nvidia-persistenced failed to initialize

deepforest · 2023-08-17 23:28:08

Ceres
installed

apt install nvidia-legacy-340xx-driver

freeartist-devuan@home:~$ nvidia-persistenced
nvidia-persistenced failed to initialize. Check syslog for more details.
freeartist-devuan@home:~$

its Devuan bug?
https://bugs.devuan.org/cgi/bugreport.cgi?bug=700

freeartist-devuan@home:~$ sudo /etc/init.d/nvidia-persistenced stop
Stopping NVIDIA Persistence Daemon
freeartist-devuan@home:~$ sudo apt install 
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
0 upgraded, 0 newly installed, 0 to remove and 2 not upgraded.
freeartist-devuan@home:~$ nvidia-persistenced
nvidia-persistenced failed to initialize. Check syslog for more details.
freeartist-devuan@home:~$

Last edited by deepforest (2023-08-17 23:32:56)

delgado · 2023-08-18 12:28:51

I don't think it is "bug 700".
700 does not harm, it is an unhappy apt, complaining about an unfinished package installation, which functionally succeeded. Anything else is in order.
The error message looks like something went wrong with nvidia-persistenced.

What about the 2 not upgraded packages?
Just hoping that apt upgrade or dist-upgrade might fix it.
And de-installation is an option too.

Marjorie · 2023-08-18 14:33:55

nvidia-persistenced has no real purpose unless you want to use CUDA.

If you don't need CUDA then you can just delete it, nvidia works fine without it.

There was (is still?) a bug in the code that dates back to 2014 that can lead to it not initialising properly.

See my post dating back to April 2020.

https://dev1galaxy.org/viewtopic.php?pid=20903#p20903

chris2be8 · 2023-08-18 16:11:24

Marjorie wrote:

nvidia-persistenced has no real purpose unless you want to use CUDA.

It's not essential for CUDA either. Both the CUDA apps I use (gmp-ecm and msieve polynomial selection) work without it running.

Marjorie · 2023-08-18 17:03:23

Indeed. It may however be helpful (albeit not essential), but it seems only on machines/servers that don't run X persistently.

"Under Linux systems where X runs by default on the target GPU the kernel mode driver will generally be initalized and kept alive from machine startup to shutdown, courtesy of the X process. On headless systems or situations where no long-lived X-like client maintains a handle to the target GPU, the kernel mode driver will initilize and deinitialize the target GPU each time a target GPU application starts and stops. In HPC environments this situation is quite common. Since it is often desirable to keep the GPU initialized in these cases, .. "

https://docs.nvidia.com/deploy/driver-p … index.html

Last edited by Marjorie (2023-08-18 17:04:14)

delgado · 2023-08-18 18:12:23

Thanks for clarifying the purpose.

stopAI · 2023-08-19 11:27:05

I found a much simpler way to avoid this problem with

nvidia-persistenced

Since this package is not needed by many, the easy way is as follows.

1. Install the required nvidia driver (in my case, the driver I need is nvidia-tesla-470-driver, because my video card does not support a higher version driver), using the command

sudo apt install --no-install-recommends

The list of recommended packages includes the

nvidia-settings-tesla-470

package, which can be installed with the simple command

sudo apt install nvidia-settings-tesla-470

Using this method avoids installing the nvidia-persistenced package.

deepforest · 2023-08-21 18:25:12

yes and no
semi-solution its "no package - no problem"
but i want go deeper)
i want to know why exactly this package have problems at Deuvan?

Marjorie · 2023-08-21 18:45:59

I thought I'd explained this in my linked 2020 post at https://dev1galaxy.org/viewtopic.php?pid=20903#p20903

The nvidia supplied init (/etc/init.d/nvidia-persistenced) file was borked.
As most other distributions (such as vanilla debian) use a systemd .service file, not a sysvinit init file to initialise they wouldn't get this problem.

I think I've identified the problem, though I'm unclear why it got triggered during an install. The problem occurs when there is an existing nvidia-persistence running on the system. When this is the case starting nvidia-persistenced fails with the above error message (as it can't get a lock on the PID file).
I've therefore added/modified a few line of code to correct this: if a running nvidia-persistenced exist it stops it and then starts a new instance. This ensures that nvidia-persistenced starts in its enabled state as that is the default.

