nvm keeps "forgetting" node in new terminal session

node.js macos npm homebrew nvm

184,830

Solution 1

Try nvm alias default. For example:

$ nvm alias default 0.12.7

This sets the default node version in your shell. Then verify that the change persists by closing the shell window, opening a new one, then: node --version

Solution 2

Alias to node itself to avoid updating the default alias along with node version updates later on.

nvm alias default node

Solution 3

In my case, another program had added PATH changes to .bashrc

If the other program changed the PATH after nvm's initialisation, then nvm's PATH changes would be forgotten, and we would get the system node on our PATH (or no node).

The solution was to move the nvm setup to the bottom of .bashrc

### BAD .bashrc ###

# NVM initialisation
export NVM_DIR="$HOME/.nvm"
[ -s "$NVM_DIR/nvm.sh" ] && \. "$NVM_DIR/nvm.sh"  # This loads nvm

# Some other program adding to the PATH:
export PATH="$ANT_ROOT:$PATH"

Solution:

### GOOD .bashrc ###

# Some other program adding to the PATH:
export PATH="$ANT_ROOT:$PATH"

# NVM initialisation
export NVM_DIR="$HOME/.nvm"
[ -s "$NVM_DIR/nvm.sh" ] && \. "$NVM_DIR/nvm.sh"  # This loads nvm

(This was with bash 4.2.46 on CentOS. It seems to me like a bug in bash, but I may be mistaken.)

Solution 4

To install the latest stable version:

nvm install stable

To set default to the stable version (instead of a specific version):

nvm alias default stable

To list installed versions:

nvm list

As of v6.2.0, it will look something like:

$ nvm list
         v4.4.2
->       v6.2.0
default -> stable (-> v6.2.0)
node -> stable (-> v6.2.0) (default)
stable -> 6.2 (-> v6.2.0) (default)
iojs -> N/A (default)

Solution 5

nvm does its job by changing the PATH variable, so you need to make sure you aren't somehow changing your PATH to something else after sourcing the nvm.sh script.

In my case, nvm.sh was being called in .bashrc but then the PATH variable was getting updated in .bash_profile which caused my session to find the system node before the nvm node.

View more solutions

184,830

Author by

frhd

Updated on July 15, 2022

Comments

frhd almost 2 years
Upon using a new terminal session in OS X, nvm forgets the node version and defaults to nothing:

$ nvm ls:
```
         .nvm
     v0.11.12
     v0.11.13
```
I have to keep hitting nvm use v.0.11.13 in every session:
```
         .nvm
     v0.11.12
->   v0.11.13
```
I've tried both the brew install, as well as the official installation script.

My .profile for the brew version:
```
#nvm
export NVM_DIR=~/.nvm
source $(brew --prefix nvm)/nvm.sh
```
And for the install.sh script:

$ curl https://raw.githubusercontent.com/creationix/nvm/v0.10.0/install.sh | bash
```
#nvm
export NVM_DIR="/Users/farhad/.nvm"
[ -s "$NVM_DIR/nvm.sh" ] && . "$NVM_DIR/nvm.sh"  # This loads nvm
```
Any clue to what I'm doing wrong?
Mild Fuzz over 8 years

how can I inspect this? I suspect this is the case.
Tom over 8 years

This is actually better than aliasing to a specific version (the other answer). It is also the approach suggested at github.com/creationix/nvm#install-script .
vaughan over 8 years

Just a note, this relies on including the v in the version when setting the remote. E.g. v4.2.4 instead of 4.2.4.
Jonah almost 8 years

Was not working for me, and I didn't have the problem maxwell states. I didn't use brew but I upgraded/reinstalled my nvm and that fixed it.
Alejandro Silva almost 8 years

this was my case thanks! i was setting the export on top on my .zshrc file, just move it to the bottom and the problem was solved
Kelly S. French almost 8 years

My version of nvm (1.1.0) does not recognize the alias parameter
Evan Mattson over 7 years

Also, depending on where the nvm init script is loaded in your profile, it's possible that a new shell doesn't load the expected version, even with the default aliased properly.
goodmanship over 7 years

If that doesn't work, make sure in your .bash_profile (or .bashrc or whatever) you don't have anything modifying PATH after source xx/nvm.sh
Eugene Tartakovsky over 7 years

That was my case. Solution was to move NVM ~/.bash_profile entry to the end of the file + set nvm alias default and nvm use to the version i needed.
LJHarb about 7 years

@KellyS.French (nvm maintainer here) nvm has never yet hit 1.0 - you may have installed it off of npm. Install only using the curl script at nvm.sh and you'll get the right version.
Kelly S. French about 7 years

@LJHarb I'll try that. It does make me wonder how my environment was translating my 'nvm' command. I installed nvm by downloading the Windows binary so is there a .bat or .ps1 equivalent of the curl script you mentioned?
LJHarb about 7 years

@KellyS.French ah! nvm does not support windows (except for BashOnWindows) - you may have nvm-windows, which is a different project.
sn0r almost 7 years

