School of Computer Science Wiki

You've loaded an old revision of the document! If you save it, you will create a new version with this data.

Sorry for the long delay in replying, I hope that this is still useful to you.
Before I get to your questions I have a few other things:

1.  I released all of the tickets that were assigned to me in the ticketing
system.  You may have noticed.  I still have super however.  At you leisure
you can go in an remove it.

2.  I have also gone though an organized all of my emails.  Something that
I've been meaning to do you the last two years.  It so much simpler when you
can organize with the delete key.  I have pulled all of the emails with
regards to software licenses into one mail box, and I will attach that mailbox
to this message (it's mbox format).

On to your questions:

Fang:  That's a good plan.  You'll need to partition the disk first though.
Since this will help me answer your question with fudge I'm going to write out
a step-wise procedure for you.  The keys to the cabinet are black.  There is
one on the shelf in 115 and one on my former key-ring.

Part one:  Prepare the hard disk.  You can do this in fang or another machine.

1.  Partition the disk.  Fang is partitioned as GPT with a BIOS boot partition
(you could choose to use MBR though -- just adapt these instructions to
fdisk).  Use gdisk to partition the drive (gdisk /dev/sdX replacing X with the
new disk).  I recommend that you look though the help before starting (?) as
issuing the wrong command here an really ruin your day.  Before creating
partitions I would print the partition table (it should be blank) to make sure
that I'm partitioning the correct disk.  I do not remember how exactly to laid
out fang.  I'm going to assume that I used a BIOS Boot partition, a EFI system
partition (future proofing), a 8GB swap partition (2x memory it's a guideline,
not a rule), a 20GB root partition and a data partition consuming the remainder
of the disk.  You may wish to look at fangs current layout first, and adjust
this as you see fit.  Layout the partitions as follows (note that the text in
[] and () is additional information for you, and will not be a part of the
interface):

- [EFI System] start: default (should be 2048) end: +100M type: EF00
	- [Linux Swap] start: default end: +8G type: 8200
	- [Root] start: default end: +20G type: 8300 (the default)
	- [Data] start: default end: default type: 8300 (the default)
	- [BIOS Boot] start: default (should be 34) end: default type: EF02

Once you're done with that sort the partitions, save and exit.  Notice how we
created the fist partition last?  The reason for that is gdisk like to align
partitions on 1MB boundaries which helps with new "Advanced Format" disks,
which have 4k sectors (instead of 1/2k sectors).  It's easiest to let gdisk do
things its way then add the bios boot partition at the end.  We need 1k for
GRUB, we're giving it close to 1MB, so this works just fine.

2.  Format the disk.  The BIOS Boot partition is not formatted.  The EFI
System partition is formatted vfat (mkfs -t vfat), the swap partition is thew
swap file system (mkswap) and the other two are ext4 (mkfs -t ext4)

Part two:  Copy the data

1.  Boot off Hedwig (if you haven't already).
2.  Create mount points in /tmp and most the two root file systems
3.  Copy the data (cp -a)
4.  unmount old disk

Part three: make fang boot

1.  Bind mount /dev /sys and /proc into the new volume
2.  Chroot in to the new volume
3.  fix the fstab.  
	- You will need to replace the UUID's as they have changed.
	- Use blkid to get the new ones (you may need to do this outside of the chroot)
4.  run grub-install
	- you need to tell it where to install, it will be the root in the new
	  disk.  /dev/sda is an example, but it will vary with the system
5.  rebuild the initrd /usr/sbin/update-initramfs -u -k all

Part three: boot an cleanup

1.  Boot fang off the new disk, and if it works put everything back together
2.  Make sure that the LDAP server (slapd), DNS server (bind) and KDC are
   *not* running.  If they are you will get some strange complains about
   login failures.
3.  If you want to make the auth components for fang work:
	- copy over /etc/ldap/slapd.d from norberta
	- clean out /var/ldap
	- start slapd
	- wait a few minutes, then test it by doing
		- slapcat > file1
		- slapcat > file2(on fluffy)
		- diff file1 file2 (you should get no output)
	- if the test worked, then it's replicating properly and you can start
	  the KDC.
	- For the DNS 
		- Make the permissions and ownership on /etc/bind match norberta 
		- Copy over /usr/local/sbin/updatezone.sh (I think that the 
		  name, it will be the only file in that directory -- it's a 
		  script that I wrote since the one in the Debian package is 
		  busted.)
	- Adjust the ldap2zone settings in /etc/crond.d to match norberta.
	- It will create zone files on the hour, once you are sure that it is
	  working start bind.
4.  You're done.

