On a test machine I had installed Debian Enlightenment. first, and then Manjaro second. Manjaro screwed up the booting. I repeatedly used the Ubuntu Boot Repair disk to get me unstuck. The Boot Repair disk is Linux agnostic. Don On Fri, 14 Sep 2018 at 12:24, o1bigtenor via talk <talk@gtalug.org> wrote:

On Fri, Sep 14, 2018 at 9:53 AM, D. Joe via talk <talk@gtalug.org> wrote:

...

On Fri, Sep 14, 2018 at 09:22:20AM -0500, o1bigtenor via talk wrote:

...

Greetings

I want to have two different copies of debian on one box with the choice of which when I boot in.

[...]

...

So I've installed both of these systems (more than once each) they have their own partitions for everything but boot and efi yet I'm only seeing one system available on grub (depending upon the last install as to which).

...

So I'm doing something wrong!! I tried using grub updating tools (# os-prober) still no joy. The web pages that I'm finding seem to be for an older version of grub and, as usual, I'm finding man pages are like reading cuneiform (which I find unintelligible).

This is likely something quite simple but I'm just not seeing it - - - please - - - some ideas/pointers?

To the best of my understanding, these tools are built with the assumption that one wants to run just the OS that invokes them.

Although the Debian wiki has some hints

https://wiki.debian.org/Grub

Looked at that page earlier - -- its about 7 to 9 years out of date at this point. The information there is current for grub 0.97 and I'm on 2.xx and there are enough differences so that I am not sure if the ideas presented there wouldn't even make things worse.

...

this seems to be more direct

https://askubuntu.com/questions/16042/how-to-get-to-the-grub-menu-at-boot-ti...

Not really the issue - - - sorry.

...

Are you hitting shift during boot time to get the grub menu, from which

you might then be able to select amongst configured choices?

...

If your successful bringing up the boot-time menu, but it isn't

configured to offer you the choices after holding down shift, you may have to muck around in /etc/default/grub from which update-grub et al seem to take their lead.

...

Take some care to dig into the boot time menu, I'm starting to see only

the default (eg, the most recently installed OS) as the obvious choice at the grub top level menu during boot time, with other options buried in a submenu. (Sorry not to be more specific about how those appear--I'm favoring a quicker response over booting up a VM to see more exactly what the strings presented are).

When you have multiple operating systems you get choices, even when you have mutliple versions of one system you can have options. I'm just not seeing the option(s) for the second system. Was told that there is a grub.conf file so that's where I'm going to be looking pdq.

Thanks for the ideas.

Dee --- Talk Mailing List talk@gtalug.org https://gtalug.org/mailman/listinfo/talk

On Fri, Sep 14, 2018 at 3:06 PM, Lennart Sorensen via talk <talk@gtalug.org> wrote:

On Fri, Sep 14, 2018 at 02:53:06PM +0000, D. Joe via talk wrote:

...

To the best of my understanding, these tools are built with the assumption that one wants to run just the OS that invokes them.

Although the Debian wiki has some hints

https://wiki.debian.org/Grub

this seems to be more direct

https://askubuntu.com/questions/16042/how-to-get-to-the-grub-menu-at-boot-ti...

Are you hitting shift during boot time to get the grub menu, from which you might then be able to select amongst configured choices?

If your successful bringing up the boot-time menu, but it isn't configured to offer you the choices after holding down shift, you may have to muck around in /etc/default/grub from which update-grub et al seem to take their lead.

Take some care to dig into the boot time menu, I'm starting to see only the default (eg, the most recently installed OS) as the obvious choice at the grub top level menu during boot time, with other options buried in a submenu. (Sorry not to be more specific about how those appear--I'm favoring a quicker response over booting up a VM to see more exactly what the strings presented are).

Certainly OSs like Debian believe they own /boot and that packaged files go in there. In general it is easiest to just keep it part of the root filesystem. The EFI boot partition on the other hand is shared by all OSs on the system. There is rarely a good reason to make a separate partition for /boot these days. Encrypted rootfs is one case or root on LVM, but other than that, not really.

OK - - - good to know. For many years it was /boot, / (or /root), /usr, /var, /tmp, swap, /usr/local and /home. I've had enough issues because / was too small, ditto for /usr. So on a new system I can drop /boot and only add /efi or ???? Regards Dee

On Fri, Sep 14, 2018 at 11:10 PM, D. Hugh Redelmeier via talk <talk@gtalug.org> wrote:

TL;DR: lots of history, some explanation of why things are done certain ways.

| From: o1bigtenor via talk <talk@gtalug.org>

| OK - - - good to know. For many years it was | /boot,

Covered well by Lennart.

