Bug#347692: Package: installation-reports
Package: installation-reports
INSTALL REPORT
Debian-installer-version: <Fill in date and from where you got the image>
2006-01-03 from ftp.debian.org
uname -a:Linux bulwark 2.6.8-2-386 #1 Tue Aug 16 12:46:35 UTC 2005 i586
Gnu/Linux
Date: started 2006-01-13, completed 2006-01-11
Method: How did you install? What did you boot off? If network
install, from where? Proxied?
Although have 2.2r2 CD wanted to try install without it, to answer
the question "What happens if I have to do scratch install and the
CD's dead?"
After reading the install manual several times, decided that since
I don't have USB or CD burner, and only ppp link to net to do the
net install, booting from floppy. See log file attached for gorry
details.
Machine: AST Pentium 75
Processor: Pentium 75
Memory: 40 Mb
Root Device: /dev/hdc:
It came with an 840 MB hard drive on hda but will only recognize
that drive on hda, any other drive on hda is not recognized. If
that drive is on hda, the BIOS doesn't see a drive on hdc.
Didn't want to use a large (for me) disk for a simple firewall
(Bulwark). Had slow noisy 540 MB hard drive wanted to use
Root Size/partition table:
Here is the table as I wanted it, and as I ended up with, but
see below for the juggling act it took to get to this.
/dev/hdc1 23M /boot
/dev/hdc5 109 /
/dev/hdc6 132 /usr
/dev/hdc7 189 /var
/dev/hdc8 89M swap
Output of lspci and lspci -n:
0000:00:00.0 Host bridge: Intel Corp. 430FX - 82437FX TSC [Triton I] (rev 02)
0000:00:07.0 ISA bridge: Intel Corp. 82371FB PIIX ISA [Triton I] (rev 02)
0000:00:08.0 VGA compatible controller: ATI Technologies Inc 215CT [Mach64 CT] (rev 09)
0000:00:00.0 0600: 8086:122d (rev 02)
0000:00:07.0 0601: 8086:122e (rev 02)
0000:00:08.0 0300: 1002:4354 (rev 09)
Base System Installation Checklist:
[O] = OK, [E] = Error (please elaborate below), [ ] = didn't try it
Initial boot worked: [E] floppy boot won't grab off ppp link
nor boot hd-media kernel because it
doesn't have the modules to read a
simple IDE disk
Configure network HW: [E] didn't detect NIC but then this is the
firewall machine, the internet is via
ppp
Config network: [E] ditto
Detect CD: [E] no cd to detect, no cd burner, no usb
did find the ISO image once
Load installer modules: [O]
Detect hard drives: [E] Floppy boot disks don't recognize IDE
drives, only usb
Partition hard drives: [E] Ok if you get it right the first time, but
if you go back to change it, hangs.
Create file systems: [O] OK, but don't know if it searches for bad
blocks. Old installer asked, and
defaulted to no. This one doesn't ask,
so did it manually prior to running the
installer
Mount partitions: [O]
Install base system: [O]
Install boot loader: [O]
Reboot: [ ]
Comments/Problems:
<Description of the install, in prose, and any thoughts, comments
and ideas you had during the initial install.>
Install logs and other status info is available in /var/log/debian-installer/.
Once you have filled out this report, mail it to submit@bugs.debian.org.
Here's my log that I kept as I went along:
Log of the bulwark's new install of Debian 3.1
Lilo is set up to boot /boot/newinstall/vmlinux
business-card iso is in /
I thought all should work, but the installer version on the floppy boot
doesn't pull base stuff out of the iso file but insists on pulling off
the net which is not available.
Download the hd-media files again, put them where they belong and try
again tomorrow.
Reading the debian CD page again, the business-card.iso does not contain
any packages. The basic netinst does. Its 180MB, downloading it now.
The install programme is able to find the iso files deeper in the file
system. /usr on bulwark's hda has room for a 180MB iso file. This
partition will be inactive during the install. I put it in
/usr/local/share
I made a boot floppy, to be tested, which uses lilo to boot kernels on
the hard drive. I looks for /hda5 /hdc5 as well as /boot/newinstall on
hda5 and hdc5. With this, I should be able to swap the two drives so
that the iso is on hdc whatever and the new install happens directly on
hda.
Next: test boot floppy per paper notes, the swap drives and start
install.
OOPS: first, I forgot to turn off automatic backups so its been backing
up the .iso file to /var, filling it so mail doesn't work. Moved the
.iso to /usr/local/share/doc.
second, the lilo floppy doesn't work.