Fudge:  This is similar to fang above.  Your goal here is to make fudge look
mostly like hopkirk.  So feel free to use hopkirk as your guide and steal some
configuration from it.  Keep in mind my warning about blindly copying config
files and other stuff though.  Also note that you can ssh into hedwig, as
working in the iLOM for this sucks.  As for the VM's on Fudge, there should be
the old skeeter, the old Dursley, and one called toby (which can be deleted).
There shouldn't be anything else.

Part One:  backup

1.  Make sure that there is nothing on Fudge that you want to save.
2.  Make a copy of fudges ssh keys (/etc/ssh/*key*)

Part Two:  Disk setup.

You can actually steal this part pretty much verbatim from part one for fang
above.  Keep in mind that the swap partition size might be different, check
the current fudge config first.  Also use hopkirk as your guide for partition
sizes.  Finally note that fudge has two disks configured as a hardware RAID 1,
hopkirk had two disks configured as a software RAID 1, so they will look
different (when you do moody, you'll want to setup a software RAID 1 there).

Part Three: Install the Operating System

1.  create a mount point in /tmp
2.  mount your root partition on your new mount point
3.  boot strap the operating system:  debootstrap wheezy <mountpoint>
4.  rbind mount /dev /sys and /proc into your newly created drive
5.  chroot into your newly created drive
6.  copy over from hopkirk:
	/etc/apt/apt.conf.d/*
	/etc/apt/sources.lisr.d/*
	/etc/apt/preferences.d/*
7.  delete /etc/sources.list
8.  aptitude update
9.  install a text editor so you don't go insane
10. set roots password (passwd)
11. create sysadmin and set it's password
	- I usually use vipw for this.  You will also need vipw -s
	- edit /etc/group to put sysadmin in the correct groups (look at
	  hopkirk, but note that some groups on hopkirk don't exist yet)
	- cp -R /etc/skel /home/local/sysadmin
	- fix the ownership and permissions on /home/local/sysadmin
	- Or you can use the adduser program
12. Setup /etc/network/interfaces (use hopkirk and the old one as your guide)
13. Create /etc/fstab (use hopkirk as your guide), and remember blkid
14. dpkg-reconfigure debconf 
	- set the priority to low
	- leave the interface at default
15. dpkg-reconfigure tzdata 
	- select America then Toronto as your time zone (unless you want all
	  of you time stamps in UTC)
16. launch aptitude and do the following:
	- install linux-image-amd64
	- install grub-pc
	- install openssh-server
	- install sudo
	- look though the recommends on these and add anything else that's
	  needed (use hopkirk as your guide)
	- Mark everything as Automatically installed (M with "--- Installed
	  Packages" selected
		- Aptitude will want to purge just about everything right now
		- Using hopkirk as your guide select the packages that you
		  want to keep (+)
		- There will be some packages that are marked as automatic on
		  hopkirk, which fudge want's to delete.  Try a + them a M on
		  these.  If fudge still wants to delete them it means that
		  the package that depends on this on on hopkirk isn't yet
		  installed on fudge.  Just + it and deal with it later (if 
		  you care, you don't have to it's just housekeeping).
17.  Copy the ssh keys that you backed up into /etc/ssh
18.  Put your ssh key in sysadmins .ssh/authorized_keys
19.  Copy over hopkirks /etc/ssh/sshd_config
20.  At this point you have a bootable base Operating System installed with an
    ssh server.  The hard part should be done.  Go ahead and reboot the
    machine (remember to remove the USB key).

Part Four:  Final setup

1.  Install all packages that are present on hopkirk but not on fudge.  You
   can skip mdadm though as that is the software RAID, which fudge doesn't 
   need.
2.  Do any necessary configuration (there shouldn't be much).
3.  Do any housekeeping (check /etc/groups and any installed packages).
4.  Transfer over VM's.
	- I would do dursley first and test it for a few days to make sure
	  that it is working and everything is good before moving skeeter

That should be it.  Every that I bootstrapped a a machine I usually forgot one
of these steps and ended up going back in once or twice (booting off of the
USB key) to fix it.  It was usually things like setting roots password or
configuring /etc/network/interfaces.  Needless to say, it was rather annoying.

Good luck with it.

James

Edit summary

Note: By editing this page you agree to license your content under the following license: CC Attribution-Noncommercial-Share Alike 4.0 International