Jun 12, 2020

Personal Server Setup

Recently I’ve been playing with VMs/VPSs as web servers, so I’ve been setting up lots of servers. This is both to self-document what I’ve been doing, and maybe it gives someone a new tool to use.

Setup the VM however with whatever you’d like. I’ve been a fan of debian on the server recently.

ssh root@<ip-addr>
create a user for me and give me sudo privileges

# my terminal doesn't have the best terminfo support out of the box; default to xterm-256color
export TERM=xterm-256color
adduser username
usermod -aG sudo username

If needed, create the .ssh directory with perms

su - username  # switch to user
cd
chmod go-w ~/
chmod 700 ~/.ssh
chmod 600 ~/.ssh/authorized_keys

create a ssh key (ssh-keygen -t rsa -b 4096 -o -a 100) for connecting to the server, ssh-copy-id it up to the server

ssh-copy-id -i ~/.ssh/file username@<ip-addr>

Add the block for the server to my ~/.ssh/config file:

Host vps
  User username
  Hostname <server ip>
  IdentityFile ~/.ssh/private_key

Then I can just connect with ssh vps

ssh onto the server and run my bootstrap script: That sets up some bash defaults: aliases, neovim configuration, installs fzf, prompts me to setup Github username/email.

sudo apt install neovim git curl
bash -c "$(curl -fsSL https://raw.githubusercontent.com/purarue/bootstrap/master/bootstrap)"

Strengthen ssh configuration: disable root login, password authentication (have to use ssh-key). Make sure the following lines exist and are uncommented in /etc/ssh/sshd_config:

ChallengeResponseAuthentication no
PasswordAuthentication no
UsePAM no
PermitRootLogin no

6b. May have to setup ufw, to setup ports

# apt install ufw
ufw allow 22
ufw allow 80
ufw allow 443
ufw enable
ufw status
ufw reload

Reload ssh: sudo systemctl reload ssh

Setup a gitlab/github ssh key and start an ssh-agent, but dont ‘eval ssh-agent’ every time you log in, just the first time, by putting this in ~/.bash_profile:

if [ ! -S ~/.ssh/ssh_auth_sock ]; then
  eval `ssh-agent`
  ln -sf "$SSH_AUTH_SOCK" ~/.ssh/ssh_auth_sock
fi
export SSH_AUTH_SOCK=~/.ssh/ssh_auth_sock
ssh-add -l > /dev/null || ssh-add ~/.ssh/github

Install nginx:

Just the base installation for now, test it by going to the IP address in the browser to make sure firewall is properly configured.

sudo apt install nginx

This is heavily modified after my applications are set up, see below.

Install lots of things to configure my applications/webapps, see vps:

# setup docker
sudo apt-get install apt-transport-https ca-certificates curl gnupg2 software-properties-common
curl -fsSL https://download.docker.com/linux/debian/gpg | sudo apt-key add –
sudo add-apt-repository "deb [arch=amd64] https://download.docker.com/linux/debian buster stable"
sudo apt-get update
sudo apt-get install docker-ce docker-ce-cli containerd.io
sudo groupadd docker
sudo usermod -aG docker "$(whoami)"
sudo systemctl restart docker
docker run hello-world  # test connection to docker socket, may require a restart/relog

# setup postgresql
sudo apt install postgresql postgresql-client
sudo su
su postgres
adduser glue_worker # primarily used for my elixir server
createuser --pwprompt glue_worker
createdb -O glue_worker glue_db
psql -d glue_db -h localhost -U glue_worker # test connection

# rust
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh

# node/npm
curl -sL 'https://deb.nodesource.com/setup_14.x' | sudo bash -
sudo apt install nodejs

# lots of other apt installs
sudo apt update
sudo apt install python3.7 docker-compose pipenv supervisor jq elixir erlang-inets erlang-dev \
    erlang-parsetools erlang-xmerl rsync goaccess apache2-utils fail2ban libssl-dev \
    htop tree unzip

# Set up environment (put this in ~/.bash_profile):
# NPM global packages are put in ~/.local/share/npm-packages to avoid permission errors/requiring sudo to install npm packages
export NPM_CONFIG_PREFIX="${HOME}/.local/share/npm-packages"
PATH="$HOME/.rvm/rubies/ruby-2.7.0/bin:$HOME/.cargo/bin:$HOME/vps:$HOME/.local/bin:$NPM_CONFIG_PREFIX/bin:$PATH"
export PATH
# rvm automatically adds its rvm function to ~/.bash_profile

# re-source/relog in to source environment variables in ~/.bash_profile; now that npm dir is set, install global npm packages
npm install -g uglifycss elm html-minifier

# install ranger for some nicer file management, and speedtest-cli in case I want to check network speed
pip3 install --user --upgrade ranger-fm speedtest-cli

# add myself to the adm group so that I have permission to view logs at /var/log/ without sudo
sudo usermod -aG adm "$(whoami)"