This has turned into quite a puzzle. The whole puzzle started with
trying to avoid a lilo problem once install is done re moving the drive
from hdc to being hda. It has turned into: Faced with a bare-metal
recovery, assuming that I have a blank drive, how can I install debian.
The short answer is that I can't without being able to burn a CD. If I
have a prveious set of boot floppies with utilities (which I do), I can
partition the drive, format it, and using split and sneaker-net could
move the ISO image, the kernel, and the initrd over to the new drive.
The question is how can I now boot that? Syslinux only boots kernels on
the floppy it's on; Lilo only boots kernels that are there when its run.
So how can I install Lilo on the drive? Its a static executable. Could
I manually create a miniture root directory tree with /etc and /boot and
with lilo on the floppy using -r, install lilo on the drives MBR? Would
GRUB work?
GRUB is great: got the package grub-disk and made a floppy from the
image file. It boots with a menu. alter the menu item (so can make
sure everything right before booting, rather than command-line).
So far, it will boot the standard linux on /dev/hda (hd0,0). The BIOS
doesn't see the second drive, so neither can GRUB.
ADDENDUM ADDED LATER HERE FOR CLARITY
BIOS NOTE: this computer will only recognize the drive it came
with on ide0. If put another drive on ide0 it doesn't see it.
If put another drive on ide1 with nothing on ide0, it sees it as
the first hard drive. Later when say swap drive, means
disconnect origional drive.
END OF ADDENDUM
It boots the install, after grub: serial --unit=0 --speed=115200
terminal serial
root (hd0,0)
kernel (hd0,0)/newinstall/vmlinuz root=/dev/ram0
ramdisk_size=12000 console=ttyS0,115200n8 debconf/priority=low
initrd (hd0,0)/newinstall/initrd.gz
It finds the /usr/local/share/doc .iso file.
-----------------
Try it for real:
having put the vmlinuz and initrd.gz files on
hda5/boot/newinstall,
swap hda for hdc. This way, we're installing on hda but the
partition containing the iso is still available.
Use grub to boot the install:
root (hd0,0)
kernel (hd0,0)/newinstall/vmlinuz root=/dev/ram0
ramdisk_size=12000 console=ttyS0,115200n8 debconf/priority=low
initrd (hd0,0)/newinstall/initrd.gz
boot
SO FAR SO GOOD! get the low-memory notice.
detecting hardware to find hard drives...
unable to load some modules:
floppy, ide-mod, ide-probe-mod, ide-generic, ide-floppy
It found the iso in equiv to /dev/hdc? /usr/local/doc
chose optional modules ppp and serial, just in case can then
carry on with ppp network install.
FAILED TO COPY FILE FROM CDROM RETRY?
retried, failed.
I wonder if the iso is too low in the directory tree. In the
installation manual for harddrive booting (A.2.5) it says that
all should be in the top level directory of the hard disk.
TRY AGAIN.
swap drives back.
can't put the iso in the / partition (hda5) because iso is too big.
it will only fit on hda7 (/var). Will try this, in case the programme
just want's the iso on the same disk. Copied rather than moved so still
have copy local to this machine.
Try again:
shutdown bulwark
swap drives but leave working system disconnected
boot grub via floppy
set for serial terminal
detecting hardware to find hard drives...
unable to load some modules, as above.
found iso file in eq to hda7.
load installer components from installer iso...
DON'T CHOOSE OPTIONAL MODULES AT THIS TIME.
loading components....
FAILED TO COPY FILE FROM CDROM RETRY?
Went down to console and used second vt.
df shows that cdrom is mounted, cd and ls /cdrom
shows its there.
using scroll-lock and good timing, able to freeze the
screen as it blips by retrying. Its failing at 18%,
Confirmed by looking at the packages files that it shows
its there in the place where it is, under
pool..../e/e2fsprogs/
either: the copy has been corrupted or it was corrupted getting
to me.
swap drives back, and reboot to working system
remove iso from hdc7
copy iso from workstation to hdc7 and verified its the same,
copied also to /usr/local/share/doc and verified its the same.
remove newinstall from /boot, can just copy from workstation.
swap drives and boot as before with grub disk.
try again:
as above, don't choose optional modules.
FAILED AS ABOVE.
swap drives back and reboot to working system
get new images via ftp from debian.
retreived hd-media vmlinuz's and initrd.gz's and verified the MD5SUMS
file, from the images directory which contains the hd-media directory
and the MD5SUMS file:
md5sum -c MD5SUMS > op 2>&1
there are no lines in op that say that md5 failed, only that
it can't open some of the files (because I didn't retreive
them). Ditto the copies I had previously retreived, and used
above.
retreived the net-inst iso file and the MD5. It matches find. However,
the previous copy I retreived was a few MB small and the MD5 fails. I
guess this teaches me (It did, I know know how to use md5sum). My
previous experience downloading individual files was over a bbs (e.g.
