[sle-beta] SLES15 beta 6, initial observation

[Joe Doupnik]

On 04/02/2018 23:07, Thorsten Kukuk wrote:
> On Sun, Feb 04, Joe Doupnik wrote:
>
>>      The second test was with one CD, as just mentioned. The peculiar item
>> observed thus far was when partitioning. I said please redo the layout with
>> the offered existing partitions, I removed all shown for the only drive, and
>> I created my usual three of BOOT (/boot, EXT2, 128MB), SWAP (swap, 512MB),
>> and ROOT (/, 6.xGB).
> No idea why people still use this extra /boot partition. This has not
> a single advantage, but a lot of disadvantages.
> Nobody would come to the idea to put /lib64 with libc and everything
> on an extra partition, so why divide the kernel or bootloader in two
> parts and put them on two different partitions? This only makes trouble
> and was only necessary in the past for broken BIOS versions. But that
> problem should have gone long ago.
>
>    Thorsten
>
-------------
     To clarify matters for Thorsten and other readers. A while ago I 
covered similar territory.
     The /boot story as I relate it goes like this. With earlier SLES 
editions I created a two partition scheme, root (/, XFS) and swap. That 
worked until a kernel upgrade was installed and then the system would 
not boot. I infer the reason was the new initrd did not have the XFS 
intelligence. Additionally, if that root had outstanding journal entries 
at boot time then grub would be unable to perform them because root was 
mounted read-only at that moment. Consequently the system may not boot.
     These aspects are not BIOS related, yet reaching a bootable 
partition can be BIOS dependent which is why a bootable partition should 
be created before other large partitions.
     However, if I create a boot partition, type EXT2 (no journal, no 
XFS), followed by swap and root (XFS) partitions then all is well 
through patches and major upgrades. My surmise is composition of a 
workable initrd and relatives is the key factor, with their 
understanding of XFS being an important thread.
     That's been my experience. Details today may have changed and to 
know would take a bunch of validation experiments. I know this can seem 
to be a reactionary view, but setting aside BTRFS the problem has been 
real enough for a long time and is easily avoided by having the /boot 
partition. Further, using a /boot partition ought not clobber a new SLES 
and certainly should not be a notable disadvantage.

     For SLES 15 beta 6, there is that new /boot popup complaint message 
to learn about. My memory says there could have been two disks, sda and 
sdb, shown in the partition menu, even though only one hard disk was 
created for the ESXi virtual machine, and the sdb one had a boot 
reference. That's my vague memory. At this moment I have restarted that 
test virtual machine after removing the second CD drive. It did not 
boot, saying no o/s, but I could enter Rescue mode. After poking about 
the boot partition (sda1) I did not find a menu.lst file which is 
normally present for SLES11 and prior. I am about to remove that virtual 
machine and rebuild with only one CD drive. In any case, the double CD 
drive presence during installation has led to trouble in beta 6, plus 
that new /boot popup message.
     Thanks,
     Joe D.