Attachext USB to WSL - easy way - no extra software needed

Hi folks
Following on from my previous post of seamless integration of GUI apps with Windows one might for all sorts of reasons want to share Linux disks over your network and currently accessing the distro directly from anywhere outside the Host machine is well nigh impossible currently.

So easy "Dummies" way !!

1) ensure you've followed instructions in my previous post about running GUI apps on WSL and integrating to windows

2) start a powershell command (windows) in admin mode and enter as shown



In this case ignore the error message -- that's because the file system on the disk is XFS whereas Windows wsl assumes ext4 -- doesn't matter.


3) simply from your desktop run a linux terminal --in my case konsole. No need tp start the wsl - ist time the console starts it will take a few seconds as behind the scenes the "VM" framework on windows is initialised - then it all works super fast.

4) run the lsblk and mount commands as shown -- I've created a user accessible mount point in /mnt/DV0 and chown -R userid /mnt/DV0 so user has r/w access to it.




Now access from Windows : just file explorer : go to the linux directory



This disk can be accessed also from remote NAS servers simply by browsing that directory on the Windows Machine !!! Caveat -- though ONLY BY RDP to the host Windows machine -- currently there's no way of accessing the WSL directly from outside it's HOST machine.

"Seemples" -- Best of both worlds -- Linux decent set of tools with decent streaming and file manager (I hate File explorer) and your favourite Windows applications --- no need for messing about with clunky VM's

What is the read/write speed that can be achieved?

vett93 said:

What is the read/write speed that can be achieved?

Click to expand...

Who can possibly answer this as it depends on lots of factors e.g. drive type, pc speed etc.

cereberus said:

Who can possibly answer this as it depends on lots of factors e.g. drive type, pc speed etc.

Click to expand...

One way to answer the question is using relative speed. For example, one can publish two sets of read/write speeds: one for Windows and one for WSL2.

I run tensorflow on WSL2 with Ubuntu 20.04 LTS. My desktop PC has 2 TB of SSD storage. I am trying to figure out how I should handle large datasets. WSL2 documentation states that it does not support USB drives. If OP can make it work, more power to him.

Hi there
The external USB 3 drive seems to perform more than adeqately -- I haven't measured it but I'd assume you'd probably use this in cases where you want to attach Linux formatted drives.

Note the main caveat so far with the WSL is that you can't access either it or any drives connected to it from OUTSIDE its host unless you logon to the Windows Host itself probably via RDP so this rather limits the usefulness of using a command line version of WSL for things like file serving

The problem is that the virtual WSL machine has a NAT address e.g 172.x.x.x while for proper network connectivity on your lan you need bridged mode with addresses 192.x.x.x for example. Note also that the VLAN doesn't support IPV6 although for a home network that shouldn't matter.

So probably Ms needs to improve this feature a bit more -- but this might be a false hope as if it becomes too good then it will be competing with its own HYPER-V system which still could do with a tweak or two for running non Windows VM's or pre win 7 OS'es.

Cheers
jimbo

jimbo45 said:

.....

So probably Ms needs to improve this feature a bit more -- but this might be a false hope as if it becomes too good then it will be competing with its own HYPER-V system which still could do with a tweak or two for running non Windows VM's or pre win 7 OS'es.

Cheers
jimbo

Click to expand...

WSL is a top feature that made me drop my Mac Pro after being 10+ years of Mac user. If Microsoft can further improve the feature, I'd imagine more people will switch over. My main use of this Windows 11 desktop is CUDA accelerated deep learning (e.g., Google's tensorflow stack) and machine learning. I need to have large store space for image files to train my ML models.

Microsoft is in an unique position for it. Hopefully, they don't screw it up.