IBM's for OS/2 files) using Zmodem which does CRC as it goes. Why
doesn't FTP use error correction?
copy files to bulwark:
images to /mnt/hdc1/hd-media/vmlinuz and initrd.gz
(kept the whole directory structure with the MD5SUMS,
verified the md5sums on bulwark)
iso to /mnt/hdc7 (also the MD5SUMS and verified file).
accidentally mount hda1 instead of hdc1 when doing the kernels and
accidentally erased the current initrd.img file, as well as the
system-map. I hope the power doesn't go off untill I fix this.
dpkg-reconfigure the kernel, but warns that no system-map. Just to be
sure also reinstalled the kernel image. It boots fine.
stop exim4 on workstation, unmount bulwark's NFS shares
ssh bulwark and shutdown.
Try again:
disconnect origional drive and boot with grub floppy
root (hd0,)
kernel (hd0,0(/hd-media/2.6/vmlinuz /
devfs=mount,dall /
root=/dev/ram0 /
ramdisk_size=12000 /
console=ttyS0,115200n8
debconf/priority=low
initrd (hd0,0)/hd-media/2.6/initrd.gz
boot ok
...
loading components...
unable to load some modules, as above
searching drives for an installer ISO image...
found on ==hda7
loading components, don't choose optional components...
IT WORKED.
detect network hardware:
I have an NIC but that goes up my network not to the
internet. I'm using the net-inst cd so I can get a
functioning base system to then use ppp.
unable to load some modules, as above.
chose no ethernet card, since not using for install
configure the network: no network interfaces were detected,
continue.
hostname: bulwark
free memory (low memory install)
Detect hardware.
prompt for module parms YES
no pcmcia
unable to load some modules, as above.
partition disks:
OOPS, of course, debian correctly sees disk as hdc
we'll see if it works.
edit partition table.
HOLD THE PRESSES!
I didn't realize that JFS was an option.
Just re-read the comparisions in the Linux Gazzette and
JFS is better than reiserfs for what I'm doing.
Since I'm using the file system to hold the image...
aboart install
swap drives,
reboot.
reading ibm's overview on JFS. Default block size is 4096 for
spped over space efficiency. Since bulwark has small disk, will
want smaller block size. I've decided to manually format the
drive.
run aptitude and install jfsutils on both bulwark and
workstaiton.
can't set block size, but JFS still looks like what I want.
at least I can check bad blocks, unlike resiserfs
(I don't know the status of my old disk's
ability to manage bad blocks)
umount /mnt/bulwark/hdc*
on bulwark, unexport hdc shares, comment-out entries in /etc/exports
umount /mnt/hdc*
comment-out entries in /etc/fstab
cfdisk /dev/hdc
change partition types: install program has type as ReiserFS,
yet reiserfs isn't on the type list, neither is JFS. Since IBM
says that JFS is the default type on AIX now, chose type AIX.
man jfs_mkfs shows options but doesn't give details. Download
info from IBM website.
for partitions on bulwark, use:
jfs_mkfs -c -L label /dev/hdc?
-c check for bad blocks
copied files and checked md5sums as before, however, only left the 2.6
images since not room for 2 plus one installed via installer, if any.
stop exim4 on workstation, umount bulwark shares
ssh bulwark and shutdown -hF now
boot with grub: attempt to access block outside partition.
turn off, swap drive, reboot to running system.
Is the problem that grub can't access jfs, or that the type in the
partition says aix?
change type to linux and see what happens.
Try again:
boot with grub: that was it.
low memory mode
scan hard-drives looking for ISO, default modules, no ask parms
unable to load some modules, as above.
first scan can't find ISO, long scan gives up right away,
CAN'T FIND THE ISO.
abort: see if file there, if it is,
then the install can't read JFS untill it gets the ISO
I was hoping it would come under what the installation
manual says about supported systems (among others).
swap drives, reboot,
iso is there on hdc7
interesting that cfdisk shows type as Linux JFS and I can't change it.
make reiserfs on hdc7: mkreiserfs -l /var /dev/hdc7
copy ISO and MD5SUMS and verify
shutdown
swap drives, reboot with grub.
