When things go wrong
Troubleshooting#
Find the message you see, or press / and search for words from it.
Golden rules: read the exact error message; change one thing at a time; don't reinstall or erase anything to "fix" a connection problem.
First: which kind of problem?#
flowchart TD
A["Can't connect with ssh"] --> B{"What does it say?"}
B -- "Could not resolve hostname" --> N["Name problem<br/>see: Could not resolve"]
B -- "Connection timed out" --> T["Network / server off<br/>see: timed out"]
B -- "Connection refused" --> R["SSH not running<br/>see: refused"]
B -- "Permission denied" --> P["Key problem<br/>see: Permission denied"]
B -- "REMOTE HOST IDENTIFICATION<br/>HAS CHANGED" --> H["Reinstalled server<br/>see: host changed"]
Connecting#
Could not resolve hostname homeserver#
The name can't be found.
- At home: try
homeserver.local, or the number from your notebook. - Away from home:
.localnames never work away from home. Is Tailscale on on this device? Check its icon or app. - Still stuck: on the server,
tailscale ip -4gives a100.x.x.xnumber. Use that.
Connection timed out#
Your device can't reach the server at all.
- Is the server on and plugged in? Lid closed without Chapter 13? Then it's asleep.
- Is the address right? Addresses can change after a router restart. Use the name, or reserve the address (Chapter 6).
- Away from home: is Tailscale connected on both devices? Check the Tailscale Machines page.
- Did the server lose Wi-Fi? See Wi-Fi stopped working.
Connection refused#
The server is reachable but SSH isn't running. At the server's keyboard: sudo systemctl start ssh.
Permission denied (publickey)#
The server doesn't accept this device's key.
- Are you on a new device? Add its key (Chapter 10).
- Is your shortcut pointing to the right key? Run
ssh -G homeserver | grep -i identityfile. - Wrong username? Check the
Userline in your~/.ssh/config.
WARNING: REMOTE HOST IDENTIFICATION HAS CHANGED!#
Normal after you reinstall the server, because it has a new identity. If you did not reinstall, stop and find out why. If you did, remove the old record on your Mac or PC:
ssh-keygen -R homeserver
ssh-keygen -R homeserver.local
Then connect again and answer yes.
It asks for my key passphrase every time#
- Mac:
ssh-add --apple-use-keychain ~/.ssh/id_ed25519_homeserver, and make sure your config hasUseKeychain yes(Chapter 7). - Windows: the ssh-agent service isn't running; redo Chapter 8, step 6.
command not found for a command from this guide#
Check the prompt: you're probably typing in the wrong place. Windows commands (Get-Service, notepad…) typed on the server fail, and so do server commands typed on Windows. See Where am I typing?. Type exit to leave the server.
Server problems#
alex is not in the sudoers file#
You set a root password during install. At the server's keyboard, log in as root with that password, then:
usermod -aG sudo alex
Log out, log in again as alex.
Wi-Fi stopped working#
At the server's keyboard:
ip -brief address
sudo ifup wlp2s0
(Your Wi-Fi name may differ from wlp2s0; it's the one that starts with wl in the first command.)
- If you edited Wi-Fi settings recently, a typo is the usual cause. Run
sudo journalctl -b -u networkingand read the last lines. - Moved the server to a new Wi-Fi network? The installer only saved the old one. Easiest fix: plug in an Ethernet cable.
Could not get lock /var/lib/dpkg/lock#
Automatic updates are running. Wait 5–10 minutes and try again. Don't delete lock files.
The server won't start after I added a hard drive#
Usually an /etc/fstab line without nofail, or a typo. At the server's keyboard, it may show an emergency mode prompt. Type your password, run nano /etc/fstab, put # in front of the new line, save, and run reboot. Then redo Chapter 15, step 5 carefully.
Disk full#
df -h
sudo du -sh /srv/* /var/log /home/* 2>/dev/null
The second command shows which folders are big.
File sharing#
Finder / Explorer can't connect to the share#
- Does
ssh homeserverwork from the same device? If not, fix that first (above). - On the server:
systemctl is-active smbdshould sayactive, andtestparm -sshould show[files]. - Login rejected? Use the Samba password, not your login password. Reset it with
sudo smbpasswd -a alex. - Using
homeserver.local(no Tailscale)? You need the home-network firewall rule (Chapter 16 tip).
Security layer#
channel 2: open failed: administratively prohibited when opening a tunnel#
Tunnels are switched off. Check sudo sshd -T | grep allowtcpforwarding says local (Chapter 28).
SSH stopped working away from home after tightening the firewall#
Use ssh homeserver-remote (Tailscale). Plain homeserver.local only works at home, and after Chapter 29 SSH is only allowed from your home network and Tailscale.
I banned myself with Fail2ban#
From another device or the server's keyboard: sudo fail2ban-client status sshd to see the address, then sudo fail2ban-client set sshd unbanip ADDRESS.
Cockpit login fails#
Use your server account password (the sudo one), not your key passphrase. Root can't log in.
gpg: problem with the agent or no passphrase prompt#
Run export GPG_TTY=$(tty) in the same window first, and keep --pinentry-mode loopback in the command (Chapter 33).
AI#
error while loading shared libraries: libgomp.so.1#
The engine needs one extra package: sudo apt install libgomp1.
llama-model is not active#
Read why:
sudo journalctl -u llama-model -n 50 --no-pager
| Message contains | Fix |
|---|---|
No such file or failed to load model |
The path in /etc/ai-runtime/model.env doesn't match the file name. Compare with sudo ls /srv/ai/models |
Permission denied |
Redo the chown/chmod lines in Chapter 20 |
out of memory, or the service is killed |
Use the smaller model, or lower --ctx-size to 1024 |
/health says the model is loading, or curl gets no answer#
Wait a minute after a restart: loading takes time. The first question after a start is also slow (about a minute).
The answer is empty#
Use the address /v1/chat/completions, not /completion.
Answers stop in the middle#
Raise max_tokens in the request (or in the chat page settings). In ask-pages, change max_tokens: 400.
bind: Address already in use when opening a tunnel#
An old tunnel is still open. Close that terminal window, or pick another number on your side.
The agent dashboard says "origin rejected" or won't load#
Use exactly -L 18765:127.0.0.1:8097 and open http://127.0.0.1:18765. Other port numbers are refused.
aisystem stops when I log out#
Linger is off: sudo loginctl enable-linger $USER, then systemctl --user enable --now aisystem.
Recoll finds nothing in a PDF#
Either the PDF is a scan (no text inside), or poppler-utils wasn't installed when you indexed. Install it, then run recollindex -c ~/.recoll-docs again.
Installing#
USB stick doesn't appear at startup#
Recreate it with Etcher (Chapter 2); try another USB port; on a PC, enable USB boot in the setup screen.
Keyboard / trackpad don't work in the installer#
Use a USB keyboard. Common on some MacBooks.
No Wi-Fi in the installer#
Use an Ethernet cable or USB-Ethernet adapter. Wi-Fi often works after installation.
Computer restarts into the installer again#
Remove the USB stick and restart.