9.1 - 9.2 upgrade
bc979 at lafn.org
Tue Oct 8 09:28:44 UTC 2013
On 5 October 2013, at 05:08, Polytropon <freebsd at edvax.de> wrote:
> On Fri, 4 Oct 2013 21:49:18 -0700, Doug Hardie wrote:
>> On 4 October 2013, at 20:03, Polytropon <freebsd at edvax.de> wrote:
>>> On Fri, 4 Oct 2013 19:42:15 -0700, Doug Hardie wrote:
>>>> On 4 October 2013, at 19:08, Polytropon <freebsd at edvax.de> wrote:
>>>>> On Fri, 4 Oct 2013 18:58:52 -0700, Doug Hardie wrote:
>>>>>> The exact sequence was:
>>>>>> Step 1: freebsd-update from 9.1 to 9.2
>>>>> Have you verified in /etc/freebsd-update.conf that "src"
>>>>> is definitely part of what should be updated?
>>>> System is not bootable - can't verify anything…
>>> Does the system (or better, its "enclosure", software-wise)
>>> allow booting a rescue system or an emergency media, such
>>> as a FreeBSD v9 live system?
>> Yes - but there is no one there who can successfully be told
>> how to run it.
> Not even inserting a USB stick (with the FreeBSD memstick data)
> or a CD?
>> We have serious communications issues - they want to use back
>> slashes and have no idea what a slash is.
> Maybe that is the result of many years of "administration" on
> "Windows" PCs. :-)
>> Even if you tell them which key to use, they know better and
>> use a back slash cause thats what Windoze uses.
> Uh... "knowing better" would disqualify them as maintainers of
> a server installation. The inability to learn (or even to read
> and follow instructions) is a dangerous thing.
>> The disk should be in the mail to me now. I will be able to
>> work with it when it arrives.
> Okay, that's also a possible alternative. To be honest, that's
> the first time I hear about this procedure. But doable.
>>> The file /etc/freebsd-update.conf should contain the line
>>> Components src world kernel
>>> if you want to make sure the source is properly updated,
>>> along with the world and kernel (GENERIC).
>> As indicated before, I don't think all the source got updated.
>> The kernel showed 9.2 after recompilation. However UPDATING
>> was not updated. Thats as much as I could check before.
> I assume that this could be possible by inconsistently updated
> sources. It would be a good start to remove /usr/src and download
> the sources of the correct version via SVN _or_ freebsd-update
> again. Before the next installation attempt, /usr/obj should be
> removed as well, just to be sure.
>>>>>> Step 5: reboot
>>>>> Attention: Into single-user mode.
>>>> Not possible since the system is located over 100 miles away.
>>>> Everything has to be done via remote console.
>>> Does this mean "SSH only" or do you have a _real_ console
>>> transmission by which you can access the system _prior_ to
>>> the OS providing the SSH access? I'm mentioning this because
>>> the traditional approach requires (few) steps done in the
>>> single-user mode where no SSH connectivity is provided in
>>> the "normal" way…
>> I have a telnet box that has serial connections to the console
>> ports. That approach has been used without any issues since
>> FreeBSD 2.5. I do disable all ports during the process via an
>> reduced rc.conf file.
> A serial console should also work, but even though I've been
> using serial consoles (and _real_ serial terminals), one thing
> I'm not sure about: Is it possible to interrupt (!) the boot
> process at an early stage to get to the loader prompt and
> boot into single user mode from there?
> boot -s
> If not, do you have the "beastie menu" (or whatever it is called
> today) enabled to go to SUM to perform the "make installworld" step?
> Anyway, if you can install everything is required with the disk
> at home, and then send it back to that "datacenter" (according
> to your characterization, the quotes are deserved), that should
> solve the problems and make sure everything works as intended.
The Thick Plottens…
I received the drives and installed them on a working system. The failed system is structured with a single partition for the system and another for swap. For some unknown reason, the BIOS got left configured to boot the extra disk if its powered up. That turns out to be handy. I can boot a working system with the corrupt drive powered off.
Booting from the corrupt drive yields the normal hardware info followed by the Beastie image and immediately by a multitude of lines (repeated many times):
Consoles: internal video/keyboard serial port
BIOS drive C: is disk0
BIOS drive D: is disk1
BIOS 639kB/1037824kB available memory
FreeBSD/x86 bootstrap loader, Revision 1.1
(doug at zool.lafn.org, Thu Oct 3 04:23:13 PDT 2013)
Can't work out which disk we are booting from.
Guessed BIOS device 0xffffffff not found by probes, defaulting to disk0:
I was able to capture these by using a serial console connected to another computer. The lines only appear on the serial console once. They scroll by on the real console many time - all too fast to read anything. Then after a few seconds of that, the screen goes black, and the system reboots. The cycle then repeats… Pressing any key does nothing. I even filled the keyboard buffer with spaces hoping to stop boot, but nothing seems to stop it.
I checked and the freebsd-update.conf include world sys and src. I rebuild everything after removing /obj just for grins and giggles. I have installed the kernel and world using DESTDIR to put it on the corrupt drive. Same messages again.
I now have the corrupt drive mounted on /mnt and am trying to update the src again. Using:
freebsd-update -b /mnt fetch
updated files list show /usr/src/sys…
and updating to 9.1-RELEASE-p7
freebsd-update -b /mnt install
This is running slower than molasses in January. Its run for almost 30 minutes and only 3 files have been updated. There must be network issues between me and the server. I'll let it run tonight but I am going to crash now. Long day. More tomorrow.
More information about the freebsd-questions