Try again:
found the ISO on =hdc7
loading components, didn't choose optional ones
detect network hardware
prompt for module parms YES
pcmcia no
unable to load some modules, as above
no ethernet card
configure network, get warning about no hardware, continue..
set hostname: bulwark
free memory
detect hardware
prompt for module parms YES
pcmcia no
unable to load some modules, as above
partition disks
manually edit partition table
IDE3 master (hdc) - 546.6 MB MAXTOR MXT-540 AT
> #1 primary 24.3 MB B K jfs /boot
> #5 logical 109.9 MB K jfs /
> #6 logical 132.1 MB K jfs /usr
> #7 logical 188.9 MB K reiserfs /var
> #8 logical 91.3 MB f swap swap
finish and write changes to disk
warning confirm YES
formatting hdc8 for swap
failed to mount hdc7 /var failed
go to vc2 and see what's up
hdc7 is already mounted to access ISO and it doesn't
want to do both.
try to continue without separate /var partition
> #7 logical 188.9 MB reiserfs
finish and write changes to disk
warning about no partition changes or formatting re
pre-exisiting files.
continue YES
install the base system
choose the kernel image: 2.6.8-2-386
UNABLE TO INSTALL SELECTED KERNEL see vc3
cpio error, no space left on device
its automatically cleared it so I don't know
which device, but I'm guessing hdc5 /
next time mount /var on hdc6 and leave /usr on /
back to partition disks
manually edit partition table
reformat so existing files don't get in the way
but don't reformat /boot or will lose hd-media
should have put hd-media with the ISO
> #5 logical 109.9 MB F jfs /
> #6 logical 132.1 MB F jfs /var
> #7 logical 188.9 MB reiserfs
write changes to disk? YES
install the base system
choose the kernel image: 2.6.8-2-386
UNABLE TO INSTALL SELECTED KERNEL see vc3
it's / that's full
continue YES
go to vc2 and try to move stuff around. can't.
tried mv usr to var/local just to get kernel installed, can't.
Next time:
without repartitioning, need a bigger / directory. Once
installed, can move things around temporarily using NFS.
abort
swap drives
boot working system.
determine new partition allotment:
hdc1 24MB jfs /boot
hdc5 110MB jfs /var
hdc6 132MB jfs /
hdc7 189MB reiserfs ISO file
make filesystems as required
jfs_mkfs -c -L ? /dev/hdc?
mkreiserfs -l ? /dev/hdc?
verify cfdisk shows correct
copy ISO and MD5SUMS to /mnt/bulwark/hdc7
verify MD5SUMS
shutdown
try again:
swap drives, reboot with grub
root (hd0,0)
kernel (hd0,0)/hd-media/2.6/vmlinuz \
devfs=mount,dall
root=/dev/ram0
ramdisk_size=12000
console=ttyS0,115200n8
debconf/priority=low
initrd (hd0,0)/hd-media/2.6/initrd.gz
boot
found iso on ==hdc7
loading components, no optional modules
detect network hardware
prompt module params YES
PCMCIA NO
unable to load some modules
no ethernet card
configure network
warning: no NIC
hostname: bulwark
free memory
detect hardware
prompt module params YES
PCMCIA NO
unable to load some modules
partition disks
manually edit partition table
IDE3 master (hdc) - 546.6 MB MAXTOR MXT-540 AT
> #1 primary 24.3 MB B K jfs /boot
> #5 logical 109.9 MB F jfs /var
> #6 logical 132.1 MB F jfs /
> #7 logical 188.9 MB reiserfs
> #8 logical 91.3 MB f swap swap
finish and write changes
warning YES
install the base system
choose the kernel image: 2.6.8-2-386
ERROR INSTALLING THE KERNEL, see vc2
df -h:
mount-point used %
/target 124.8MB 100%
/target/boot 6.2MB 28%
/target/var 41.9MB 40%
install manual says you need a minimum of 110MB for basic
system. What gives?
will have to make a partition table setup that allows the
install but has free-space to make final production partitions.
Need multiple partitions because want to mount /usr ro and keep
/boot small and ro, and / small.
abort install
swap drives,
reboot
remount bulwark shares and start exim4 on workstation
determine new partition allotment:
IDE3 master (hdc) - 546.6 MB MAXTOR MXT-540 AT
> #1 primary 24.3 MB B K jfs /boot
> #5 logical 130.9 MB F jfs /var
> #6 logical 150.0 MB F jfs /
> #7 logical 150.0 MB reiserfs ISO image
> #8 logical 91.3 MB f swap swap
unexport hdc shares on bulwark
cfdisk /dev/hdc
make filesystems as required
jfs_mkfs -c -L ? /dev/hdc?
mkreiserfs -l ISO /dev/hdc7
mount shows correct
df --si shows available:
hdc1 17MB
hdc5 130MB
hdc6 149MB
hdc7 117MB : what is using 34MB? bad clusters?
