The Language Server That Froze Everything
Date: 2026-01-26 Issue: System freeze during coding session Culprit: Runaway language server process Collateral Damage: NAS mounts broken, Tailscale routes missing
The Freeze
Mid-session. Working on some code. System locks up completely. Not even the mouse moves.
Hard power cycle. Boot back up. Time to figure out what happened and what broke.
Finding The Culprit
After the reboot, I started checking what was still running from before the crash. One process caught my attention:
PID 6035: /opt/google/antigravity/resources/app/extensions/antigravity/bin/language_server_linux_x64
- CPU: 41.6%
- Memory: 10.3% (3.3GB)
- Workspace: ~/Development
41% CPU. 3.3 gigabytes of RAM. While doing nothing. The IDE wasn't even focused.
Checked its network connections:
| FD | Remote IP | Owner |
|---|---|---|
| 15 | 34.54.84.110:443 | Google Cloud |
| 68 | 142.251.46.138:443 | |
| 70 | 216.239.36.223:443 |
Three active HTTPS connections to Google's cloud endpoints. Constantly sending data. Constantly burning CPU. This is what caused the freeze - the language server hit some kind of runaway loop.
Immediate fix: pkill -f language_server_linux_x64
Long-term fix: Disable the extension when not actively using AI features, or find a less hungry alternative.
Collateral Damage #1: Tailscale Routes
After the hard shutdown, Tailscale was running but the route to the Andromeda network was missing:
ip route show | grep 192.168.50
# Nothing
The 192.168.20.0/24 subnet (where Meridian-Mako-Silo and the other Andromeda systems live) wasn't reachable.
Fix:
sudo ip route add 192.168.20.0/24 dev tailscale0
Permanent fix: Updated /usr/local/bin/fix-tailscale-routes.sh to add this route on startup.
Tailscale was also complaining about DNS:
Tailscale health check warning: "can't reach configured DNS servers"
That resolved itself once the route was back.
Collateral Damage #2: NAS Mounts
Checked the NAS mounts:
| Mount | IP | Status |
|---|---|---|
| Meridian-Mako-Silo | 192.168.20.50 | ✅ Working (guest) |
| Caph-Silo | 192.168.20.7 | ❌ Failed |
| Matar-Silo | 192.168.20.170 | ❌ Failed |
dmesg told the story:
CIFS: Status code returned 0xc000006d STATUS_LOGON_FAILURE
CIFS: VFS: \\192.168.20.7 Send error in SessSetup = -13
Wrong SMB credentials. The credential files had a password that doesn't match what's configured on the Synology.
The credential files at /etc/cifs-Caph-Silo and /etc/cifs-Matar-Silo contain:
username=logistics-officer
password=<wrong password>
This isn't a crash-related issue - it's been broken since the initial setup. I just never noticed because Meridian-Mako-Silo (guest access) was working fine and that's the one I use most.
Fix needed: Either update the password on the Synology NAS devices, or update the credential files with the correct password.
Documentation Discrepancy
While investigating, I found a documentation error. An old setup doc said Caph-Silo was at 192.168.20.8. The actual IP is 192.168.20.7 (confirmed via ping).
Added to my todo list: fix the docs.
Startup Script Issue
The tailscale mount script at /etc/local.d/tailscale-mounts.start only waits for 192.168.20.50 (Meridian-Mako-Silo) before trying to mount. It doesn't check for .7 or .170.
So even if the Synology credentials were correct, the mounts might fail on boot because the script doesn't wait for them to be reachable.
The Stack That Manages This
Current network management on Capella-Outpost:
| Component | Role |
|---|---|
| dhcpcd on enp5s0 | Primary network (via OpenRC) |
| Tailscale | VPN overlay, manages /etc/resolv.conf |
| /etc/local.d/fix-tailscale-routes.start | Fixes route conflicts |
| /etc/local.d/tailscale-mounts.start | Mounts NAS after Tailscale |
It's a bit fragile. If any piece doesn't run in the right order, things break.
Quick Recovery Commands
For future reference:
# Test NAS connectivity
ping -c1 192.168.20.50 # Meridian-Mako-Silo
ping -c1 192.168.20.7 # Caph-Silo
ping -c1 192.168.20.170 # Matar-Silo
# Check Tailscale route
ip route show | grep 192.168.20
# Add missing route
sudo ip route add 192.168.20.0/24 dev tailscale0
# Mount NAS manually
sudo mount /mnt/Meridian-Mako-Silo # Works (guest)
sudo mount /mnt/caph # Fails (wrong creds)
sudo mount /mnt/matar # Fails (wrong creds)
# Kill the offending language server
pkill -f language_server_linux_x64
Outstanding Tasks
- Fix Synology SMB credentials (Caph-Silo & Matar-Silo)
- Make 192.168.20.0/24 route persistent
- Update tailscale-mounts.start to handle all 3 NAS devices
- Update docs with correct Caph-Silo IP
- Monitor Antigravity language server for recurrence
System freeze. Language server burning 41% CPU while "idle". Three active connections to Google's cloud. 3.3GB of RAM.
Sometimes the tools that are supposed to help you are the ones causing the problems.