Cannot SSH from WSL to VM, although it's SSH is possible from host outside WSL.

[pedrofgodinho]

Vagrant version

Vagrant 2.2.14

Host operating system

Windows 10 Education Version 10.0.19041 Running WSL2

Guest operating system

Ubuntu Bionic

Vagrantfile

Default file from vagrant init hashicorp/bionic64 (Comments Removed)

# -*- mode: ruby -*-
# vi: set ft=ruby :

Vagrant.configure("2") do |config|
  config.vm.box = "hashicorp/bionic64"
end

Debug output

https://gist.github.com/Fowlron/35838ae9d801e2a6713ce392a821cedf

Expected behavior

vagrant up and vagrant ssh should be able to properly connect to the running VM.

Actual behavior

Although the VM is starting and running properly, I cannot SSH from WSL into the VM. Note that SSH is possible from the windows host, as ssh vagrant@127.0.0.1 -p 2222 on powershell will successfully connect to the running VM created by vagrant.

Steps to reproduce

Starting on an empty directory (/mnt/d/vagrant/vagrantexample)

1. vagrant init hashicorp/bionic64
2. vagrant up

Note also that my env includes VAGRANT_WSL_ENABLE_WINDOWS_ACCESS=1 and VAGRANT_WSL_WINDOWS_ACCESS_USER_HOME_PATH=/mnt/d/vagrant.

[nmofonseca]

This seems to be exactly the same issue as in https://github.com/hashicorp/vagrant/issues/11716. I am also facing the same issue

[Karandash8]

Hi there,

Faced the issue too.

If I may suggest a change. The following would be a fix:

• the second port forwarding entry is created on 0.0.0.0 that allows to ssh to VirtualBox VM from WSL2 on Windows IP.
• Windows IP is automatically selected for vagrant ssh command instead of localhost.

Pretty much what I have in my workaround: https://github.com/Karandash8/virtualbox_WSL2. If that makes sense I can polish it a bit and open a PR.

[nmofonseca]

Hello @Karandash8 ,

I will give it a go and let you know how it goes. Cheers and thanks for your work.

[nmofonseca]

Hello @Karandash8 ,

I have been testing your plugin and it's working fine, thank you so much for all that work and for doing that.

Just a qq that you may know how to fix it, vagrant ssh asks for password, I believe that is more an issue re. vagrant keys being setup on /mnt/nc and it's permissions, do you have any way of wor around that?

Cheers,

[Karandash8]

Hey @nmofonseca,

Glad it worked for you!

Strange, I don't have any problem with ssh keys, it just works. I have nothing special in my vagrantfile, just: config.vm.synced_folder '.', '/vagrant', disabled: true and I keep my vagrantfile on WSL filesystem not on Windows one.

[Devorkin]

@nmofonseca I experience the same issue as you - when running vagrant ssh the process hangs for a second or two and then prompts for a password. I did find, that once I've disabled the VAGRANT_WSL_WINDOWS_ACCESS_USER_HOME_PATH export in my .bashrc - I was able to run a new machine on my env. and run vagrant ssh successfully using the Vagrant private key.

I've tried to copy the "insecure_private_key" file to my specific VAGRANT home path - but Windows WSL got wired permissions. I did try to play with it for a bit - no luck yet.

[nmofonseca]

Hey @nmofonseca,

Glad it worked for you!

Strange, I don't have any problem with ssh keys, it just works. I have nothing special in my vagrantfile, just: config.vm.synced_folder '.', '/vagrant', disabled: true and I keep my vagrantfile on WSL filesystem not on Windows one.

Apologies for the delay and taking so long to reply, wow! since March ridiculous.

I haven't used it for a while, to be honest, but I also kept my Vagrant File in WSL but I do sync using NFS my local folder with the Vagrant machine.

The vagrant ssh not working for me is not a blocker issue because I use Ansible on provision to install my own keys so I can ssh anyway in the end and run and test my ansible playbooks against the Vagrant machines, that was my main use case for Vagrant.

Once again @Karandash8 thank you so much for the work you did and for providing this plugin.

[nmofonseca]

@nmofonseca I experience the same issue as you - when running vagrant ssh the process hangs for a second or two and then prompts for a password. I did find, that once I've disabled the VAGRANT_WSL_WINDOWS_ACCESS_USER_HOME_PATH export in my .bashrc - I was able to run a new machine on my env. and run vagrant ssh successfully using the Vagrant private key.

I've tried to copy the "insecure_private_key" file to my specific VAGRANT home path - but Windows WSL got wired permissions. I did try to play with it for a bit - no luck yet.

I didn't spend much time on it either @Devorkin , I work around it by installing my own ssh keys using Ansible, but I never tried the option you mentioned, if I can I will give it a try also.

[ivanmanokhin]

@nmofonseca I experience the same issue as you - when running vagrant ssh the process hangs for a second or two and then prompts for a password. I did find, that once I've disabled the VAGRANT_WSL_WINDOWS_ACCESS_USER_HOME_PATH export in my .bashrc - I was able to run a new machine on my env. and run vagrant ssh successfully using the Vagrant private key. I've tried to copy the "insecure_private_key" file to my specific VAGRANT home path - but Windows WSL got wired permissions. I did try to play with it for a bit - no luck yet.

I didn't spend much time on it either @Devorkin , I work around it by installing my own ssh keys using Ansible, but I never tried the option you mentioned, if I can I will give it a try also.

The problem with access rights to mounted directory has been resolved this way:

sudo bash -c 'cat << EOF >> /etc/wsl.conf
[automount]
enabled=true
options=metadata,uid=1000,gid=1000,umask=022
EOF'

After this, VAGRANT_HOME can be moved. But then vagrant can only be run from under WSL, otherwise access rights problems will appear Windows.

[entrptaher]

After hours of issues. I found a solution for me. Keeping it here for future reference.

Most tutorials tells us to use a vagrantfile which binds the port 22 to 2222 of the 127.0.0.1, which causes issues with the firewall and so on. When I removed the restriction to the host ip (keeping it empty), it instantly got connected for me.

Another solution is to use something like tailscale/zerotier instead of localhost. Which also worked great so far.