# Run my `vps_install` script to setup all of my application data/verify I have all of my packages installed: https://github.com/purarue/vps
# sets up logging for all my applications, centralizes that in ~/logs, sets up basic HTTPAuth for nginx using apache2-utils, sets up supervisor for process management
git clone git@github.com:purarue/vps ~/vps
cd ~/vps
./vps_install  # clone/setup all my applications
./generate_static_sites  # clone/generate all my static sites

Setup my DNS info on my domain name registrar, point it to the IP address of the VPS, follow instructions to set up certbot for HTTPS:

Make sure the server_name directive exists in the server block running on port 80 in /etc/nginx/sites-available/default before trying to do certbot, so it can grab the domain name from there.

sudo apt install certbot python-certbot-nginx
sudo certbot --nginx
sudo certbot renew --dry-run # test renewal

Disable logrotate for certain logs. I’m not a fan of it rotating to the .tar.gz files, I prefer to have the one giant log file to I can parse/script with it easier.

I do this for /var/log/auth.log, nginx and fail2ban:

cd /etc/logrotate.d/
sudoedit rsyslog  # remove auth
sudo mv nginx nginx.disabled
sudo mv fail2ban fail2ban.disabled

Setup the basic nginx server blocks to have nginx redirect from HTTP to HTTPS and from my https://www.purarue.xyz to just https://purarue.xyz. In /etc/nginx/sites-available/default:

# redirect from HTTP to HTTPS
server {
  listen 80 default_server;
  listen [::]:80 default_server;
  server_name _;
  return 301 https://$host$request_uri;
}

# redirect from www to non-www URL
server {
  listen 443 ssl;
  server_name www.purarue.xyz;
  ssl_certificate /etc/letsencrypt/live/purarue.xyz/fullchain.pem; # managed by Certbot
  ssl_certificate_key /etc/letsencrypt/live/purarue.xyz/privkey.pem; # managed by Certbot
  include /etc/letsencrypt/options-ssl-nginx.conf;
  ssl_dhparam /etc/letsencrypt/ssl-dhparams.pem;

  rewrite ^/(.*) https://purarue.xyz/$1 permanent;
}

server {

  listen [::]:443 ssl http2 ipv6only=on;
  listen 443 ssl;
  ssl_certificate /etc/letsencrypt/live/purarue.xyz/fullchain.pem; # managed by Certbot
  ssl_certificate_key /etc/letsencrypt/live/purarue.xyz/privkey.pem; # managed by Certbot
  include /etc/letsencrypt/options-ssl-nginx.conf;
  ssl_dhparam /etc/letsencrypt/ssl-dhparams.pem;

	root /var/www/html;
	index index.html;
	server_name purarue.xyz;

# .... location blocks continued for different servers
}

Its also possible to configure this at the DNS level using a CNAME, but I like being able to see which requests are getting redirected/whose going to www in my nginx logs.

Configure linux/nginx for better performance/more connections/open files (especially since I use phoenix as my main server):

## /etc/nginx/nginx.conf
# at the top
worker_rlimit_nofile 37268;
events {
    worker_connections 37268;
}

## /etc/security/limits.conf
# at the bottom
* soft nofile 37268
* hard nofile 37268
root soft nofile 37268
root hard nofile 37268

## /etc/systemd/system.conf
# uncomment this line and set to:
DefaultLimitNOFILE=37268

# In both of these files:
## /etc/pam.d/common-session
## /etc/pam.d/common-session-noninteractive
# at the bottom, add:
session required pam_limits.so

Restart the system, and check that nginx’s file limit has increased:

ps -ef | grep nginx

cat /proc/<nginx_pid>/limits

Setup fail2ban to stop unauthorized ssh attempts/temporarily banning suspicious behavior. I pretty much followed this tutorial, and enabled this badbots filter. Remember to whitelist your own IP:

ignoreip = 127.0.0.1 ::1 <your public ipv4 address>

Setup some server monitoring.

Install netdata and my fork of superhooks (in my supervisord.conf in my vps repo) for server and supervisor process monitoring respectively:

bash <(curl -Ss https://my-netdata.io/kickstart.sh)

I use the go module for nginx, and use webhooks for discord (netdata docs), to get notifications whenever something goes wrong on my server (e.g. high CPU usage, too many 300x/400x requests, an application using too much ram, lots of other alerts that come default with netdata…), and whenever one of my supervisor processes unexpectedly crashes.

Both netdata and goaccess (nginx log parser) are password protected with apache2-utils’s htpasswd, which is setup in vps_install. To set it up:

sudo htpasswd -c /etc/nginx/.htpasswd username, and then on the routes:

location /logs {
  alias /home/username/.goaccess_html;
  try_files $uri $uri/ =404;
  auth_basic "for logs!";
  auth_basic_user_file /etc/nginx/.htpasswd;
}

location /netdata/ {
  proxy_pass http://127.0.0.1:19999/;
  auth_basic "for netdata!";
  auth_basic_user_file /etc/nginx/.htpasswd;
}

References: