How To Install Node.js with NVM (Node Version Manager) on a VPS

Great article. Straightforward instructions.

Great, thanks!

The only line that didn’t work for me is “source ~/.bash_profile”.

I (on Ubuntu 12.04) had to do “source ~/.bashrc” and “source ~/.profile” .

@erelsgl: Thanks, I’ve updated the article. It should be .profile.

Another problem I just encountered is that, after reboot, the node version goes back to the previous one (see also this SO post: http://stackoverflow.com/questions/9170713/node-js-version-goes-back-to-0-4-form-0-6-on-reboot-nvm ). To solve this, I had to add the line “nvm use xxx” to my .profile script.

@erelsgl You could also do nvm alias default XX.XX.XX

From what I see, .profile starts nvm and that will set the $PATH correctly.

I have found that when using non-login shell (for example an SSH terminal), the .profile does not get sourced. I found some info about how to deal with it:

1. http://superuser.com/questions/176404/bash-not-loading-profile-in-new-session-on-linux
2. http://superuser.com/questions/183870/difference-between-bashrc-and-bash-profile/183980#183980

I ended up moving: [ -s $HOME/.nvm/nvm.sh ] && . $HOME/.nvm/nvm.sh # This loads NVM from .profile to .bashrc

In order to select the correct path anytime the server restarts, with or without console, I had to put it in a “@reboot” record in the root crontab:

@reboot PATH=/root/.nvm/v0.10.18/bin:$PATH

Another way to solve this is to use nvm-global: https://github.com/xtuple/nvm

Great article, nice and clear – thanks!

This tutorial didn’t work on the LAMP droplet, it asked me to install git first (the tutorial says you can do it later). So I tried installing git and THAT didn’t work <grumble> Here are the required steps that work for the LAMP stack image: apt-get install git-core (which fails) apt-get update (which works) apt-get install git-core (which now works) curl https://raw.github.com/creationix/nvm/master/install.sh | sh (which now works) n=$(which node);n=${n%/bin/node}; chmod -R 755 $n/bin/*; sudo cp -r $n/{bin,lib,share} /usr/local (which fails) cd ~ cp -r .nvm/v0.10.13/bin/node /usr/local (which works)

I very much appreciate the effort Digital Ocean is making in these tutorials, but they are either not configuring some of their droplet images properly or these instruction don’t work across all versions of unix. This tutorial was marked ‘beginner’ and obviously, it was far from that.

@scott it didn’t say ubuntu or did he mention a droplet. If you would have used Arch or Fedora there wouldn’t even be an apt-get, also explaining for every distro doesn’t make a lot of sense and would just clutter the tutorial. Updating repos and installing apps should count before beginner level, as its the first thing you learn. Also Updating repos should be the first thing to do on any distro anyway as you can’t expect DO to do daily updates of their droplets. But i agree that if you need git for the first command, it should say so in the beginning, but the ‘copy’ command of yours in not a good solution, as its explicit for one node version and not the one you are using right now. But maybe it should mention to move yourself to you home folder (cd ~) before executing.

NVM install script needs to be updated. Here is the newest curl, courtesy of the github page:

curl https://raw.githubusercontent.com/creationix/nvm/v0.10.0/install.sh | bash

BE CAREFUL! The command listed above, is actually not well thought out and could lead to SUDO breaking (at the least) on many systems. Consider the following

chmod -R 755 $n/bin/*; 

This recursively will change permissions on all executable files within /bin to 755, which is not a good thing in the case of SU or SUDO, which requires the sticky bit being set and will result in the following, unwanted behavior as the final part relocates the entire /bin to /usr/local/bin…

$ which sudo
/usr/local/bin/sudo

Perhaps a safer way to do this command, would be to remove the wildcard ( * ) and do something like this…

chmod 755 $n/bin/nodejs;  chmod 755 $n/bin/node; chmod 755 $n/bin/nvm; chmod 755 $n/bin/npm;      

Even wild card on say /bin/n* would include - /usr/bin/nproc - used for limiting system processes and resources… either way… I really suggest NOT using that command and add the following code for each users ~/.profile

export NODE_PATH=/usr/local/bin/node
export NVM_DIR=/usr/local/bin/nvm
export PATH="$NODE_PATH/bin:$PATH;$NVM_DIR/bin:$PATH;"

This is not a complete example but using environment variables is a lot safer than changing system binary permissions and copying these without the correct permissions, setuid, stickybit etc… I appreciate the article and thanks for sharing! :-)

curl https://raw.github.com/creationix/nvm/master/install.sh | sh

should be

curl https://raw.github.com/creationix/nvm/master/install.sh | bash

This helped a lot. Thanks.

Hi - I maintain nvm. Please update your link to use the latest tagged release (per the readme on the github repo). Installing from master is dangerous, as you may install unstable code. Similarly, please pipe it to bash and not to sh (as is in the readme).

I’ve updated the command, thanks!

hmm does following the above mean you can only install new packages under root permissions? I previously had installed with npm but it was in my users home directory, which I understand is not ideal, n=$(which node);n=${n%/bin/node}; chmod -R 755 $n/bin/*; sudo cp -r $n/{bin,lib,share} /usr/local command to move it to the /usr/local/bin folder. Now I can no longer install packages, like ember-cli, without using sudo, which you are not supposed to do according to their documentation.

The only problem I have with this is that node isn’t in my path. That is to say that running $ node --version gets me bash: node: command not found. Anyone got any thoughts on this?

In case of the node.js app running on a server with a non-interactive shell, the PATH is missing /usr/local/bin. What’s the best practice for that ? Should I export it in the init script ?

                PATH=/usr/local/bin:$PATH

Your n=$(which node);n=${n%/bin/node}; chmod -R 755 $n/bin/*; sudo cp -r $n/{bin,lib,share} /usr/local command has messed up with my server and now I cannot sudo, because I started getting this error:

[aaaa@bbbb ~]$ sudo su
sudo: effective uid is not 0, is sudo installed setuid root?

Any suggestions?

Very good tutorial. I like the part when installing Node globally and using it as a standard user. Thanks for sharing.

Great article! But I’m unable to start an application with this, I’m trying to install an application generated with Express, but npm install does not create ./bin/www required by the app. Is there way to remove all of this so I can try another nodejs installation methods? Regards!

for every new terminal wndow, i have to source the ~/.profile and run nvm use <node version> any sollution to this?

Thanks!

Perhaps worth adding that it’s necessary to deactivate your node version through nvm after running the command to copy it to /usr/local/ in order for that to be used. Stumped me for a little while. Otherwise, thanks for the great article!

I tried to run the big command

n=$(which node);n=${n%/bin/node}; chmod -R 755 $n/bin/*; sudo cp -r $n/{bin,lib,share} /usr/local

but the first time I got

cp: not writing through dangling symlink ‘/usr/local/bin/node’

I tried again, this time without any feedback, which I think was ok. I then run sudo -s; which node I got

/home/myuser/.nvm/versions/node/v6.2.0/bin/node

To get it right I had to restart the terminal.

Broke my sudo D:

You should listen to mrhassell’s comment further up this comment thread and change that part of the article to something safer. Luckily this just happened on my personal server and reinstalling Ubuntu won’t be hard. Here’s the whole story.

Now I know not to trust odd-looking commands even on digitalocean…

I’m getting this response when I run the command to fix which node: cp: cannot create regular file ‘/usr/local/bin/node’: Text file busy

Hi, I’m running a Centos 7 and I followed all the steps from above, so far everything goes ok, but I’m having this issue, all my logged in users can see and make use of node js, except for one: the root user, it cannot see or execute node, even thought when I execute:

which node

I get this message: “/usr/bin/which: no node in (/sbin:/bin:/usr/sbin:/usr/bin)”

It only happens with the root user.

By the way great article.

I know that isn’t the best way to resolve the “user variable environment” issue, but here is the way that I found to set node in env everytime that change node version by using nvm use <version> command:

{ ... }
      if [ -n "${NVM_USE_OUTPUT-}" ]; then
        n=$(which node);n=${n%/bin/node}; chmod -R 755 $n/bin/*; sudo cp -r $n/{bin,lib,share} /usr/local
        nvm_echo "$NVM_USE_OUTPUT"
      fi
{ ... }

This code is a piece found in ~/.nvm/nvm.sh that log when the nvm use <version> command is runned successfully.

The difference here, is that sudo password will be asked always that you run nvm use <version>

[]'s

Latest install script:

wget -qO- https://raw.githubusercontent.com/creationix/nvm/v0.33.6/install.sh | bash

source: https://github.com/creationix/nvm#install-script

**Great ** this helped alot.