Like every January 1, a new year begins. 365 days to enjoy all the possibilities of free and open source software. You will find several suggestions inta list of New Year's resolutions for Linux users and fans of free software.
Of course, these are only suggestions and if you search you will be able to find many other things to do.
New Year's resolutions for Linux users
Install Unusual Linux Distributions
It's true that many of us are compulsive installers, but we almost never go beyond common distributions. 2023 is a good time to try others that require a little more of our attention and skills.
Red Hat Enterprise Linux
It is a distribution aimed at the corporate market and requires the payment of a subscription, however, there are several ways to try it for free.
Traditionally, CentOS was used by companies that did not want to pay for Red Hat technical support. CentOS was first a stand-alone distribution that was compiled from Red Hat's source code. Over time, the company began to collaborate more closely with the community in charge of the project.
When IBM acquired Red Hat, things changed and CentOS became the test bed for future Enterprise releases. That is to say that a certain technology is tested in Fedora, then it is implemented in CentOS and, when it is finally mature, it is added to RHEL.
As often happens in the world of open source, IBM's decision led to the emergence of several alternative projects and competitors such as SUSE Linux and Oracle offering free versions. Red Hat's response was to extend the terms of its free license for developers.
Traditionally it could only be used on a machine for testing purposes. It can now be used in production for free on up to 16 computers including major public clouds, such as AWS, Google Cloud Platform and Microsoft Azure. In the latter case, the costs of using the platforms will have to be paid.
In order to access the program, you must register at this page creating a Red Hat account or signing in with your GitHub, Twitter, or Facebook credentials. Please note that subscriptions are handled from this page.
It only remains to download and install the distribution.
You will wonder why you should do it.
Red Hat is not only the most widely used distribution in the corporate sector, It also develops tools for working with containers, cloud deployments, and application creation.. In addition, you can download a complete documentation of these.
linux from scratch
If you don't like existing Linux distributions or you want to understand the function of each of the components of a Linux distribution, this is what you are looking for.
linux from scratch it is not a distribution. It is an instruction manual on how to get and assemble all the parts to create your own Linux distribution. The project has a continuation called Beyond Linux From Scratch that allows us to bring it closer to any of the usual Linux distributions.
The project also includes additional documentation, patch repositories and automation tools of the construction of the distribution.
As I said above, installing LFS will not only give you a better understanding of what each component of a Linux system does, you will also be able to try replacing packages at your own risk.
LineageOS
Last year I already did the test to install this operating system based on the Android source code for mobile phones on an old smartphone and was very happy with the experience. This year I plan to revive a Motorola G5 Plus by changing the video module and transforming it into my main phone with the LineageOS equivalent of Android 12.
It is worth mentioning that Not all phone models are supported by the official variants, but if you search on Google you might find some developed by third parties. Or, you can create it yourself. The documentation is extensive and can be found all over the web.
I take this opportunity to wish you a very happy start to the year and invite you to tell us if you have any New Year's resolutions related to open source. Below is the feedback form.
You are right. Very nice, but could you show me how to make the linux kernel not look for disks asynchronously?
I've had this problem for quite some time now and I can't solve it for more than a year googling. I've even compiled the linux kernel with the corresponding parameter, but it's no use. Of course I have put the command scsi_mod.scan=sync in /etc/default/grub but it doesn't help either.
As it is a module I have also tried said command "scsi_mod.scan=sync" in a file in /etc/modprobe.d/ but neither.
I have also tried it "without the dot" (as recommended in many forums) "scsi_mod scan=sync". But nothing.
I use Debian11 Bullseye.
Greetings.
That would rather be the New Year's resolution for Linus Torvalds, but if I find a solution I'll let you know
I'll let you know if I get it too.
The only Debian kernels that still maintain disk order are:
Debian10 All.
Debian11 5.10.0-10 official kernel: 5.10.84-1 (2021-12-08) (All following random order).
Debian12 5.16.1 (All following random order).
I've blocked them from updating and I'm still there.
