Debian User Forums

Hi all!

I've recently installed Debian 11 bullseye, netinst, for 64-bit PC (amd64) on a brand new desktop machine with the following characteristics:
- Intel Core i7 9700 9th gen.
- RAM: Kingston ValueRAM 32 GB
- Motherboard: Gigabyte H310M 2.0
- SSD Hard Drive 480 GB

The installation run without issues, connected to internet via ethernet. The partitioning was standard (/home, /var , ... ) as per default sizes.
The machine came with a Wireless N PCI Express Adapter TL-WN881ND, whose driver could not be located during installation, being a non-free file (apparently the missing file is 'rtlwifi/rtl8192eefw.bin').
I then tried to install the driver manually (package firmware-realtek). It run without problems. Afterwards I reboot and notice that it fails to start GNOME Display Manager (attach1).
The I login using alt+f2 to check 'systemctl status gdm.service' and the output is in attach2.
As a last step, I try to remove the conflicting package, but I'm unable to do it (attach3).

I'm sorry for the poor quality of the pictures, my cell phone is not in the best shape.

I know that I can do a fresh reinstall, especially considering that there are no personal files to backup, but I would like to know if this has a solution, in case this happens in a future.
Hope someone could help me! I'd really appreciate it. Note that I'm quite novice using Linux, so please be as descriptive as you can.

Best,
Joaquín

What is the output of command?

Thanks for your answer, Victor. The output is attached.

So, basically your problem is that your /var volume is full. You was warned about that ('No space left on device' on the first screenshot).
Why have you moved you /var to a separate partition? Just make partitions for /, /home and swap. 30 to 40 GB is usually quite enough for Debian root partition.

The other question is, what consumes 9 GB in /var; check your /var/log and /var/cache directories, what are their sizes; clean your logs and cache.
Or reinstall your system with disk repartitioning.

Also, why have you mounted your /tmp to a disk partition? It is usually mounted to tmpfs in memory.

I actually selected the partitioning option according to image 'a0'. I didn't explicitly tell the installer the corresponding space for each partition. I supposed that the standard sizes would be ok, though I guess I was wrong.
/var directory is quite heavy due to /var/log . There are three super heavy files (see 'a2' attached). An excerpt of 'syslog' is in attachment 'a1'. The file is full of the depicted error, which means nothing to me due to my computational ignorance.
Next weekend I'll try reinstalling, partitioning according to your indications, and post the outcome here for future reference.

Thanks for your time!

Joaquín

By the way, after removing those three large files, I am able to boot correctly. Perhaps it won't be necessary to reinstall.

Best,
Joaquín

Yes, it's not necessary to reinstall. But this problem indicates that something in your system goes... not well enough, so to say. It's strange that logs are not rotated. It's better to disable and mask rsyslogd service which generates those ridiculous logs. You can access system logs via systemd's command journald. I configured my journald to have storage=volatile and readKMsg=no (see man journald.conf). So, no system logs are stored to disk, but I can access general system messages via 'journalctl' command and kernel messages via 'sudo dmesg' command. Less logs stored to disk = less SSD wearing.

But the problem with your system lies a bit deeper. Your logs are so bloated because they are filled with error messages spawned by the kernel (as I can see by the last screenshot). Something is wrong with your PCI configuration. I suppose you haven't installed all necessary firmware packages. Search for 'firmware' keyword in Synaptic or any CLI package manager (I prefer to use aptitude) and install everything you need. Hopefully you enabled non-free repos. You can see your hardware configuration by running 'lspci' and 'lsusb' commands, there you can see the vendor names (firmware packages are commonly labeled by vendors).

Also, install CPU microcode for your architecture (intel-microcode or amd-microcode packages).

Also, it's better to have /tmp mounted on RAM. If it's mounted on disk, it's, in the first place, slower, and, OTOH, this will accelerate wearing of your SSD. Edit your /etc/fstab as root and comment out the string related to /tmp. Systemd will automatically mount /tmp on tmpfs in that case. Or you can do it manually as do I: (Set the size at your liking, usually some 20% of your RAM would be enough, but it's better not to set it less than 512MB; those mount options are rather strict security-wise, you can omit the 'noexec' option because it may break some install scripts, but I only encounter this with rkhunter in Debian; no other good program in any distro seems to try to run executables stored in /tmp; generally only malware would try to do that).

EDIT: BTW what is connected to your PCIe bus? I mean, the device 0000:00:1c.0. Is that NVidia card?

Victor Brand wrote: ↑2021-09-12 23:29Also, it's better to have /tmp mounted on RAM. If it's mounted on disk, it's, in the first place, slower, and, OTOH, this will accelerate wearing of your SSD. Edit your /etc/fstab as root and comment out the string related to /tmp. Systemd will automatically mount /tmp on tmpfs in that case. Or you can do it manually as do I:

Code: Select all

tmpfs /tmp tmpfs noatime,nosuid,nodev,noexec,mode=1777,size=2G 0 0

In Bullseye, tmpfs is automatically mounted to memory at boot time, no fstab entry required. Also advice to limit writes to SSD is ancient history, modern ones have better life expectancy than hard drives.

