Help with Linux Mint - this is why I do not like Linux!

Don’t feel bad, you are not the only one that shares that feeling. Those two videos are a bit dated, but from that time more or less no progress has been made in the directions that were outlined at that time. By the way, those two videos are from Linus Torvalds, so I would thing good enough authority with respect to Linux.

The title says everything here - “why desktop Linux sucks”:

And this is about how easy switching between distros is - “why Linus doesn’t use Ubuntu / Debian”:

What he described there is basically what you are facing right now. Both Max and Harold are using just one Linux distro, or rather one version of that distro (with unknown updates applied). Once thigs work on their primary boxes, the code moves to production, as Max is just too small to have a decent QA and Chia doesn’t have QA by design (as they call it beta, and hope that us, farmers will do the testing). Due to shortcomings that Linus explained, neither party tests with other releases / distros, as the number of different combinations more or less grows exponentially. Therefore, any time someone tries to deviate from that original distro/version/HW, things may happen. In this case (plotter), nVidia’s drivers are mostly always at fault.

Another part of the problem is that neither Max, nor Harold are Win developers, so things sometimes go ugly there as well - like MM v1 CPU plotter that has problems with dual CPU boxes (e.g., Max is using cross-platform libs, but those are mostly optimized for Linux, but the only goal on Win is to be able to run the binary). I think that Harold is doing a better job in the final Win build, but maybe I am mistaken here.

I am using headless CentOS/Rocky Linux more or less day in and out and initially tried to use Fedora Desktop (all those are RHELs) for MMX plotter but failed miserably (couldn’t get MMX plotter installed at all). After that, I switched to the latest Ubuntu LTS version, and it installed OK, but after an hour or so of plotting, the box was always hosed (reboot needed). I have tried to find the culprit, but more or less was on my own. Then I noticed that Max was using the previous Ubuntu LTS version, so I installed that one, and things started to work kind of right away. Most likely, due to my H/W (Dell t7610), MMX plotter was crapping on the final file transfers still making the OS unstable after about 5 hours, so I had to write my own scripts to do those xfrs (plotter dumps plots to NVMe, and my scripts move those plots to HDs - more or less it can run 24/7).

1 Like

I can confirm your link has the correct steps for a clean install of the nvidia 535 drivers. After removing all cuda and nvidia packages via methods mentioned above, I rebooted and Ubuntu was using the base generic graphics driver. I used the steps below from your link in this post to install only the nvidia 535 driver (no cuda toolkit) and I was able to run a test of 5 cuda plots as well as start the Beta 1 farmer/harvester without issue/clean logs. Next I need to upgrade that setup to RC1 but I’m not expecting any driver issues.

These are the steps to use from your link:
–add the ppa repository if you haven’t already added it before
sudo add-apt-repository ppa:graphics-drivers/ppa

sudo apt-get update

sudo apt-get install nvidia-driver-535

The only bump I did hit was within the ‘NVIDIA X Server Settings’ control panel, where you can manually set the fan speed (I like to crank it before plotting)… It opened and looked fine but I was getting an error of “Failed to set new Fan Speed!” when moving the slider and clicking ‘Apply’ after the new driver. This was easily corrected with the steps from the 2nd post in the following link (reboot required):

1 Like

Yeah Linux upgrades are a pain. I recently updated my Ubuntu machine and it messed up drive mounts. Some of the disk mount names were reset. I thought some of my drives failed as the number of plots decreased. Turns out I needed to add directory again and remove the old ones in Chia GUI to point to the new mount names! Crazy.

Unfortunately I’m still going round in circles, really thought this was going to work tonight.

Thinking it may be the missing nvidia-settings I tried to install that, but no joy there either.

This is exactly the same problems I faced previously.

The second screenshot looks to be a base Gnome package. Here’s one page on it:
https://launchpad.net/ubuntu/+source/gdk-pixbuf

What do you see when you look for the package from the second screenshot:
dpkg -l | grep -i libgdk-pixbuf

I’m seeing the following:
$ dpkg -l | grep -i libgdk-pixbuf
ii libgdk-pixbuf-2.0-0:amd64 2.42.8+dfsg-1ubuntu0.2 amd64 GDK Pixbuf library
ii libgdk-pixbuf2.0-bin 2.42.8+dfsg-1ubuntu0.2 amd64 GDK Pixbuf library (thumbnailer)
ii libgdk-pixbuf2.0-common 2.42.8+dfsg-1ubuntu0.2 all GDK Pixbuf library - data files

You may try following/working that chain of errors with the technique mentioned at this link and once you resolve the system packages, all the needed nvidia packages should install with the driver:

If you don’t care for the data/other configurations on the machine, as Voodoo mentioned, you may consider wiping and starting with a fresh install.

This is what I get.

I think I’m going to do a fresh install and see how that goes. Thanks for your help, I’ll update how I get on later.

Just to update, another evening lost.

