The 9router v0.4.41-46 was auto killed after startup, but not OOM-killer

[traderlife8]

========================================
Choose Interface (v0.4.41)
 Server: http://localhost:20128

★ Web UI (Open in Browser)
☆ Terminal UI (Interactive CLI)
☆ Hide to Tray (Background)
☆ Exit

Press Enter to go back to menu...已杀死

[traderlife8]

~$ 9router
[9router][runtime] npm install better-sqlite3@12.6.2 --no-audit --no-fund --prefer-online --no-save (cwd: /home/zh/.9router/runtime)
已杀死

[indrapalijama]

revert to v0.4.39

[zvwgvx]

i have the same problem #1116

[decolua]

Please try the latest version:

npm i -g 9router@latest --prefer-online

已杀死 = OOM-killer. If still happens on 0.4.45, share free -h and dmesg | tail.

[zvwgvx]

Confirmed, the issue is fixed in 0.4.45. Thanks!

[traderlife8]

v0.4.45 still has this problem. my system is Ubuntu 24.04.4 LTS, how to share free -h and dmesg | tail ?

[nautonk]

v0.4.45 still has this problem. my system is Ubuntu 24.04.4 LTS, how to share free -h and dmesg | tail ?

Me too. Works good on v0.4.39 so revert back to v0.4.39.

free -h output:

ubuntu@instance-20220624-0828:/root$ free -h
total used free shared buff/cache available
Mem: 23Gi 1.8Gi 17Gi 218Mi 4.2Gi 21Gi
Swap: 999Mi 0B 999Mi

dmesg | tail output:

ubuntu@instance-20220624-0828:/root$ sudo dmesg | tail
[1106625.625950] docker0: port 1(veth67f3444) entered disabled state
[1106625.627350] device veth67f3444 left promiscuous mode
[1106625.627358] docker0: port 1(veth67f3444) entered disabled state
[1106685.619415] docker0: port 1(vetha10fe0e) entered blocking state
[1106685.619422] docker0: port 1(vetha10fe0e) entered disabled state
[1106685.619520] device vetha10fe0e entered promiscuous mode
[1106685.627955] eth0: renamed from vethde96cbe
[1106685.646489] IPv6: ADDRCONF(NETDEV_CHANGE): vetha10fe0e: link becomes ready
[1106685.646527] docker0: port 1(vetha10fe0e) entered blocking state
[1106685.646530] docker0: port 1(vetha10fe0e) entered forwarding state

uname -a output:

ubuntu@instance-20220624-0828:/root$ uname -a
Linux instance-20220624-0828 5.15.0-1081-oracle #87~20.04.1-Ubuntu SMP Mon Apr 28 12:53:11 UTC 2025 aarch64 aarch64 aarch64 GNU/Linux

Thank you

[traderlife8]

9router.zip
my hermes use LLM=MiniMax-M2.5 fixed the problem! see the 9router.zip

9router Exit-on-Startup Problem Investigation Report

Date: 2026-05-15
Problem: 9router service exits immediately after startup, systemd shows exit status=0/SUCCESS
Environment: Linux zh-i7-7700-105-U (Ubuntu), Node.js, systemd

---

1. Symptoms

systemctl status 9router
● 9router.service - 9Router - AI Router & Token Saver
     Active: activating (auto-restart) since Fri 2026-05-15 16:00:44 CST; 8s ago
   Process: 13505 ExecStart=/home/zh/.npm-global/bin/9router (code=exited, status=0/SUCCESS)

Every startup cycle exits immediately and restarts, with logs showing:

[9router][runtime] npm install better-sqlite3@12.6.2 ...
Exiting...
Exiting...

---

2. Investigation Process

Step 1: Check Service Logs

journalctl -u 9router -n 30

Found that each cycle runs npm install better-sqlite3 then immediately exits.

Step 2: Run 9router Manually

Direct foreground run also exits immediately, exit code = -9 (SIGKILL).

