Installing node.js on a VPS

In this article I will tell you what problems will have to be solved in order to get a VPS with a really working node.js service. These are all basic things, but maybe someone else will come in handy.

Install node

Problem: the repository may not have a package for the correct node version. Alternatively, you may have two projects that require different versions of the node.
Solution: on development machines in such cases use nvm. On the combat server, it can also be easily used, just have to write a special script to start your server.

Access to port 80

Problem: the application must have superuser rights in order to read port 80 (for details, see CAP_NET_BIND_SERVICE, man capabilities ). But launching a node with such rights is considered unsafe.
Solutions:

• use a proxy server like Nginx or HAProxy .
• redirect requests from port 80 to any one with a large number using iptables . Practically done with one command: iptables -t nat -I PREROUTING -p tcp --dport 80 -j REDIRECT --to-port 8000
• start with the rights of the superuser, but immediately after connecting to the socket the process will programmatically lower its privileges. Read more here .

Load balancing

Problem: node is single-threaded and cannot effectively use multi-core processors.
Solutions:

• in fact, all he can. See cluster API
• and there is also node-http-proxy or fugue .
• you can use Nginx (here it’s written like: one , two ). Or HAProxy .

')

To use node directly or through Nginx?

The question was discussed a lot where, but there is no consensus On the one hand, nginx is a reliable, proven solution over the years; on the other hand, node itself is able to do the same.

As a rule, one of those options is offered:

• Node does everything at once, right up to distribution of static files.
• node is behind haproxy.
• The node is behind Nginx and is used only for generating pages, everything else (static distribution, balancing, https operation) does Nginx

Restart of the fallen process

Problem: how to restart the server if it suddenly falls?
Solutions:

• If you handle exceptions correctly, there will be almost no drops.
• Upstart (which you will most likely use for the initial launch of the service) is able to restart the fallen processes.
• Well, a bunch of solutions for monitoring and restarting: forever , monit , supervisord , daemontools .

Updating a service with git

Described here