exportfs -r
copy ISO and MD5SUMS to /mnt/bulwark/hdc7
verify MD5SUMS
df --si shows for hdc7: used 148 MB avial 2.7MB
stop exim4 on workstation, unmount bulwark shares
unexport shares on bulwark
shutdown bulwark
swap drives, reboot with grub
root (hd0,0)
kernel (hd0,0)/hd-media/2.6/vmlinuz \
devfs=mount,dall
root=/dev/ram0
ramdisk_size=12000
console=ttyS0,115200n8
debconf/priority=low
initrd (hd0,0)/hd-media/2.6/initrd.gz
boot
OOPS
Freeing unused kernel memory: 204k freed
attempt to access beyond end of device
ram0: rw=0, want 18890, limit16384
kernel panic: No init found.
didn't verify hda's MD5SUMS.
try rebooting with grub again, just in case
OK.
found iso on ==hdc7
loading components, no optional modules
detect network hardware
prompt module params YES
PCMCIA NO
unable to load some modules
no ethernet card
configure network
warning: no NIC
hostname: bulwark
free memory
detect hardware
prompt module params YES
PCMCIA NO
unable to load some modules
partition disks
IDE3 master (hdc) - 546.6 MB MAXTOR MXT-540 AT
> #1 primary 24.3 MB B K jfs /boot
> #5 logical 130.9 MB F jfs /var
> #6 logical 150.2 MB F jfs /
> #7 logical 150.2 MB reiserfs ISO image
> #8 logical 91.3 MB F swap swap
finish, write changes, warning, YES
df before install base system
device size used avail % mountpoint
hdc7 143.2M 140.7M 2.5M 98% /hd-media/
hdc6 142M 188k 141.8M 0% /target/
hdc1 21.9M 6.2M 15.7M 28% /target/boot/
hdc5 123.3M 148k 123.2M 0% /target/var/
install base system
df before install kernel
hdc7 143.2M 140.7M 2.5M 98% /hd-media/
hdc6 142.0M 96.0M 46.0M 68% /target/
hdc1 21.9M 6.2M 15.7M 28% /target/boot/
hdc5 123.3M 41.2M 82.1M 33% /target/var/
choose the kernel image: 2.6.8-2-386
same error.
Since ISO on #7 just fits, don't need to change it.
#1 for /boot is fine
#5 for /var can be shrunk
#6 for / needs to be bigger.
Need to repartition but leave room to resort afterwards.
Will need to ensure after install, and resorting, that there is still
room in / to upgrade the kernel. To do this, we need to break /usr out
of the / partition. After install, the hardest to move about is /.
Try this:
IDE3 master (hdc) - 546.6 MB MAXTOR MXT-540 AT
> #1 primary 24.3 MB B K jfs /boot
> #6 logical 159.9 MB f jfs /
> #7 logical 120.7 MB f jfs /var
> #5 logical 150.2 MB reiserfs ISO image
> #8 logical 50.0 MB f jfs /usr
> #9 logical 41.3 MB f swap swap
(after install, 120 /var becomes usr, ISO becomes var, 50 /usr
joines with 41 swap to be 91 swap)
Since are only repartitioning empty or disposable partitions, don't need
to abort install.
back to partition the disk
set it up
warning YES
warning: kernel unable to read partition table, Linux won't know
anything about changes untill reboot.
continue
couldn't creat jfs on #7
continue
puts me back to partitioner
undo changes
go back
execute a shell
cfdisk /dev/hdc
no hdc device. Have to find out what its called under devfs
df
/dev/discs/disc0/part7 is hdmedia
cfdisk /dev/discs/disc0
no
cd /dev;ls
cd /discs;ls
cd disc0
/dev/ide/host1/bus0/target0/lun0 !#?@$
ls
disc part1 part2 part5 part6 part7 part8
cfdisk /dev/ide/host1/bus0/target0/lun0
no good, errors on part5
exit
abort
reboot to my emerg bootfloppy and repartition from there
cfdisk /dev/hdc
bad logical partition5, partition overlap
looks like the installer's partitioner messed up the partition
table
swap drives. will have to recopy the iso and boot files after
repartition.
reboot
cfdisk /dev/hdc: same error
cfdisk -z /dev/hdc: manually repartition
make filesystems as required
jfs_mkfs -c -L ? /dev/hdc?
mkreiserfs -l ISO /dev/hdc7
mount shows correct
copy ISO and MD5SUMS to /mnt/bulwark/hdc7
verify MD5SUMS
df --si shows for hdc7: used 148 MB avial 2.7MB
stop exim4 on workstation, unmount bulwark shares
unexport shares on bulwark
shutdown bulwark
swap drives
reboot with grub
OOPS
Freeing unused kernel memory: 204k freed
attempt to access beyond end of device
ram0: rw=0, want 18890, limit16384
kernel panic: No init found.