Setting a path variable has nothing to do with the nvm version selection, surely?
XoXo over 6 years

you can also specify a lts version in later version of nvm. e.g.: nvm alias default lts/boron
Jeremy over 6 years

I cannot get this to work on an Amazon linux image. I have tried the above and defining use, etc... I have to manually run: nvm use 8.8.1 everytime. this is a pain. NVM is too broken. I am just manually installing node. No issues this way.
Raja about 6 years

I would recommend using &&s instead of ;s in that command. If the which fails to find node, we don't want to proceed to chmod /bin/*!
clozach about 6 years

Yes! Not sure if it's 'cause I'm using fish shell, but the accepted answer didn't work for me. n works a treat!
YannickHelmut almost 6 years

I actually had to uninstall the version that was used by default since nothing else worked out. $ nvm uninstall 6.11.3 and then I switched to 6.11.5...
Redsandro over 5 years

This helped me! Default was stuck.
bobbytables over 5 years

Give this guy a medal! I had additional loading of other *rc files in my zshrc after the NVM_DIR export it drove me crazy. Thank you!
ajay over 5 years

Thanks a lot! This was the only solution which sets node path in bash as well and worked for me. I just used "nvm install node" to install latest node version rather than using "nvm install v7.8.0"
Admin over 5 years

This was the solution for me. Moved the following to the bottom of .bashrc: export NVM_DIR="$HOME/.nvm" then . "/usr/local/opt/nvm/nvm.sh"
Denis Howe over 5 years

nvm.sh should modify your PATH after you set it (e.g. with the above). This let nvm put its version directory before /usr/local/bin where the "system" version of node lives.
Denis Howe over 5 years

The behaviour is very confusing. The first nvm install X sets your nvm alias default to X, which is what you'll get in a new shell until you do another nvm use or nvm install. Even if you do nvm use or nvm install, that won't update nvm alias so you're stuck with the first default version unless you update the alias yourself.
png about 5 years

This worked great and the suggestion to use nvm list helps users visualize what is happening with their versioning. Thanks!
Carsten Führmann over 4 years

This answer worked for me on the Windows Subsystem for Linux (WSL) with Ubuntu 18.04
WakeskaterX about 4 years

ffs this was such an easy answer, this has been bugging me for a while and thought it was going to be some annoying path thing to fix, thank you.
Trees almost 4 years

Thank you! This was driving me bonkers.
Trent almost 4 years

Instead of pinning Node to a version, you can specify 'lts/*' as the version identifier to use the latest LTS version.
Jason R Stevens CFA over 3 years

I don't think you need upped privileges via sudo in 4. I'll edit the answer to remove, as there appears to be no benefit to privilege escalation here, as the answer works within the home directory. If that's not the case, I am curious to hear why--thanks!
Ahmed Mahmoud about 3 years

That works perfectly fine, but why use stopped working?
vikramvi about 3 years

Please clarify why you didn't use "source ~/.nvm/nvm.sh" after export statement ?
Raja almost 3 years

@vikramvi We did, on the last line. \. is the same as source. The [ -s ... ] just checks that the file exists and is not empty. If you are asking why not move just the source line to the bottom, and leave the export NVM_DIR where it was originally, the only reason is to keep the NVM lines together for clarity. You could separate them if you wanted to.
Akash Kumar Seth almost 3 years

Thanks. this solution is easiest if any one is facing issue on ZSH.
Buh Buh almost 3 years

nvm alias default stable
Krishna Pravin over 2 years

If setting default alias does not work, 1. Try moving the export NVM_DIR=.... in bash profile to the last 2. If that also does not work, uninstall nvm and reinstall it and install the node version you want Don't spend too much time finding a fix. reinstall is very quick
sails44 over 2 years

Yep, this worked for me. Otherwise I was seeing the default revert with each new terminal.
AliRehman7141 over 2 years

Thanks @arseniy
Rohan Shukla over 2 years

Doesn't persist. I tried installing nvm, everything works, but when I remove that node version it says it's not installed, but when I check node version it comes up?
MuhsinFatih over 2 years

I don't understand, why would an entirely unrelated path in the $PATH variable affect node version? Was your "other program" modifying the node path?
Raja over 2 years

@MuhsinFatih What seems to happen is when export PATH="$ANT_ROOT:$PATH" executes, for some reason bash uses an earlier copy of PATH, ignoring the directory that nvm added. The result is that we don't have nvm's directory on the PATH at all. It gets "forgotten".
andresp about 2 years

In fish with jorgebucaran/nvm.fish I get nvm: Unknown command or option: "alias" (see nvm -h). I guess there you need to use set --universal nvm_default_version $VERSION
Adriana Hernández about 2 years

Thanks! In my case it worked by running this in vs code terminal: nvm use 16.14.2 and then restart VS Code. Beautiful!
B T about 2 years

Thank you! None of the other solutions worked for me either. Maybe this is a recent nvm bug.
Wayne Mao almost 2 years

It can be simplify to nvm use default >> /dev/null 2>&1 I use zsh and I found that the accepted answer doesn't work. I think on zsh, this might be the best(only?) choice.