This was 4 years ago (and 10 years since that file was created) so maybe it's now something else.

Last edited by Marjorie (2023-08-21 18:48:24)

deepforest · 2023-08-22 01:20:45

Marjorie Thanks for interesting information!
What do those nvidia script? And will it work at OpenRC?

Last edited by deepforest (2023-08-22 01:26:13)

Marjorie · 2023-08-22 21:10:32

Not sure about OpenRC, the nvidia supplied nvidia-persistenced is a standard sysVinit script to go in /etc/init.d and I use sysVinit.
In my linked post I include a copy of my patched version.
https://dev1galaxy.org/viewtopic.php?pid=20903#p20903
If you replace the supplied /etc/init.d/nvidia-persistenced with this it should just work and not throw up any errors.
It worked fine on my PC. This PC has AMD not Nvidia graphics but my Nvidia card was still working on another machine last time
I checked.
O course they may have changed the original since 2020, but I doubt it - my 2020 version was originally written in 2013!

deepforest · 2023-08-23 19:32:46

And why this solution works good at Deadalus(openrc) and nvidia-persistenced too
https://www.if-not-true-then-false.com/ … dia-guide/
but Debian Ceres 340.108 driver have problem with and nvidia-persistenced?

Marjorie · 2023-08-24 14:51:30

Not sure what you mean by Debian Ceres 340.108 driver.
Ceres is the Devuan equivalent of Debian Sid.
Playing around with bleeding edge stuff (Sid/Ceres) is asking for trouble.
I would only suggest doing it if you have such new hardware that you can't get it working any other way.
And isn't Nvidia 340.108 a legacy driver? Not exactly new hardware.

deepforest · 2023-08-25 18:59:10

cuz only sid/ceres support 340.108 driver

kapqa · 2024-01-04 19:26:53

hello ,

i have this issue "persistenced" on upgrade from ascii to beowulf with FX Quadro 770m<
the solution from "StopAI" worked perfectly to get me back into Desktop with Graphics Accelerations.

what i did>

1. uninstall official Nvidia/driver 340.108 in recovery.
2. install as suggested "sudo apt install --no-install-recommends nvidia-legacy-340xx-driver"
3. reboot, voil'a

now it is understanding that the next step, if wanted to upgrade further, would be to jump from Beowulf stright to Ceres, or uninstall again, and regain X without Nvidia, and go from there to upgrade path ZChimaera Daedalus Ceresa.

EDIT> i tried first with the official driver package from Nvidia, since this worked in the previous iterations of Debian/Devuan, but would no longer function on Devuan 3.0 for some reason or another.

Last edited by kapqa (2024-01-04 19:30:07)

stopAI · 2024-01-05 14:19:59

Hello.
I am glad that the method I suggested helped you.

For upgrades, you should go from Chimaera to Daedalus and Ceres. If my opinion is important to you, then stick with open source driver. You will have less problems during the upgrades.

GlennW · 2024-01-05 23:40:52

Hi. Kapqa, will you be upgrading each release before you move to the next version?

I think the Open Source nVidia module is called nouveau.

imo, it's only the installation of the Official-nVidia (.run) package that is a bit tricky sometimes.

DKMS can carry quite well, but I stopped using dkms a while back. Because I felt I didn't need all the dev stuff just for that.

I think if you upgrade each release as you go, you'll have a great weekend.

Cheers!

btw, my install steps are...
1. OS and GUI default, nouveau module gets loaded and configed
2. install dev packages and requirements for the required build environment, including any updates, reboot as needed.
3. copy text files from backup, grub, sysctl.conf, /etc/modprobe.d (nouveau blacklist entry)
4. reboot and at the grub prompt, to recovery-mode (use e to edit) adding "vga=794 modeset.nouveau=0 nokmsboot" to the last "linux=" line
5. install nvidia- .run package. reboot.
:-)

I hope this is helpfull

Last edited by GlennW (2024-01-05 23:42:42)

aluma · 2024-01-06 09:47:13

@Marjorie

I completely agree!

After adding the Ceres repository, Apt removed some of the DE packages and no dist-upgrade helped. The problem is known, newer versions of packages break dependencies.

The equipment may be old, but my i7 laptop has two identical GT650Ms installed, which do not work with the tesla-470 driver suggested by nvidia-detect. (Debian's list of supported card models has errors).
At the same time, the 2D speed of the Nvidia driver is 10...15 times higher than with nouveau.