just like last time.
reboot with grub.
same kernel panic
swap drives, reboot working system
verify md5sums on vmlinuz and initrd.gz OK
reboot with grub: OK
Try again:
found iso on ==hdc7
loading components, no optional modules
detect network hardware
prompt module params YES
PCMCIA NO
unable to load some modules
no ethernet card
configure network
warning: no NIC
hostname: bulwark
free memory
detect hardware
prompt module params YES
PCMCIA NO
unable to load some modules
partition disks
IDE3 master (hdc) - 546.6 MB MAXTOR MXT-540 AT
> #1 primary 24.3 MB B K jfs /boot
> #5 logical 160.0 MB K jfs /
> #6 logical 120.2 MB K jfs /var
> #7 logical 150.2 MB reiserfs ISO image
> #8 logical 50.1 MB K jfs /usr
> #9 logical 41.8 MB f swap swap
finish, write changes, warning, YES
install the base system
base system install error:
the dbootstrap programme exited with an error
(return value 1)
check /var/log or vc3
error: no space left on device
during install of libc6:
failed in buffer_write(fd) (12,ret=-1)
df -h
hdc sz us av % mntpoint
5 151.4 13.2 138.1 9 /target
1 21.9 4.3 17.6 20 /target/boot
8 46.5 39.7 68 85 /target/usr
6 113.5 33.6 79.9 30 /target/var
Don't know what's wrong.
continue
continue
install the base system
warning about existing files on target, to go back and reformat
partition disks
set #s 5,6,8,9 to format
install the base system
same base system installation error.
abort
reboot with grub: OK
execute shell:
df -h: size used avail % mountpoint
tmpfs: 100MB 21.0 79 21 /
#7 143.2 140.7 2.5 98 /hd-media
loop 0 100% /cdrom
free: Total Used Free Shared Buffers
Mem: 37912 35224 2688 0 156
swap: 0 0 0
Total: 37912 35224 2688
partition disks:
same, set to format, except #1,7
finish, write changes, warning YES
shell:
df -h: (note: goes off right hand side of screen
tmpfs: 100MB 13.9 86.1 14 /
loop 0 100% /cdrom
#5 151.4 188k 151.2 0 /target
#1 21.9 4.3 17.6 20 /target/|||||
#8 46.5 140K 46.4 0 /target/|||||
#6 113.5 148k 113.3 0 /target/|||||
free:
Mem: 37912 30036 7876 0 272
Swap: 40812 0 40812
Total: 78724 30036 48688
what if it's not enough swap?
install the base system
same error: was watching on vc2 with free: memory ok. on df-h,
/usr gets to 98% (unfortunatly, there's no 'top' to watch it
closely.
assume that /usr needs to be bigger.
partition disks
IDE3 master (hdc) - 546.6 MB MAXTOR MXT-540 AT
> #1 primary 24.3 MB B K jfs /boot
> #8 logical 109.9 MB f K jfs /
> #9 logical 170.3 MB f K jfs /usr
> #5 logical 150.2 MB reiserfs ISO image
> #6 logical 50.1 MB F K jfs /var
> #7 logical 41.8 MB f swap swap
I don't know why it renumbers the partitions like this.
finish, write changes, warning, YES
kernel unable to read table, tells me to reboot later re-boot
failed to make swap on # 7
abort. Don't know what partition table looks like because
cfdisk can't get partition table.
swap drives
reboot working system
fsck of reiserfs failed. I think that the installer is messing up the
partition table when it goes back after an initial failure.
cfdisk /dev/hdc: bad partition table with overlapping partitions.
use cfdisk -z /dev/hdc, to make it like this:
IDE3 master (hdc) - 546.6 MB MAXTOR MXT-540 AT
> #1 primary 24.3 MB B jfs /boot
> #8 logical 109.9 MB jfs /
> #9 logical 170.3 MB jfs /usr
> #5 logical 150.2 MB reiserfs ISO image
> #6 logical 50.1 MB jfs /var
> #7 logical 41.8 MB swap swap
make filesystems as required
jfs_mkfs -c -L ? /dev/hdc?