EDIT: BTW what is connected to your PCIe bus? I mean, the device 0000:00:1c.0. Is that NVidia card?

If so, firmware-misc-nonfree may be required.

sunrat wrote: ↑2021-09-13 00:43 In Bullseye, tmpfs is automatically mounted to memory at boot time, no fstab entry required.

I wrote about that. I use manual entry in fstab for security reasons because my mount options are more strict than the default ones. OP has his /tmp mounted on disk (see screenshot of df output) for some reason, I believe there is a particular fstab entry for that.

sunrat wrote: ↑2021-09-13 00:43 Also advice to limit writes to SSD is ancient history, modern ones have better life expectancy than hard drives.

When I bought SSD last year I used to worry about its lifetime and wearing. I installed Debian on ext4 partition and I had some 0.5 TB of writes during 1.5 months. Then I used to run Artix Linux installed on a btrfs subvolume with zstd:2 forced compression and used to update it once in a week or two. There was additional 1 TB of writes in a half of a year. I didn't install rsyslog and also moved browser cache to /run/user/1000 (i.e. to RAM).

Thus unnecessary logging and caching could potentially write a lot of data to your hard drive, even more than those pesky Arch updates. (Now I have Bullseye installed on another subvolume of the same btrfs partition. Debian runs lightning fast and smoothly.) Moreover, disabling unneeded services and moving cache to RAM makes your computer run faster. I believe those rsyslogd and journald persistent logs are unnecessary on a home PC (contrary to a server). Who ever reads those logs by the way? I configured auditd for apparmor logs though, because this is something I do read.

sunrat wrote: ↑2021-09-13 00:43

EDIT: BTW what is connected to your PCIe bus? I mean, the device 0000:00:1c.0. Is that NVidia card?

If so, firmware-misc-nonfree may be required.

Yes, this. But there may be other unpleasant surprises regarding nvidia proprietary driver and wayland. We have not enough information about that. Although I suppose the OP is interested in revealing and eliminating the source of those continuous error messages.

Hi guys,

I don't think I have an nVIDIA card. The outputs of lspci and lsusb are in the attachment. There you can see what's in 00:1c.0. Does this shed some light on the problem?
I'm gonna have to postpone the execution of the indications you gave me for next weekend, since I don't have much free time right now. I very much appreciate your guidance!

Best,
Joaquín

Please copy/paste terminal output as text inside CODE tags, much easier to see than photos.
Your graphics are UHD Graphics 630 which is the onboard graphics on your CPU. This should work out of the box although you will need to install the package intel-microcode from non-free for the CPU.

Hi guys!

Please copy/paste terminal output as text inside CODE tags, much easier to see than photos.

Sorry for that, by the time I said I was able to boot, I hadn't still disabled 'rsyslog', so the huge logs still were created and didn't allow me to use any application, so I managed to pick the errors logging in rescue mode and using my cellphone and another PC to post here. I'll try to minimize pictures from now on.

So, I'm proceeding with your indications:

It's better to disable and mask rsyslogd service which generates those ridiculous logs. You can access system logs via systemd's command journald. I configured my journald to have storage=volatile and readKMsg=no (see man journald.conf). So, no system logs are stored to disk, but I can access general system messages via 'journalctl' command and kernel messages via 'sudo dmesg' command.

I disabled 'rsyslog' and the huge logs clearly stopped being created, which is great for me, but still haven't configured journald. I'll go on with that later on.

Also, it's better to have /tmp mounted on RAM. If it's mounted on disk, it's, in the first place, slower, and, OTOH, this will accelerate wearing of your SSD. Edit your /etc/fstab as root and comment out the string related to /tmp. Systemd will automatically mount /tmp on tmpfs in that case.

As a first step, I commented out the line associated with /tmp. No manual mounting done for the moment. If this has cons, let me know and I'll go on manually.

Search for 'firmware' keyword in Synaptic or any CLI package manager (I prefer to use aptitude) and install everything you need.

I have a problem here. Using Synaptic, I search for 'firmware', mark all of the packages to install, and an error pops up, saying 'Could not mark all packages for installation or upgrade. The following packages have unresolvable dependencies. Make sure that all required repositories are added and enabled in the preferences': I think the repositories selected are alright, see image attached.

Then I try using 'Aptitude', getting this:

Your graphics are UHD Graphics 630 which is the onboard graphics on your CPU. This should work out of the box although you will need to install the package intel-microcode from non-free for the CPU.

That's one of the packages seemingly present and up-to-date.

How should I proceed?

Best,
Joaquín

Hi all,

It seems that the problem I'm experiencing with the kernel errors is not that uncommon. Here is some more additional info for the interested reader:

https://itsfoss.com/pcie-bus-error-severity-corrected/

I'm just wondering how harmful might it be to use the computer ignoring this behavior. In the article this error doesn't seem like a big deal. If anyone has some advice, I'll be happy to read it.

Best,
Joaquín