I know that this problem does not affect people who only have ONE disk in Linux or millionaires who have a supercomputer with hundreds of disks and want the system to boot fast. But those of us who have a disk with the OS and four more data disks are not happy to see this at the beginning:
sda Volume3
sdb Volume1
sdc Debian11
sdd Volume4
from Volume2
And what else does it matter, right? Sure, sure... But the Clonezilla Live kernel does the same thing. As long as you are not aware of the order of the discs, you are lost. Because Clonezilla uses the traditional names sda, sdb,… and a kernel that also changes disks. I often have to restart Clonezilla Live itself until it shows me the disks in order. Sometimes the disks seem ordered but it is the flash drive itself that has caught the sda. Again to reboot the system with Clonezilla.
When I make an image of the Debian11 OS and the order is as shown above, the disk image will be made with reference to the drive in sdc. One bad day I have, and I want to restore the image backup, it will alert me that it is going to be restored to the sdc drive. But if that day the Clonezilla linux kernel has given it a different order, it is very easy for me to load the data from another of the disks.
It only happened once and I didn't know what I had done. The system was still with the old image and a partitioned twin image with identical UUIDs on another disk from which all data was missing. I remembered all the relatives of Linux Torwalds and his collaborators. Good thing I had a daily backup on another PC, and I was able to restore all the data.
Clonezilla won't let me restore to a drive other than the one I was imaged with. If it is the "sdc", by the nose it forces you to restore in the "sdc". As you have not fixed well which disk is now "sdc" you are going to make it good. I don't know why Clonezilla asks if it has already decided. Ah well, it must be one last warning before the execution.
I'm going to have to deal with this problem for the life of Debian11. And in Debian12 they already have kernel 6.0.0-6 and I have 5.16.1. Why am I going to have to use Debian12 with that kernel?
This is a problem that I consider very fat. And that is a reason to go to another system.
They say that "Linux you can modify it to your liking." But for a common user that is practically impossible. You have to do a master. I don't think this is within my reach.
I'm going to have to study the whole Linux SCSI system in depth and I'm not a programmer. I have other things to do than dig deeper into Linux.
So don't be surprised that Linux is barely used by people.
In order.
There is a mailing list of Debian users. Maybe they can help you. https://lists.debian.org/debian-user-spanish/
Hello again.
Sorry, but I'm pretty bounced with this problem.
I went to comment on the first site I found. After googling for more than 11 months.
It is not the right forum. Linuxadictos I think it is a portal for news and information about Linux but not to fix problems.
I'm sorry. I'll look elsewhere.
Calm. You raised the issue in a respectful manner and made me aware of something I didn't know existed. I recommend the Debian mailing list that I told you about in the other comment.
Most of the answers to other users who raise this same problem is that they use the UUIDs. The problem is that most applications still use the traditional designation:
gparted
Dolphin
clonezilla live
etc ...
When ALL applications use the UUIDs or labels I won't care about the old nomenclature anymore.
They could use aliases to the UUIDs (which are not for humans).
I can say that this is the biggest problem I've ever had with Linux.
I was wrong.
The Debian12 kernel that doesn't mess up the disks is:
5.16.0-1 (corresponding to official 5.16.7-2)
Hello again.
I am sending you a link so you can read the problem yourself and how the kernel developers are working for the rich who have computers with hundreds of disks.
https://www.suse.com/es-es/support/kb/doc/?id=000018449
in this translated paragraph:
"Starting with kernel 5.3, the order in which SCSI devices are tried and named has become non-deterministic. This is the result of a change that was pushed to add asynchronous device polling. Polling is done asynchronously per device, so even devices on a single bus can appear in "random" order. The logic behind the change is that if you're going to have dozens of disks, you want them to start as soon as possible, rather than polling/failing/waiting synchronously; in an environment where there are hundreds of disks and even more partitions, this change is even more important. »
To the domestic user, “fuck it”.
Well, my PC boots just as fast with ordered disks as disordered. I don't see any advantage. It will be milliseconds.
Can you imagine Windows changing the name of the boot disk C: to F: or to D:?
Well nothing, when I can no longer continue with decent old kernels, I'll leave Linux.
Thanks for the information. What you are suggesting is very interesting.
By the way, one of the New Year's resolutions I suggest is to try FreeBSD.
