SOLVED: Beowulf: quite good, but ......

berni51 · 2020-06-22 11:12:08

Firstly: I installed Beowulf on 3 machines without any problem: One update from ASCII on a Thinkpad, two new installations on a AMD machine and another Thinkpad. Absolutely no problems till now.

But there is my Intel NUC! Until beginning of last week the NUC runs ASCII - perfect, everything OK during the last 2 years.
Then I ran an upgrade to Beowulf. Update went fine, machine runs good. Except the shutdown, which made what it wants. Sometimes the machine was shutdown correctly, sometimes it ended in sleeping, sometimes it crashes.

OK, I then made a complete new installation. As before, the installation went well, everything seemed to work. However the strange shutdown behaviour was as before.
After some hours working with the machine the problems became more and more:
- black screen and dead system after longer pauses and without doing anything
- systems hangs during start: Loginmanager appeared, I entered name, but when entering the password, it crashes
- during work with any application again crashes.
Each crash came suddenly with a black screen and a dead system. No hint in the log files, but the file file system must be repaired after each crash.

After one day with these crashes I was sure that I have a hardware problem, So I changed, one after the other, the following
- RAM
- SSD
- all connected USB components

None of this helped! Sometimes the machine ran several hours, sometimes it crashed every few minutes.

Before changing the mainboard I tried 2 other OS: Debian Buster and Archlinux. Debai ran one full day without any crash, archlinux now the 3rd day.

So meanwhile I am convinced that Beowulf is incompatible with my NUC. Remember that ASCI did the job for a long time. I have no more ideas what to do. Maybe I should try Chimaera?
Any hints are welcome, because I like my NUC as well as Beowulf.
Further it may be dangerous for Devuan cause archlinux is quite nice, even it is extremely full of systemd. But it runs perfect after some work - I have forgotten how hard is is, to built a Linux from scratch.

Thats what I want to tell you today.

Regards
Berni

... and sorry for the long poting :=(

Last edited by berni51 (2020-06-27 16:11:26)

HevyDevy · 2020-06-22 13:18:30

I run a NUC okay with devuan, what is the model you are using? Post up the system specifications of it.

Marjorie · 2020-06-22 14:22:19

I also have a NUC running Beowulf without issue, albeit it's my mailserver, running headless and rarely rebooted: the last time was was for a kernel image upgrade to 4.19.0.9 (which was unproblematic).

Mine's a very basic NUC5CPYH 85254, 4GB, 120 GB SSD, mbr, LVM apart from /boot and I run it in legacy bios mode, not EFI.

berni51 · 2020-06-22 16:59:26

Sorry for missing data. Mine is a NUC 6CAYB with Intel 500 graphics, 8 GB of RAM and 512 GB ssd, EFI boot.

Marjorie · 2020-06-22 19:22:06

OK, so not dissimilar to mine: your is Celeron 6 gen 4 core vs the 5 gen 2 core on mine so it's not as if it has recent/high spec. hardware that could be causing problems.

Your'e operating under EFI, I'm not. However I'm assuming your previous ascii install was EFI too?

I using sysVinit. There have been problems reported on Beowulf with shutdown with OpenRC.

Are you certain your Beowulf install iso (btw which one did you use?) hasn't become corrupted. Have you checked it?

rolfie · 2020-06-22 19:29:31

Are you using LVM? I am using openrc on encrypted systems with LVM, no issues with shutdown except that you need a patch to speed the closing of the volumes and the LVM: https://dev1galaxy.org/viewtopic.php?id=2690&p=2 #34.

Read the thread if you experience similar problems. Maybe it helps.

rolfie

berni51 · 2020-06-23 07:31:01

@Marjorie: the previous ASCII installation ran with legacy BIOS. I am using Openrc for my Beowulf. The stick with the Beowulf iso image was tested and I used it for all my installation. As mentioned before, three of them are still running without any problems.

@rolfie: No, I don't use LVM. So I have 1 possible faults less.

Yesterday I downloaded another Beowulf image (again desktop-live), installed it on a brandnew ssd, added some essential programs for me, played a little bit backgammon - and after 2 hours I clicked on the wrong pixel and there was it again: black screen, system dead, hardware still alive.

Beowulf really makes me crazy

PedroReina · 2020-06-23 09:02:25

I would try with a different kernel on your Beowulf install. Even with a strimlined custom kernel.

HevyDevy · 2020-06-23 09:21:36

by black screen is it completely black or is there text on screen as well like a kernel crash?

I have the same nuc as you except for less ram. It might be something to do with the graphics or possibly a kernel parameter in use, have you added in any kernel parameters in /etc/sysctl.conf or in /etc/default/grub ?

What does below command say?

lspci -k | grep -EA3 'VGA|3D|Display'

Last edited by HevyDevy (2020-06-23 09:27:41)

berni51 · 2020-06-23 09:27:18

OK. Is there a version you would recommend?
My trial with Archlinux and its 5.xxx kernel was really positive.

