I was tempted by Clear Linux, because of all the promises made and the benchmarks I saw. It’s linux, so I expected some suffering, but at least some performance gain in compensation. My thinking was that these guys at Intel are a bunch of highly professional people, they may be able to get this thing right.
I wanted to set up a virtualized environment with a GUI and it seemd like Clear Linux was a perfect fit, although I know that Linux is always hiding some surprises for you.
I would like to run it on a small machine with an Atom J1900 (has VT-x) and a 120 GB SSD, that has previously run Windows Server 2012R2 with Hyper-V and up to 10 virtual machines without a hiccup (some on an external SATA drive).
So I start up installing Clear Linux 32850 and of course it’s not trivial to get a simple thing like the installation running. So I have to find the right flashing tool (Etcher worked), switch the UEFI settings in the BIOS, play around with the settings for the USB storage and stuff to finally get into a prompt, where I have to start the installation myself. Why? What decade are we in now? But well aware that the biggest surprises are yet to come, I wander on in my journey.
Fine I setup a password for the root and enter the installation process. So the first thing I stumble upon is a hopelessly outdated disk partitioning, I would date it 1984 by aperance.
A slight hint in the installation process indicates that I might need to name the partitions in some certain way, for Clear Linux to recognize them, how is not specified, until you do something and get a warning. Nobody considered to include that information in the installation process, you have to go and ask Google. Why?
Searching for help online I find that you can even add “_F” to the naming of the partition, to automatically format it, which doesn’t work if the drive is not already formatted with the required format. To be sure, the _F has to be added differently depending on if it’s a SWAP drive or BOOT drive. WHY?
So the partitioning ask me about filesystem and in the Linux-world-interpretation-of-user-friendliness it suggests 8300… And no specifications anywhere about any other possible options like EF00 or 8200. Because?
After realizing that this is definitly not plug-n-play, I wonder where the option to format the drive manually is? Didn’t find any… eeeeh? So I leave the installation and format the relevant boot drive from the command line and restart the installation.
Now I’m finally allowed to proceed with the installation and it seems like I have a successful install until after a reboot. For some reason I can’t login with the root password, in disbelief I try to enter it close to 10 times, also hitting caps-lock. Fortunately I made an admin user with the same password that successfully logs in. Then I try changing the root password and it accepts the password that it didn’t accept at the login. WTF?
Then I realize that the specified minimum for the disk size, doesn’t allow for a GUI. So I decided that I wanted to do a new fresh install, as I didn’t want to insult the Linux installation with resizing the partition. So I made a fresh install, only increasing the disk size of the root partition and adding the desktop UI bundle. Aparrently the added UI bundle failed to install and the installation process was abruptly halted with a message about what information I had to send to the developers.
OK, so trying again with the exact same settings and voila! Installations completes and boot fails…
I don’t understand why anybody chooses to run Windows, this is so much more fun! But I’m not sure I would trust my data to somebody who can’t even make a reliable installation process.