This first came up for me with UNIX System V release 2 on a PC. The boot loader only understood a a special simple file system (BFS). This file system was pretty useless for real work so only a minimal set of stuff was held there. I *think* it was mounted as /boot. It was a lot like the kernel + initrd.

There's an interesting bootstrap problem that it addresses: how simple can you make the bootloader and still make it do what you need.

LILO was an elegantly simple solution but it pinched:

- It used BIOS for disk I/O. Eventually BIOSes could not address the whole of a disk and this bizarre limitation bled into the bootloader.

- The bootloader did not understand any filesystems so a tool was needed to put absolute disk addresses into the bootloader whenever a kernel or initrd was moved, added, or deleted. Forget to do this and your system became unbootable.

On my first iteration on linux I do remember lilo, vaguely (grin it is a while ago!).

- Secure Boot? I don't see how that could fit into the LILO framework.

- UEFI? I think that a new bootloader, as simple as LILO, but a lot better, could be written using the API provided by UEFI.

GRUB is quite complex. Among other things, it needs to understand any filesystem from which it can boot. For another, it needs its own disk device drivers so that it can get around the BIOS limitations outlined above.

I have a little PC that has a too-small "disk" (eMMC) to hold Windows and Linux. I wanted to use an SD card to hold Linux. The UEFI firmware doesn't have a driver for the SD card. So the machine cannot boot from the SD card. If GRUB had such a driver, then I could put /boot on the SD card, but it doesn't. Still, it would be quite reasonable for it to have such a driver.

| / (or /root),

Those are quite different.

/ is the root filesystem

/root is the home directory for user "root". It is conveniently on / so the root user can be logged in without /home being mounted.

Thank you very much - - - - had not ever seen an explanation of the whats and whyfors on this. Kn ow for next time! (Installing that is)

| /usr, /var, /tmp, | swap, /usr/local and /home. I've had enough issues because / was too small, | ditto for /usr.

If you use LVM (I don't) you get a level of indirection that lets you increase partition sizes pretty freely. So you don't have to get allocations right at the beginning.

I've never used LVM so that might be useful but I've been able to figure things out so they've worked fairly well so I will look into LVM but I'm not sure I will be using it.

/tmp is quite reasonably on a separate special-purpose filesystem. On Fedora 28, the default is to use the filesystem type "tempfs".

In the days of LILO, /boot needed to be addressable by LILO so it was constrained. That was one reason to put /boot in a separate filesystem.

Once upon a time, / was all you needed for single user mode. You kept everything else in things like /var and /usr so that there was a great chance that a broken system could still be booted in single user mode. / was almost, but not quite, read-only. Long gone.

Sometimes fleets of boxes shared, say, /usr, through NFS. This meant some kinds of system administration were easier and disk resources were less wasteful yet each machine was faster than a diskless workstation.

Now /, /var, /usr/var, /local, /usr/local, /bin, /opt, might as well be on one filesystem.

One of my linux mentors, who calls himself a linux dinosaur (started on a System V on a pc) even advocates for a separate partition for /var/log to forestall any software issues choking off disk usage.

I keep /home separate so that my user data is mostly in its own filesystem. Then replacing the OS is almost simple. Too bad the dotfiles need to adjust for different OSes.

| So on a new system I can drop /boot and only add /efi or ????

/boot/efi, not /efi. Of course that isn't the name of the filesystem it is the mount point.

I was remembering it being called just EFI on the debian partitioning setup pages in the install but wasn't sure of its name. Likely another of those places where we are, as in English, divided by a common language.

(Some people insist on using BIOS, not UEFI. In that case, you probably want a separate /boot and no /boot/efi. I think UEFI is a better choice.)

As I have my main systems with more than 2 TB of hdds bios is no longer an option. I find when installing a new system that things go easier when I use my CDResuceDisc and its copy of gparted to set up the partitions the way I want and then mounting the install disc to effect the installation. Thank you for the ideas and tips!! One of my frustrations with linux has been that it can be very difficult to find clear understandable information on many parts of the system. Much of the documentation seems to be written for someone who is well versed in things and is looking for a clue or a reminder on 'how things work'. Your note is how I wish man pages were written. If only there were examples of the command usage, happens rarely but is starting to happen, with what to do when there are errors or problems would be wonderful but the assumption always seems to be that the user just knows. My question is 'how does one learn?' - - - even the Socratic method, which can be quite challenging for a student works from what the student 'knows' to educating them in what they 'don't' know. I think that these lacunae in documentation are because its far more fun to build things than it is to describe the hows and the whys and the whatfors. So your explanations offered are very welcome!!! Regards Dee

On Sun, Sep 16, 2018 at 12:10 AM, D. Hugh Redelmeier via talk <talk@gtalug.org> wrote:

| From: o1bigtenor via talk <talk@gtalug.org>

Note: I am only an amateur sysadmin.