Step 3: strace Signal Tracing

strace -f -e trace=signal -o /tmp/9router_strace.log /home/zh/.npm-global/bin/9router

Found the key line:

kill(14854, SIGKILL)   # Some process killed the main process

Further tracing of strace log revealed two child processes executing:

execve("/bin/sh", ["/bin/sh", "-c", "kill -9 16841 2>/dev/null"], ...)
execve("/bin/sh", ["/bin/sh", "-c", "kill -9 16852 2>/dev/null"], ...)

Step 4: Root Cause Identified — ps aux Self-Kill Bug

The killAllAppProcesses() function uses ps aux | grep 9router to find old processes, but:

• The grep 9router command line itself contains the "9router" string
• The strace -f /home/zh/.npm-global/bin/9router command line contains "9router"

Both get matched and killed with kill -9, which also kills the currently running 9router main process.

Step 5: Adding --skip-update Flag

After fixing the ps aux filter logic, the process was still being killed. Further analysis revealed:

checkForUpdate() makes an HTTPS request to registry.npmjs.org to check the latest version.

• No foreign network access in China → 3 second timeout wait
• Safety timeout → another 8 seconds wait
• systemd keeps restarting the service (Restart=always)
• Signal conflicts during restart cause the process to get SIGKILL

After adding the --skip-update flag to skip the version check, 9router starts successfully.

---

3. Fixes

Fix 1: Improved ps aux Filter Logic (cli.js lines 241-254)

File: /home/zh/.npm-global/lib/node_modules/9router/cli.js

Before:

lines.forEach(line => {
  const isAppProcess = line.includes("9router") || line.includes("next-server");
  if (isAppProcess) {
    const parts = line.trim().split(/\s+/);
    const pid = parts[1];
    if (pid && !isNaN(pid) && pid !== process.pid.toString()) {
      pids.push(pid);
    }
  }
});

After:

lines.forEach(line => {
  // Exclude: grep itself (cmdline = "grep <pattern>"),
  // strace wrappers (cmdline contains strace ... 9router),
  // any strace arguments (-- separator)
  const isExcluded =
    line.includes('grep ') || line.includes('grep\t') ||
    line.includes('strace') || line.includes(' -- ');
  const isAppProcess = !isExcluded && (line.includes("9router") || line.includes("next-server"));
  if (isAppProcess) {
    const parts = line.trim().split(/\s+/);
    const pid = parts[1];
    if (pid && !isNaN(pid) && pid !== process.pid?.toString()) {
      pids.push(pid);
    }
  }
});

Rationale: Exclude lines containing "grep ", "grep\t", "strace", or " -- " to prevent accidentally matching the grep/strace processes themselves.

Fix 2: Add --skip-update Flag to systemd Service

File: /etc/systemd/system/9router.service

[Service]
ExecStart=/home/zh/.npm-global/bin/9router --skip-update

To apply:

sudo systemctl daemon-reload
sudo systemctl restart 9router

---

4. Verification

$ systemctl status 9router
● 9router.service - 9Router - AI Router & Token Saver
     Active: active (running) since Fri 2026-05-15 16:11:47 CST; ...

$ ss -tlnp | grep 20128
LISTEN 0  511  0.0.0.0:20128  0.0.0.0:*  users:(("next-server (v1",pid=18280,fd=21))

$ curl -s http://localhost:20128/
/dashboard

9router is running normally, Web UI is accessible.

---

5. File清单

File	Description
/home/zh/.npm-global/lib/node_modules/9router/cli.js	Fixed main program
/home/zh/.npm-global/lib/node_modules/9router/cli.js.bak	Original backup
/etc/systemd/system/9router.service	systemd service configuration
/home/zh/.npm-global/bin/9router	Symlink → ../lib/node_modules/9router/cli.js
/home/zh/.9router/	User data directory (config, DB, logs)

[traderlife8]

cli.js
Use this fixed cli.js , and bash run: '9router --skip-update' is OK.
Another issue that cause the 9router process to be killed maybe in the 'update' process.
Both of these issues need to be fixed in order to completely fix the problem of the '9router' process being killed after startup.

[indrapalijama]

cli.js this is fixed cli.js

hi @decolua , please check this, latest version still contains same error (auto killed)

[phatjkk]

i got the same issue in latest version on debian 12