@HevyDevy: there is no text on the screen, totally black. And very often it happens, when I enter the login password into the slim dm.
lspci only says "not found" when looking for a VGA|3D|Display

And there is no special parameter in /etc/sysctl.conf or in /etc/default/grub.

Last edited by berni51 (2020-06-23 09:37:47)

rolfie · 2020-06-23 11:23:50

There is a 5.6 kernel in the backports already. I am running this on two machines, and a 5.5 on my file server.

rolfie

HevyDevy · 2020-06-23 12:08:02

@HevyDevy: there is no text on the screen, totally black. And very often it happens, when I enter the login password into the slim dm.
lspci only says "not found" when looking for a VGA|3D|Display

Thats odd, that command should display something, mine displays what graphics hardware is used and the kernel driver and module in use which should be i915. Do what rolfie suggests and install the 5x kernel in backports, might fix the issue, might not?

Last edited by HevyDevy (2020-06-23 12:10:25)

berni51 · 2020-06-23 12:56:13

@rolfie: Thx, I will try. Anything special to consider?

@HevyDevy: simply grepping for VGA says:
00:02.0 VGA compatible controller: Intel Corporation Device 5a85 (rev 0b)

Any many thx for all of your assistance so far.

HevyDevy · 2020-06-23 12:58:57

berni51 wrote:

@rolfie: Thx, I will try. Anything special to consider?
@HevyDevy: simply grepping for VGA says:
00:02.0 VGA compatible controller: Intel Corporation Device 5a85 (rev 0b)
Any many thx for all of your assistance so far.

Is that with Devuan or Archlinux?

berni51 · 2020-06-23 13:07:02

Thats the answer from Devuan.
My yesterdays installations runs, and if I only use the console, its rock solid. Beowulf only seems to crash under xorg.
Do you know if there is another firmware for the i915 graphics?

HevyDevy · 2020-06-23 13:35:58

its under non-free firmware, did you install from non-free or disable it somehow.

The NUC 6CAYB needs Intel "Broxton" GuC firmware as mentioned in below link.

https://packages.debian.org/buster/firm … sc-nonfree

But im pretty sure this should be automatically installed unless you chose not to install non-free firmware.

rolfie · 2020-06-23 14:13:17

Make sure you have contrib main non-free enabled in your sources. It maybe worth also checking if there is an update of the firmware package available in backports.

rolfie

berni51 · 2020-06-23 14:49:23

OK, I did it! Followed your hints and now have a kernel 5.5 Beowulf. Will now heavily work with the system to see, what happens. A deep feeling says that everything is good now. I will report in some hours. Thanks to all so far.

Berni

berni51 · 2020-06-23 20:24:42

Now, more than 5 hours later, Beowulf is still running. No crash, no shutdown. Looks good, but I let the NUC working over night - I should not be too optimistic.
Many thanks for the wonderful help help here, without this help maybe I had given up and would be an Archlinux user.

Good night
Berni

berni51 · 2020-06-24 07:22:53

Good morning! Good news: Beowulf is still running.

There is only one pending question: the i915 firmware for the Intel 500 graphic is not loaded yet by xorg.
Is there a way I can force xorg to use this firmware? Or may it be better not to do anything on a running system?

Berni

HevyDevy · 2020-06-24 08:50:03

hi Berni51, are you sure there is no output from below command?

lspci -k | grep -EA3 'VGA|3D|Display'

My nuc which is the same as yours says below.

lspci -k | grep -EA3 'VGA|3D|Display'
00:02.0 VGA compatible controller: Intel Corporation Device 5a85 (rev 0b)
        Subsystem: Intel Corporation Device 2067
        Kernel driver in use: i915
        Kernel modules: i915

or try lsmod

lsmod | grep i915

Last edited by HevyDevy (2020-06-24 08:59:03)

berni51 · 2020-06-24 09:28:37

HevyDevy, you are right - and I am so silly. Indeed I have the same output as you. Before I grepped without -EA3. So everything is in order.
Thx and sorry

Last edited by berni51 (2020-06-24 09:30:45)

HevyDevy · 2020-06-24 09:44:58

There is more to be done which i have just found out, seems xserver-xorg-video-intel package provides the free as in freedom i915 driver but you can go further and enable non-free drivers if you like from package firmware-misc-nonfree in either stable or backports.

This is how i did it if you are interested, but you must download firmware-misc-nonfree and install the firmware first...

create the file /etc/modprobe.d/i915.conf

add to it

options i915 enable_fbc=1 enable_guc=3

then do commands below

sudo update-initramfs -u
sudo update-grub

then reboot

check to see if i915 non-free (DRM) is loaded by doing below command

sudo dmesg | grep -iE "huc|guc|dmc"

source: https://gist.github.com/Brainiarc7/aa43 … dd835efe57

This is what should be shown more or less from the dmesg...

