronyc tracking shows system clock offset > 3 seconds, or hwclock drift > 5 seconds from system time → OpenClaw.Time.ClockDriftDetected (WARNING)\r
  • Remediation: Force sync — chronyc makestep or ntpdate -u pool.ntp.org; investigate why drift occurred (suspended VM, unreachable NTP server)\r \r ---\r \r

10. Zombie & D-State Processes\r

\r Zombie processes indicate child process leaks; D-state (uninterruptible sleep) processes signal I/O hangs that block system operations.\r \r Judgment rules:\r

• Zombie count > 10 → OpenClaw.Process.ZombieProcessesHigh (WARNING)\r
  • Remediation: Identify parent processes (ps -eo pid,ppid,stat,comm | awk '$3~/Z/'); the parent is not reaping children — restart or fix the parent process\r
• D-state process count > 0 → OpenClaw.Process.DStateProcessesFound (CRITICAL)\r
  • Remediation: D-state processes are blocked on I/O — check disk health (dmesg | grep -i error), NFS mounts (mount -t nfs), and storage subsystem; these processes cannot be killed normally\r
• Total process count > 80% of kernel.pid_max → OpenClaw.Process.TotalProcessCountHigh (WARNING)\r
  • Remediation: Identify process-spawning storms (ps -eo user --sort=user | uniq -c | sort -rn | head); increase kernel.pid_max if needed\r \r ---\r \r

11. Systemd Journal & Log Disk Usage\r

\r Systemd journal grows unbounded on long-running servers, silently consuming disk space — a common hidden root cause of "disk full" events.\r \r Judgment rules:\r

• Journal disk usage > 2 GB → OpenClaw.Logs.JournalDiskUsageHigh (WARNING)\r
  • Remediation: journalctl --vacuum-size=500M; set SystemMaxUse=500M in /etc/systemd/journald.conf and restart systemd-journald\r
• /var/log total size > 5 GB → OpenClaw.Logs.VarLogOversized (WARNING)\r
  • Remediation: Identify large files (find /var/log -type f -size +100M); configure logrotate; clean old rotated logs\r \r ---\r \r

12. Filesystem Integrity\r

\r Read-only filesystem (from ext4/xfs journal errors) prevents writing session data, logs, and PID files. Inode exhaustion produces "No space left on device" even with free disk space.\r \r Judgment rules:\r

• Any non-virtual mount has ro flag, or /tmp write test fails → OpenClaw.Disk.ReadOnlyFilesystem (CRITICAL)\r
  • Remediation: Check dmesg for filesystem errors; run fsck on the affected partition (requires unmount or single-user mode); may indicate disk hardware failure\r
• Any real filesystem inode usage >= 80% → OpenClaw.Disk.InodeUsageHigh (WARNING)\r
  • Remediation: Find directories with many small files (find / -xdev -printf '%h\ ' | sort | uniq -c | sort -rn | head -10); clean up session/temp files\r
• dmesg contains EXT4-fs error, XFS error, or read-only remount messages → OpenClaw.Disk.FilesystemErrorsDetected (CRITICAL)\r
  • Remediation: Back up data immediately; run fsck at next maintenance window; check disk SMART status (smartctl -a /dev/sdX)\r \r ---\r \r

13. Firewall & Outbound Connectivity\r

\r Firewall rules blocking inbound or outbound traffic are the #1 cause of "port not reachable" and "API connection refused" in self-hosted deployments.\r \r Judgment rules:\r

• DROP or REJECT rules detected on INPUT or OUTPUT chains → OpenClaw.Network.FirewallDropRulesDetected (WARNING)\r
  • Remediation: Review rules — ensure required ports (gateway port, 443 outbound) are allowed; use iptables -L -n -v for detailed hit counts\r \x3C!-- - Outbound TCP connect to external port 443 fails → OpenClaw.Network.OutboundHTTPSBlocked (ERROR)\r
  • Remediation: Check OUTPUT chain rules; verify cloud security group allows outbound 443; check if proxy is required (HTTP_PROXY/HTTPS_PROXY) -->\r
• ufw status shows default deny incoming (informational only) → OpenClaw.Network.UFWDefaultDeny (INFO)\r
  • Remediation: No action required if intentional; ensure gateway port is explicitly allowed (ufw allow \x3Cport>/tcp)\r \r ---\r \r

14. Transparent Hugepages\r

\r THP causes latency spikes and memory fragmentation for Node.js workloads. Multiple database and runtime vendors recommend disabling it on servers.\r \r Judgment rules:\r