Installed Mint 21.2, rebooted and got.

error: file '/grub/i386-pc/normal.mod' not found.
grub rescue>

Tried another reinstall, as suggested on a forum, this time on reboot, no boot device found.

Edit: This last error is something to do with UEFI, if I go into the boot menu, and select Ubuntu under UEFI it boots.

I would hold my nose and install Ubuntu on a clean SSD with the version that Chia is using (if you don’t like Ubuntu or are unfamiliar with that command dialect). You just need a 128 GB SSD to run OS, so any cheap will do. (Plotter / farmer will never be touching that SSD.) Bear in mind that you are still fighting just the installation, and the next slew of issues may come while trying to plot if things go further south.

You are running into package hell that Linus described in that video. Solutions that look like are addressing those issues you have but are coming from Ubuntu side may at the end of the day be still only relevant to Ubuntu (if Mint is out of sync, and looking at the output you and @Whatshisname provided so far you don’t have the same packages / versions).

By the way, the latest Rocky Linux - 9.2 (as such all RHELs) kind of dropped support for my Dell. I mean, I cannot use UEFI anymore, BIOS needs to be in Legacy mode (everything else works fine). Just switching BIOS from UEFI to Legacy was not fixing it, I had to run a clean install with SSD being in Legacy mode right away. That happened during a standard update from RL 9.1 to 9.2, and there was no warning no nothing - standard Linux user level error handling. Maybe you are facing a similar issue with the latest update.

Ubuntu was the first version of Linux I tried, but I had no end of problems, hence I tried Mint, and that worked fine until I need to get Cuda working.

Previously I’d installed every OS in Legacy mode on this Dell, but tonight it would not boot from the USB stick unless I selected the option under UEFI, kind of weird not had that issue before.

Anyway its currently working, and just setting up RDP access, then I’ll see how it goes.

Seems we finally have success:

Two Bladebit processes running, and creating plots.

3080 = Completed Plot 2 in 157.73 seconds ( 2.63 minutes ) Windows 300 seconds
P4 = Completed Plot 1 in 461.92 seconds ( 7.70 minutes ) Windows 600 seconds

That equates to 56TB a day.

3 Likes

Great to hear! Those are some huge time differences. Especially for the 3080.

So what version of Linux did you settle with???

Linux Mint 21.2 - its based on Ubuntu.

Regarding the times I did use what I believe to be the latest plotter, which at some point I do need to test on Windows, as there may have been some speed improvements, but I suspect its down to Linux.

Being a dual CPU system I may see some decrease in plotting times if I matched the CPU (Numa) used to the GPU, this might be why there isn’t such a significant decrease in time on the P4 compared to the 3080.

Got a weird one, when I copy and paste in Linux, I get extra characters even if the copy was from within Linux, but not every time, any ideas?

^[[200~--no-cuda~

Which should be

--no-cuda

So why the weird characters?

Edit:

Found this [SOLVED] Weird characters while pasting in terminal / Newbie Corner / Arch Linux Forums

Seems to be bracketed paste mode whatever that is, but entering printf “\e[?2004l” from the CLI is supposed to fit it. Just plain weird!

1 Like

lol… I get this all the time too… Due to years in the MS IT ecosystem as a career and using Windows keyboard commands constantly while working, I instinctively use ‘ctrl-v’ to paste in Linux terminal before remembering to add ‘shift’ to the combination. In Ubuntu’s terminal, the first ctrl-v won’t print/paste anything but it will without a doubt add the ^[[200~ and trailing ~ to whatever’s on your clipboard that’ll be visible when you paste. I’ll then start using ‘ctrl+shift+arrow’ key to highlight the text intending to delete it but start getting ‘DDDD’ added to the end before catching myself lol… but yeah, strange results.

I only get this in the terminal though, in the text editor or nano (and maybe other programs), ctrl-v and other keyboard navigation seems to work for me.

2 Likes

It was in terminal, and yes I often forget to add shift, but it never happened in the previous version of Mint, I’ll see how it goes.

I was most amazed when I hit automatically hit F2 to rename a file and it actually worked!

I RDP in to Mint, so working from Windows, often copy and paste between the two

@DigitalSpaceport take a read of this thread.

So I created a raid 0 array of two drives in Linux.

But cleaning up some left over mount point folders from the previous two drives I accidently deleted the mount folder for it. To fix it I removed the array using mdadm, then recreated it.

All went well, except when I reboot it doesn’t auto mount, and I now can’t mount it either, I’m sure the first time round it did mount correctly after a reboot, but I might be imagining that!

What I get now.

and

and

Any idea’s please?
Also which file system would be quickest?

Sometimes the name changes after first reboot md0 ↔ md127 ? …no not that.

Did you make a filesystem on it?
mkfs.ext4 /dev/md0

I formatted it from within Disks, then used it, then rebooted, so it was working, I created some plots to see how much quicker the write time was.

At work at the moment, so can’t try anything.