I installed firmware-misc-nonfree from backports.

 > ~$ sudo dmesg | grep -iE "huc|guc|dmc"
[    2.520372] Setting dangerous option enable_guc - tainting kernel
[    2.524202] i915 0000:00:02.0: firmware: direct-loading firmware i915/bxt_dmc_ver1_07.bin
[    2.524446] [drm] Finished loading DMC firmware i915/bxt_dmc_ver1_07.bin (v1.7)
[    2.545629] i915 0000:00:02.0: firmware: direct-loading firmware i915/bxt_guc_ver9_29.bin
[    2.545827] i915 0000:00:02.0: firmware: direct-loading firmware i915/bxt_huc_ver01_07_1398.bin
[    2.548628] [drm] HuC: Loaded firmware i915/bxt_huc_ver01_07_1398.bin (version 1.7)
[    2.554733] [drm] GuC: Loaded firmware i915/bxt_guc_ver9_29.bin (version 9.29)
[    2.560310] i915 0000:00:02.0: GuC firmware version 9.29
[    2.560314] i915 0000:00:02.0: GuC submission enabled
[    2.560316] i915 0000:00:02.0: HuC enabled

Im actually getting better screen picture now, i was getting ghosting a bit but it seems to have gone!

Last edited by HevyDevy (2020-06-24 09:53:53)

berni51 · 2020-06-24 10:16:26

Hey, great - thanks for the cookbook. Will try today eveneing.

Head_on_a_Stick · 2020-06-24 18:26:25

HevyDevy wrote:

Tseems xserver-xorg-video-intel package provides the free as in freedom i915 driver

No, that package provides the Intel DDX driver. The kernel provides the i915 driver.

HevyDevy wrote:

you can go further and enable non-free drivers if you like from package firmware-misc-nonfree

There are no non-free Intel drivers. The firmware-misc-nonfree package provides the non-free firmware that is required by all recent Intel cards for full performance.

The parameters you are applying are for framebuffer compression (fbc, to reduce power consumption and memory bandwidth needed for screen refreshes) and to enable GuC/HuC firmware loading. Note that GuC/HuC firmware can cause freezing on some systems, which is why it is disabled by default.

The officially official Devuan Forum!

#1 2020-06-22 11:12:08

SOLVED: Beowulf: quite good, but ......

#2 2020-06-22 13:18:30

Re: SOLVED: Beowulf: quite good, but ......

#3 2020-06-22 14:22:19

Re: SOLVED: Beowulf: quite good, but ......

#4 2020-06-22 16:59:26

Re: SOLVED: Beowulf: quite good, but ......

#5 2020-06-22 19:22:06

Re: SOLVED: Beowulf: quite good, but ......

#6 2020-06-22 19:29:31

Re: SOLVED: Beowulf: quite good, but ......

#7 2020-06-23 07:31:01

Re: SOLVED: Beowulf: quite good, but ......

#8 2020-06-23 09:02:25

Re: SOLVED: Beowulf: quite good, but ......

#9 2020-06-23 09:21:36

Re: SOLVED: Beowulf: quite good, but ......

#10 2020-06-23 09:27:18

Re: SOLVED: Beowulf: quite good, but ......

#11 2020-06-23 11:23:50

Re: SOLVED: Beowulf: quite good, but ......

#12 2020-06-23 12:08:02

Re: SOLVED: Beowulf: quite good, but ......

#13 2020-06-23 12:56:13

Re: SOLVED: Beowulf: quite good, but ......

#14 2020-06-23 12:58:57

Re: SOLVED: Beowulf: quite good, but ......

#15 2020-06-23 13:07:02

Re: SOLVED: Beowulf: quite good, but ......

#16 2020-06-23 13:35:58

Re: SOLVED: Beowulf: quite good, but ......

#17 2020-06-23 14:13:17

Re: SOLVED: Beowulf: quite good, but ......

#18 2020-06-23 14:49:23

Re: SOLVED: Beowulf: quite good, but ......

#19 2020-06-23 20:24:42

Re: SOLVED: Beowulf: quite good, but ......

#20 2020-06-24 07:22:53

Re: SOLVED: Beowulf: quite good, but ......

#21 2020-06-24 08:50:03

Re: SOLVED: Beowulf: quite good, but ......

#22 2020-06-24 09:28:37

Re: SOLVED: Beowulf: quite good, but ......

#23 2020-06-24 09:44:58

Re: SOLVED: Beowulf: quite good, but ......

#24 2020-06-24 10:16:26

Re: SOLVED: Beowulf: quite good, but ......

#25 2020-06-24 18:26:25

Re: SOLVED: Beowulf: quite good, but ......

Board footer