• THP enabled is set to [always] → OpenClaw.Kernel.THPEnabled (WARNING)\r
  • Remediation: echo never > /sys/kernel/mm/transparent_hugepage/enabled; persist via systemd unit or /etc/rc.local\r
• THP defrag is set to [always] → OpenClaw.Kernel.THPDefragEnabled (INFO)\r
  • Remediation: echo never > /sys/kernel/mm/transparent_hugepage/defrag; reduces latency spikes from compaction\r \r ---\r \r

15. TCP Connection Overload\r

\r Excessive network connections exhaust file descriptors, memory, and conntrack table capacity, degrading system-wide performance.\r \r Judgment rules:\r

• Total TCP connections > 10000 → OpenClaw.Network.TcpConnectionCountHigh (WARNING)\r
  • Remediation: Identify top connection-holding processes; check for connection leaks; consider connection pooling\r
• CLOSE_WAIT count > 500 → OpenClaw.Network.CloseWaitAccumulation (ERROR)\r
  • Remediation: CLOSE_WAIT indicates the local application is not calling close() on sockets — identify the leaking process and restart it; this is an application bug\r
• ESTABLISHED count > 5000 → OpenClaw.Network.EstablishedConnectionsHigh (WARNING)\r
  • Remediation: Review whether all connections are legitimate; check for connection pool exhaustion or slow clients holding connections open\r
• Ephemeral ports in use > 80% of available range → OpenClaw.Network.EphemeralPortExhaustion (CRITICAL)\r
  • Remediation: Widen range sysctl -w net.ipv4.ip_local_port_range='1024 65535'; enable tcp_tw_reuse; check for connection leaks\r \r ---\r \r

16. Headless Browser / Chromium Dependencies\r

\r OpenClaw skills that use browser automation (Playwright, Puppeteer) require Chromium shared libraries and headless mode. The diagnostic first tests whether Chrome can actually launch in headless mode. Dependency diagnosis is only performed when Chrome fails or is absent.\r \r Judgment rules:\r

• Chrome headless launch test (--headless=new --dump-dom about:blank) succeeds → no issue, skip dependency checks\r
• Chrome headless launch test fails → proceed with dependency diagnosis below:\r
  • Any of the 7 critical shared library stems (libnss3, libatk-bridge-2.0, libgbm, libxkbcommon, libdrm, libgtk-3, libasound) is absent from ldconfig -p → OpenClaw.Browser.ChromiumDependenciesMissing (ERROR)\r
    • Remediation: On Debian/Ubuntu: apt install -y libnss3 libatk-bridge2.0-0 libgbm1 libxkbcommon0 libdrm2 libgtk-3-0 libasound2; on RHEL/CentOS: yum install -y nss atk at-spi2-atk mesa-libgbm libxkbcommon libdrm gtk3 alsa-lib\r
  • ldd on chromium binary shows one or more "not found" entries → OpenClaw.Browser.ChromiumBinaryLddFailures (CRITICAL)\r
    • Remediation: Install the specific missing libraries identified by ldd; run ldconfig after installation to update the dynamic linker cache\r
  • /proc/sys/kernel/unprivileged_userns_clone is 0 → OpenClaw.Browser.UserNamespaceDisabled (ERROR)\r
    • Remediation: sysctl -w kernel.unprivileged_userns_clone=1 and persist in /etc/sysctl.d/99-userns.conf; or configure Chromium with --no-sandbox (less secure, not recommended for production)\r \r ---\r \r

17. Locale & Encoding Configuration\r

\r Missing or misconfigured locale causes garbled text output, incorrect sorting in logs, and subtle bugs like backspace deleting two characters over SSH (when client sends UTF-8 but server expects ASCII). OpenClaw's text processing relies on correct UTF-8 support.\r \r Judgment rules (use the persistent LANG value read from /etc/default/locale or /etc/locale.conf, not the runtime $LANG which may be overridden to C by the diagnostic runner):\r

• Persistent LANG is empty, unset, or set to POSIX/C → OpenClaw.Locale.LocaleNotConfigured (ERROR)\r
  • Remediation: On Debian/Ubuntu: apt install locales && dpkg-reconfigure locales, then set LANG=en_US.UTF-8 in /etc/default/locale; on RHEL/CentOS: localectl set-locale LANG=en_US.UTF-8\r
• The persistent LANG value does not appear in locale -a output (configured but not generated/installed) → OpenClaw.Locale.LocaleNotGenerated (WARNING)\r
  • Remediation: On Debian/Ubuntu: uncomment the locale in /etc/locale.gen and run locale-gen; on RHEL/CentOS: localedef -i en_US -f UTF-8 en_US.UTF-8\r
• Persistent LANG does not contain UTF-8 or utf8 → OpenClaw.Locale.NonUTF8LocaleDetected (WARNING)\r
  • Remediation: Change to a UTF-8 variant: localectl set-locale LANG=en_US.UTF-8; re-login for the change to take effect\r