For a while now, I’ve had a spare 512GB NvMe M2 drive. I stuck it in a USB3+3.1 elcosure and I haven’t found a solid plan for it yet. I would like to turn it into a “Rotating” linux stick.
Basically, In this stick I wish to set up a common
/home and /EFI
and start with LEAP 15.1 root, LEAP 15.2 root and Tumbleweed root, all sharing the same home.
The idea is to basically rotate from current, future and rolling distributions to test and make sure everything is up to date and compatible with all (or most) of my machines then clone the best combination of / and /home into my machines.
I have done something like this before, where I had LEAP 42.3, 15.0 and 15.1 installed into a same SSD using the same /home HDD then I wiped the partitions in the SSD I no longer use.
Are there any hints or warning that I should watch out for? The barebone will look something like:
75GB LEAP 15.1 /
75GB LEAP 15.2 /
75GB Tumbleweed /
75GB CentOS/Ununtu/Fedora… etc. /
I don’t know that I would put any boot files or partitions on removable media for “permanent” OS.
I’d also be wary about hidden files specific to how a Desktop is configured in different OS… If DE versions are same or close enough maybe there wouldn’t be an issue but I can imagine all sorts of problems if there are different DE and widely different versions of the same DE.
I’d recommend instead that boot files and partitions be placed on an internal, fixed drive
/home would be mounted specific to the running DE
but you can create common mounts for directories that hold only data files like ~/Documents and ~/Downloads.
Of course, this assumes you’re running only one OS at a time, if for instance you were to run virtualization and Guests and HostOS might access the same files and directories, you could run into contention issues and lost data.
TSU, In terms of installing everything in the external SSD is not to operating the computer from it for long term, it is to check and make everything compatible with the machine then to move the partition into an internal drive.
I think what you are suggesting may be more suitable in terms of partitioning. I should have “Desktop, Documents, Downloads…” personal data partition but embed /home with the matching DE root.
I guess for my purpose then is that would it be feasible to re-partition during cloning and separte /home from / after installation? In past, I did move my /home out of single root partition and it worked. I guess I will need to keep some spare space in the SSD then to do this cleanly on a machine.
I use a spare 128GB NvMe drive in an USB3+3.1 enclosure as rescue system. It is setup with openSUSE Tumbleweed and GPT-partitions for /EFI, /, /home, a partition for GRUB2 to be able to BIOS boot and an additional NTFS-Partition which allows me to move data from MS Windows systems to my Linux boxes. GRUB2 is configured to be able to boot on UEFI and BIOS systems.
Basically it starts up and can be used on any of my machines however graphics is always a topic to keep an eye on e.g. machines with NVIDIA or even NVIDIA Optimus graphics do always need special measures to boot and to run happily.
If you want to use several OSs in parallel with just one /home you should keep in mind that configuration data is written to a users HOME-directory (e.g. by the desktop environment). Different OSs might use the same packages but with different versions which in turn probably need slightly different configuration data/files. Using different USERIDs in every OS might help to avoid problems.
Now, you’re giving me a great idea! A single /home partition with individual users dedeicated to a specific desktop envrionment. I think I am going to leave Nvidia cards alone and try to work with the most compatible nouveau at the time. Basically the machines I use are:
Lenovo T480 (Optimus)
Lenovo T430 (Integrated)
Lenovo W530 (Optimus)
Toughbook CF19 (Integrated)
Alienware 15 2015 (Optimus)
Custom desktop with Nvidia 1070Ti
I am getting a curious thought now… What would happen if I have bumblebee+Nvidia setup for an OS, and boot from a laptop with only integrated graphics, and what happens if I just install Nvidia for the OS then boot from a laptop with only an integrated graphics.
I guess in a setup like yours graphics is the most tricky part. I decided to go for the “least common denominator”. All my machines have Intel graphics (Optimus = Intel + NVIDIA) so i configured my rescue system to enforce the use of the intel graphic (which is supported by a kernel driver).
I’ve done that regularly in the past but I had a weird variation of issues. For Lenovo T430 with Integrated HD 4000 + Nvidia NVS5400m optimus, it turns out the VGA is hard-wired to be ported through NVS5400m only and Display port is hard-wired to port through integrated GPU only. So it always gave me trouble:
Only Intel = no VGA access
Only use Nvidia = no Display port access
Use Nouveau = access to VGA + Display port, but weird “double mouse pointer” issue. Basically everything was rendered twice and when I connect to an external monitor, I will get a second mouse pointer in the centre of the monitor which I cannot move.
Bumblebee only started working properly in Leap 15.0.
Similar issues with W530. I’ve since then opted for bumblebee+Nvidia driver to make everything work on optimus capable laptops.
The external ports (HDMI, …) of my laptop are only accessible when the NVIDIA-card is active. I have seen similar behavior on quite a few other laptops with NVIDIA-Optimus-configuration.
I stopped using bumblebee or suse-prime because my demands are not very high as far as graphic capabilities are concerned. And the Intel graphic runs fine on any of my machines. So most of the time i stick to the Intel cards.
On rare occasions when i have to use the HDMI-port of my laptop (e.g. doing presentations abroad) i just switch the NVIDIA-card on, load nouveau and when i have finished using the HDMI-port i just reboot.
I think the very best and real purpose of this stick is to experiment, so I will try the bumblebee, Nvidia and etc. on non-optimus laptops. I’m curious though, how did you get your laptops to only work off of intel card? Do you turn off the Nvidia card from BIOS, or do you do something equivalent setting nomodeset in grub/ blacklisting nouveau or etc.?
In my laptops UEFI i can select whether both graphic cards will be enabled (this setting is called MSHYBRID) or only the NVIDIA card (this setting is called DISCRET).
Because i want to use the Intel card i have to set my UEFI to MSHYBRID so that both cards are enabled.
There is no need to do anything in GRUB but i have to block any NVIDIA driver (nouveau, nvidia, …) from being loaded automatically on system start up. Then i have a xorg.conf file which tells X which card (the Intel card) and which driver (the driver build into the kernel) to use.
To avoid that the NVIDIA card drains my battery i have installed bbswitch which on system start up switches the NVIDIA card off completely and switches it back on when the system is shut down. But the sole purpose of bbswitch is to save battery. The setup would work without bbswitch as well.
When i want to use my HDMI-port (wired to the NVIDIA card) i switch the NVIDIA card on (by software), load the nouveau driver manually and restart the X-server. In order to go back to “normal” i reboot.
I don’t like to use the proprietary NVIDIA driver because it replaces parts of the X-stack with own modules so that a switch between cards becomes a bit tricky (Nevertheless its possible).