Hello. Have you tried using noasync? It is used with the mount command.
mount -o noasync /partición /mnto
mount -o sync /dev/sda1 /mntMy disks have the noauto option in fstab. That is, they are not mounted at startup. However, if you run
ls -l /dev/disk/by- you will see that they already have the names sda, sdb,…
Also your command is useless. Indicates that you are mounting a partition on the sda disk, that is, that the disk already has the name sda associated with it. Are you going to be able to change it to sdb? You can not. I imagine that you are one of those who have only one disk. So you can live happily. This problem does not affect you. And even if you have several disks, if you don't use Clonezilla or dolphin, you can live in peace.
For daily life, the mess in the old nomenclature does not affect you. Most applications only need the reference to partitions mounted on /media/ or /mnt. It's just that some applications like Dolphin will show you the disk labels out of order when you go to mount the disks. Others, like Gparted, will show you the drives with their current names out of order. And if you are going to use Clonezilla Live, be very careful not to make a mistake, because its creators continue to use the real names sdX.
I have tried the solution given in the suse link that I indicated above but it does not work. And that the suse article is from September of last year. It should only work for the suse distro.
Although it doesn't help me much, I'm reading now:
https://documentation.suse.com/sles/12-SP4/html/SLES-all/cha-grub2.html
in section
12.3.2 Adjusting the Disk Order
At least suse gives a lot more information. And it's going through my head to switch to Suse. She was one of the distros I tried many years ago.
I add that in the command
mount -o sync /dev/sda1 /mnt
the sync option is for synchronizing data on disk. You are telling it that the data should be written directly to disk and spend less time in memory.
It has nothing to do with scanning (sync or async) the disks at startup, which is what my problem is about.
More misinformation from kernel.org
https://docs.kernel.org/scsi/scsi-parameters.html
The concrete parameter is: scsi_mod.scan
It literally puts:
scsi_mod.scan= [SCSI] sync (default) scans SCSI busses as they are
discovered. async scans them in kernel threads,
allowing boot to proceed. none ignore them, expecting
user space to do the scan.
They say it is set to sync by default. This is false. That was BEFORE. But now the new kernels boot in async mode.
The gentlemen of kernel.org lie or they still have outdated information.
The worst thing is that putting the boot options that many websites recommend doesn't work.
/ Etc / default / grub
GRUB_CMDLINE_LINUX_DEFAULT=”scsi_mod.scan=sync” —-> DOES NOT WORK
Putting the command in a scan_sync file
/etc/modprobe.d/scan_sync
with the content:
scsi_mod scan=sync —–> DOES NOT WORK
They have messed it up so much that their own instructions don't work.
But I'm calm, because in 1 or 2 years they will think about home users again and they will put some solution like generating special kernels for home users. And if they don't, you'll see the share of Linux go up a lot for supercomputers and Microsoft will see its number of home users go up.
Look I'm going to stop sending you messages because I wrote a very long one and I sent it but it doesn't seem to appear.
I forgot about the healthy habit of copying the content in a text editor before posting it in a forum.
Just tell you that you have not understood what I told you. The sync option in the mount command is for data synchronization and has nothing to do with the sync option for kernel scanning of disks at startup.
There is no noasync option (only sync or async). It reads "man mount."
Your command indicates the sda1 partition of a disk called sda but your command does not change it to sdb.
The only solution I see for my problem is to start the PC only with the system disk and then connect the disks in the order I want. This will assign the disks to them in the order you want. But, as you will understand, that would not be a practical solution.
And let's see if you understand it: it has nothing to do with not mounting the disks in a directory. The disks are given the names sdX as the kernel picks them up.
I thought systemd was the culprit but it's not because I installed Devuan (which doesn't use systemd) and they're still messed up. The culprit is the kernel. And the solutions given in kernel.org itself DO NOT WORK. Not even compiling the kernel.
I am not going to write more here because I do not want to waste my time and lose my messages.
I thank you for your kindness.
If I do, I'll post it again on your forum. But I'm tired of the subject.
Farewell.
Hey.
Messages that have commands must be approved manually. Today is Saturday. Even bloggers deserve a break.