mkreiserfs -l ISO /dev/hdc7
mount shows correct
copy ISO and MD5SUMS to /mnt/bulwark/hdc7
verify MD5SUMS
stop exim4 on workstation, unmount bulwark shares
unexport shares on bulwark
shutdown bulwark
swap drives
reboot with grub
only chose modules for regualar IDE drive support
found iso on ==hdc7
loading components, no optional modules
detect network hardware
prompt module params NO
PCMCIA NO
unable to load some modules
no ethernet card
configure network
warning: no NIC
hostname: bulwark
free memory
detect hardware
prompt module params NO
PCMCIA NO
unable to load some modules
partition disks
IDE3 master (hdc) - 546.6 MB MAXTOR MXT-540 AT
> #1 primary 24.3 MB B K jfs /boot
> #5 logical 109.9 MB K jfs /
> #6 logical 171.3 MB K jfs /usr
> #7 logical 150.2 MB reiserfs ISO image
> #8 logical 50.1 MB K jfs /var
> #9 logical 40.8 MB f swap swap
finish, write changes, warning, YES
df -h shows all partitions mounted and ready
free: Total Used Free
Mem: 37912 30216 7696
Swap: 39804 0 39804
Total: 77716 30216 47500
install the base system
done.
Pre-kernel free: Total Used Free Shared Buffers
Mem: 37912 31880 6032 0 2364
Swap: 39804 11976 27828
Total: 77716 43856 33860
Pre-kernel df -h
# Size Used Avail % Mount
5 103.6 16.2 87.4 16 /target
1 21.9 4.3 17.7 19 /target/boot
6 162.2 80.1 82.1 49 /target/usr
8 46.5 41.2 5.4 88 /target/var
choose kernel-image-2.6.8-2-386
YAHOO!
back to menu: next selected is install grub
but first, shell for df -h
5 103.6 55.9 47.7 54 /target
1 21.9 10.3 11.6 47 /target/boot
6 162.2 91.4 70.8 56 /target/usr
8 46.5 42.8 3.7 92 /target/var
----------------------------------
200.4
I saw target go to 65: 10
-----
210MB total space needed so far
install grub boot loader on hard disk
confirm install: YES
finish the installation
complete, remove media
does soft restart, BIOS doesn't see a hd, have to go pull plug
no serial console
hangs up at 'starting cron'
Ctrl-Alt-Del does soft restart, have to pull the plug
add console=ttyS0,115200n8 to kernel line on grub
have to add it to file later
seems to hang forever at cron, but then continues with
error unable to switch to charset maping to ISO-8859-1
with terminal 'serial'
disabling unsupported locale 'en_US'
(That's OK, I'm Canadian Eh?)
get base system setup menu
set up ppp for ISP
configure APT
it can't fetch because PPP isn't on
besides, don't want to fetch packages until after resort
the file system.
select and install packages. It didn't ask anything
configure the mail transfer agent
I'll have to manually set the mail auth stuff
and the network
finish configuring the base system
sensible-editor /etc/inittab
uncomment the vc's and change Ctrl-Alt-Del to -t10 +1
sensible-editor /etc/fstab
add entry for /mnt/hdc7
remove ISO file and MD5SUMS
run aptitude:
Mark as auto anything that I don't know that I want
this removes 29.3 MB of stuff:
at,dhcp-client,ed, groff-base, hotplug, info, ipchains,
iptables,libgdbm3,libssl0.9.7,libtextwrap1,libusb-0.1-4,
locales, man-db, manpages, modutils, nano, pciutils,
pppoe, ppoeconf, tasksel, telnet, usbutils, wget.
I know that I'll end up needing some of these when I install new
stuff after reorganizing the partitions, but by the same token,
some of them would have to be downloaded fresh anyway due to
security updates.
df -H: device size used avail % mount
/dev/hdc5 109M 58M 52M 53 /
tmpfs 20 0 20 0 /dev/shm
/dev/hdc1 23 12 12 48 /boot
/dev/hdc6 171 72 99 42 /usr
/dev/hdc8 49 11 38 23 /var
/dev/hdc7 115 34 117 23 /mnt/hdc7
hdc7 is reiserfs. Make it into JFS:
umount /mnt/hdc7
jfs_mkfs -c -L /var /dev/hdc7
edit /etc/fstab, change reiserfs to jfs
mount /dev/hdc7: used is 148k not 34M
Will do the shell-game shuffle from the other system,
where I can use ssh and NFS rather than serial console.
shutdown -hF now
swap drives
reboot
fsck failed:
e2fsck -ccCDfk /dev/hda1: OK, hda6:OK
reiserfsck /dev/hda7:OK, hda8:OK
it was probably that hdc7 is listed in fstab as reiserfs not jfs
fixed that.
all ok
shutdown -rF now
OK
Its up.