It's a pity that Nvidia driver support for them has been discontinued.

In my opinion, this decision by Debian programmers simply repels users. I just installed opensuse on my laptop, where the Nvidia driver is in the repository without all the Debian dances with a tambourine.

GlennW · 2024-01-06 21:37:25

When I last did a clean install of Daedalus, last week I found I needed to install seatd

(I use sddm (X as root), and Plasma).

But upgrading from Chimaera did not have the same fault.

Nietz · 2024-01-07 00:35:52

Hello @Deepforest

I encountered this issue as well. It became an opportunity to reevaluate why I was using their driver. From https://download.nvidia.com/XFree86/Lin … nced.htmln:
"
Background

A Linux daemon utility, nvidia-persistenced, addresses an undesirable side effect of the NVIDIA kernel driver behavior in certain computing environments. Whenever the NVIDIA device resources are no longer in use, the NVIDIA kernel driver will tear down the device state. Normally, this is the intended behavior of the device driver, but for some applications, the latencies incurred by repetitive device initialization can significantly impact performance.

To avoid this behavior, nvidia-persistenced provides a configuration option called "persistence mode" that can be set by NVIDIA management software, such as nvidia-smi. When persistence mode is enabled, the daemon holds the NVIDIA character device files open, preventing the NVIDIA kernel driver from tearing down device state when no other process is using the device. This utility does not actually use any device resources itself - it will simply sleep while maintaining a reference to the NVIDIA device state.
"
and
"
The daemon utility nvidia-persistenced is installed by the NVIDIA Linux GPU driver installer, but it is not installed to run on system startup. Due to the wide variety of init systems used by the various Linux distributions that the NVIDIA Linux GPU driver supports, we request that package maintainers for those distributions provide the packaging necessary to integrate well with their platform.

NVIDIA provides sample init scripts for some common init systems in /usr/share/doc/NVIDIA_GLX-1.0/sample/nvidia-persistenced-init.tar.bz2 to aid in installation of the utility.

nvidia-persistenced is intended to be run as a daemon from system initialization, and is generally designed as a tool for compute-only platforms where the NVIDIA device is not used to display a graphical user interface. As such, depending on how your package is typically used, it may not be necessary to install the daemon to run on system initialization.
"

I realize this functionality is useful, but I think that it is intended more for ultra low latency "compute-only" platforms with no GUI.

Thank you for bringing this into the forum.

The officially official Devuan Forum!

#1 2023-08-17 23:28:08

nvidia-persistenced failed to initialize

#2 2023-08-18 12:28:51

Re: nvidia-persistenced failed to initialize

#3 2023-08-18 14:33:55

Re: nvidia-persistenced failed to initialize

#4 2023-08-18 16:11:24

Re: nvidia-persistenced failed to initialize

#5 2023-08-18 17:03:23

Re: nvidia-persistenced failed to initialize

#6 2023-08-18 18:12:23

Re: nvidia-persistenced failed to initialize

#7 2023-08-19 11:27:05

Re: nvidia-persistenced failed to initialize

#8 2023-08-21 18:25:12

Re: nvidia-persistenced failed to initialize

#9 2023-08-21 18:45:59

Re: nvidia-persistenced failed to initialize

#10 2023-08-22 01:20:45

Re: nvidia-persistenced failed to initialize

#11 2023-08-22 21:10:32

Re: nvidia-persistenced failed to initialize

#12 2023-08-23 19:32:46

Re: nvidia-persistenced failed to initialize

#13 2023-08-24 14:51:30

Re: nvidia-persistenced failed to initialize

#14 2023-08-25 18:59:10

Re: nvidia-persistenced failed to initialize

#15 2024-01-04 19:26:53

Re: nvidia-persistenced failed to initialize

#16 2024-01-05 14:19:59

Re: nvidia-persistenced failed to initialize

#17 2024-01-05 23:40:52

Re: nvidia-persistenced failed to initialize

#18 2024-01-06 09:47:13

Re: nvidia-persistenced failed to initialize

#19 2024-01-06 21:37:25

Re: nvidia-persistenced failed to initialize

#20 2024-01-07 00:35:52

Re: nvidia-persistenced failed to initialize

Board footer