And I am not even that good - - - grin but i'se a learning that too!

| One of my linux mentors, who calls himself a linux dinosaur (started on | a System V on a pc)

Picky picky: System V isn't Linux. It's UNIX.

Ja ja - - - got it. but it were unix before linux, although today one might wonder listening to some.

I first used UNIX seriously with 4th or 5th Edition in 1975. But I first became strongly interested in 1974 from reading "The UNIX Time-Sharing System" in the Communications of the ACM.

I think that a number of GTALUG folks used Linux before I did. I switched my desktop from Solaris on SPARC to Linux about 1997 (I played with Linux before that).

| even advocates for a separate partition for /var/log | to forestall any software issues choking off disk usage.

Good point. What causes disk-full events for you? If you can isolate them to separate filesystems, that's good.

Well - - - its one thing I haven't done yet. Haven't been bit by it yet either. Likely would do after the first instance - - - grin.

Here are the ones that have hit me:

- too many log messages (/var/log). That can easily be cured by log system settings.

As I'm finding log message useful (strange for someone with a background primarily as a 'user') I think I would rather have more here than less although some of the messages are quite - - - -carefully now - - - useless (when there are a few thousand lines exactly the same in a row) but none the less I would rather enhance than truncate this function.

- too much email (/var/spool)

Today I get maybe 30 to 75 emails. I remember a time when I could get over a thousand. Today I would be quite a bit more into the deleting than I was then too.

- too many saved update packages (due to a PackageKit / dnf bug) (/var/cache)

- too many packages installed (/)

- too many core dumps (/var/spool/abrt) A firefox dump usually takes 4GiB or so on my system.

HDDs are so big and cheap these days that the simplest solution is to wildly over-allocate for each partition.

Unfortunately, SSDs are so much faster that you ought to use them and their capacity costs more. I use both. I haven't bothered to put any problematic directories on the HDD.

| One of my frustrations with linux has | been that it can be very difficult to find clear understandable information | on many parts of the system. Much of the documentation seems to be | written for someone who is well versed in things and is looking for a clue | or a reminder on 'how things work'.

There are other challenges

- Linux is very large. Mostly things get added and very little gets deleted. (When I first used UNIX, the largest RAM was 256KiB. It ran off a disk that was 2.5MB. You can bet is was simpler.)

In comparison to the dominant desktop I don't think that's true but I do know that the economy of code AND logic that was used in coding back then is in quite a different universe than the present iterations.

- Documentation gets out of date.

One of the things that I have been finding very often - - - even with this present search for answers. There is a lot of information out there that is wildly out of date but there isn't even a date on the document to alert one to its fossilization. I have taken to dating all of my documents for myself to try and reduce the issue. Very seldom does information from the major players get marked with a date. It is only with some experience that this outdated stuff can be recognized. Finding information for current system thinking is often almost impossible. (Coders hate documenting!) Even more challenging is the all to common attitude that the newest is the best stuff. I want stuff that works - - - not stuff that has not been thrashed so its flaws aren't known. For business use something that works is far more important than stuff that is the newest (with huge bling).

- Linux is very general-purpose. How you should set it up depends on many things. Consider the difference between Linux embedded in a home router and Linux running a supercomputer. (Most routers, smartphones, tablets, supercomputers, websites, cloud systems, etc. run Linux.)

- each distro is a different. Each installation can be different.

Even more challenging is that Linux appears to be tweaked for those where the tweaking and working under the hood IS the object of the OS. There doesn't seem to be that much that is just for the USE. There are mountains of things that are cool from a developers point of view but for a 'users' - - - well - - - its all to often not much use. IMO far too often change is done for change's sake. Another issue is that there is a lot of behind the scenes activity that is quite antithetical to the visible. It feels a lot like the frog swamp is being heated very carefully!

One key to understanding Linux is to understand its anatomy. Then you can study the organs of interest in isolation.

I'm a trying but finding good tools and guides is not very easy whereas for the M$ world there is a plethora of tools and guides even if far too many of those are that good. Thanks Hugh - - - appreciate the pointers and nudges! Dee