Here's what we want eventually:
Here's how things are now:
df -H
device size used avail % mount
/dev/hdc1 23M 12M 12M 48 /boot
/dev/hdc5 109 58 52 53 /
/dev/hdc6 171 72 100 42 /usr
/dev/hdc7 149 0 149 1
/dev/hdc8 49 11 39 22 /var
/dev/hdc9 40M swap
Here's how we want things set up when finished:
/dev/hdc1 23M /boot
/dev/hdc5 109 /
/dev/hdc6 132 /usr
/dev/hdc7 189 /var
/dev/hdc8 89M swap
So far, / and /boot are fine, and all are formated jfs.
Next step:
move /usr from hdc6 to hdc7
unmount all hdc
cfdisk /dev/hdc and remove hdc6 partition and remake to 132MB
jfs_mkfs -c -L /usr /dev/hdc6
reexport
mv from hdc7 to hdc5
Now have:
/dev/hdc1 23 12 12 48 /boot
/dev/hdc5 109 58 52 53 /
/dev/hdc6 131 70 62 54 /usr
some free space
/dev/hdc7 149 0 149 1 not
/dev/hdc8 49 11 39 22 /var
/dev/hdc9 40M swap
unexport and unmount all hdc
cfdisk /dev/hdc and remove hdc7 partition and remake to 189MB
jfs_mkfs -c -L /var /dev/hdc7
reexport
mv from hdc8 to hdc7
Now we have:
/dev/hdc1 23M 12 12 48 /boot
/dev/hdc5 109 58 52 53 /
/dev/hdc6 132 131 70 62 /usr
/dev/hdc7 189 11 178 6 /var
/dev/hdc8 49 0 49 1 not
/dev/hdc9 40M swap
unexport and unmount all hdc
cfdisk /dev/hdc and remove hdc8 and hd9 and remake to remainder
as swap
makeswap -c -L swap /dev/hdc8
swapon /dev/hdc8 (to test it): swapon -s shows it;
swapoff /dev/hdc8
remove hdc <7 from fstab and exports and remove from /mnt
Disk is now sorted out. Now make /etc files reflect the new reality.
mount -a; exportfs -r
now can edit from the comfort of mc on workstation.
first, ensure grub is set up correctly.
we can also remove the hd-media installation files
edited /etc files to reflect the choices we made on the last system, but
did not blindly copy files. This will have to be done again
when the other packages are installed.
rebooted bulwark and workstation
newbulwark doesn't have modconf; it relied on hot-plug and helpers to
make it automatic-sort of. It didn't see my NIC.
rather use modconf so don't need the hot-plug database
ppp up for apt-setup;
ttyS2 overrun. Don't know why; don't have setserial yet.
probably because the modem is on int 5 not 4
why isn't setserial on the net-install cd?
How set int without setserial? Note: modem is legacy. I
suppose if need by I could remove it and move the
jumper to change the irq.
After reading serial and modem howto, it seems that need
setserial. ldd /bin/seterial, yes bulwark has the libs
required. Sneakernet /bin/setserial over to bulwark.
setserial -g /dev/ttyS2, yes its on irq4
setserial /dev/ttyS2 irq 5
That did it. apt-setup is getting packages.
(added setserial to my boot floppy utility disk)
priorities for packages:
security updates for what I have
debconf english
ipmasq
modconf
setserial
minicom
lrzsz
nfs client and server
ssh "
fetchmail
dnsmasq
bwm
shutdown -rf now
does a soft reset and bios doesn't find any drives and waits for
floppy. Have to pull plug. Grub works correctly with serial
terminal. Try kernel boot param reboot=c (cold), b (bios), h
(hard). Supposed to default to cold, which should work.
tested and works:
dnsmask, setserial, debconf english, ipmasq, minicom, nfs, ssh,
fetchmail, bwm
changed grub's menu.lst, added reboot=b
shutdown -rf now: not expecting it to reboot since running kernel didn't
get reboot=b
shutdown -rf now
didn't work, try reboot=h
shouldn't work, but try reboot=w. didn't work, leave it at h
exim4 is not rewriting addresses.
copied whole /etc/exim4 dir from backups. If it doesn't work,
the problem isn't in exim4
didn't work.
using mail from workstation, ok
using mail from bulwark to dtutty@porchlight,
says permission denied on lookup.
on workstation is owned by root.root ugo+r
on bulwark owned by dtutty.dtutty ug+r
changed it.
OOPS: I don't do things under root unless I really need to.
Unfortunatly, I copied the bulwark /etc/ backup file to my
working directory to be easily accessible. This changed the
ownership and modes to my defaults. Have root redo this.
works.
chrony set up as ntp client to time.nrc.ca and time.chu.nrc.ca
and as server to rest of home network,
chrony set up on workstation as ntp client to bulwark
/usr/local and /var/local copied back, autobackups set up
cron jobs set up.
Reply to: