customize-image.txt 12.1 KB
Newer Older
1
2
3
4
5
Instructions for creating a site-customized image from a "generic" image
provided by Utah.  The basic procedure is:

  - load the image on a testbed machine
  - customize both FreeBSD and Linux filesystems
6
  - save customized image
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42

We will provide you with a generic disk image.  This disk image is one
which has both a FreeBSD partition and a Linux partition.  You will need
to customize both.  By convention FreeBSD is in DOS partition #1 and Linux
in DOS partition #2.

You will need to have the Emulab network-booted, memory-filesystem-based,
FreeBSD systems (hereafter known as the MFSes).  If you have not done this
yet, do it now.  If you do not know what I am talking about, contact
testbed-ops@emulab.net.

In particular, the customization of the disk partitions is done using a
node booted into the "admin" MFS.  If you have followed the Emulab setup
directions and have added nodes to your testbed to the point where they
are now in the "hwdown" experiment, they should already be in the admin MFS.
At this point you should be able to pick one and slogin as root from your
boss machine.

If the node does not respond to slogin and you have serial consoles hooked
up, connect to the console, reboot the node and wait for the Emulab pxeboot
prompt:

	Type a key for interactive mode (quick, quick!)

So hit the space bar (quick, quick!) and you go into interactive mode
where you can tell it to boot from the admin MFS:

	loader:/tftpboot/freebsd

Now on with the show.


A. Load the image on a testbed machine.

We will assume the machine is called "pc1" in the following directions.

43
* Put the Utah-provided generic image in /usr/testbed/images on your
44
  boss machine.  The generic image will be loaded, and the custom
45
  image saved, using this directory.
46
47
48

* Boot pc1 into the admin MFS as described above

49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
* Figure out what your root disk is.  You can look through the boot
  time output of FreeBSD on pc1 to find it.  If you missed that output,
  you can ssh into pc1 and run "dmesg":

      sudo ssh pc1 dmesg

  If you have IDE disks, the disk you want will probably be "ad0".
  If you have SCSI disks, it will be "da0".  For SATA disks, it will
  probably be "ad4".  RAID controllers are variously things like: "ar",
  "aacd", "twed", depending on the controller you have.  If you cannot
  find anything in the output that looks like a disk, you may have an
  unsupported disk controller.  Contact testbed-ops@emulab.net if this
  happens (and be sure to have your "dmesg" output handy!)

  If your disk is anything other than "ad0", there are several files
  under both FreeBSD and Linux that will need to be "fixed".  The good
  news is that the post-disk-load customization pass should do this
  for you.  Just make sure that the Emulab database node_types table
  has the correct value of disktype for each node type.  But, you still
  need to know the disk type for the following steps, so lets set a
  shell variable:

      set DSK=<your-disk-here> # e.g. "ad0", "da0", "ad4"
	or
      DSK=<your-disk-here>; export DSK

75
* Use 'imageunzip' to load the disk.
76
  If all has gone well, the node should be up and accessible via ssh.
77
78
  To copy the generic image onto the test machine disk, do the following
  from boss:
79

80
81
      sudo ssh pc1 imageunzip - /dev/$DSK \
	< /usr/testbed/images/FBSD410+RHL90-GENERIC.ndz
82

83
  Image loading should take anywhere from 45 seconds, to several minutes.
84

85
86
87
88
89
90
91
92
  If the ssh returns with "Killed" then imageunzip ran out of memory.
  By default, imageunzip will consume memory without bound for buffering
  of pending disk writes.  If it grow too big, the system will kill it.
  In this case, retry the imageunzip with "-W <num-MB>" where <num-MB>
  is a number of MB maximum to use for disk buffering.  If the node you
  Using about half of the available physical memory should be safe
  (e.g., if the machine are loading has 512MB of memory, try "-W 256").

93
94
95
96
97
Now you can begin customizing the FreeBSD and Linux partitions.


B. Customize FreeBSD:

98
First, login as root from boss and set that magic disk variable:
99

100
101
      sudo slogin pc1
      set DSK=<your-disk-here>  # you *will* be in csh here
102

103
and mount the FreeBSD filesystems on the disk:
104

105
106
107
      mount /dev/${DSK}s1a /mnt
      mount /dev/${DSK}s1e /mnt/var
      mount /dev/${DSK}s1f /mnt/usr
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138

Now you can update the necessary files as follows.
	
The MFS has a much scaled-down set of binaries.  To get access to a more
full-featured system, you can run binaries from the disk image itself:

     ldconfig /mnt/usr/lib /mnt/usr/X11R6/lib /mnt/usr/local/lib
     set path=($path /mnt/sbin /mnt/bin /mnt/usr/sbin /mnt/usr/bin)

Now update the following files:

* /mnt/root/.ssh/authorized_keys

  Put in local boss root pub key.  Leave in Utah (Emulab) pub key if
  acceptable (if you want/need our help debugging).  Otherwise, remove it.

* /mnt/etc/localtime

  Copy the correct file over from /mnt/usr/share/zoneinfo

* /mnt/etc/master.passwd

  Change the root password.  The password needs to be changed in the
  etc/emulab subdirectory as well:

      chroot /mnt passwd root
      <set password>
      cp -p /mnt/etc/master.passwd /mnt/etc/emulab/

* /mnt/etc/ssh/ssh_host*

139
140
141
  We use the same host key for all images and all OSes.  If you correctly
  customized your MFSes, you have already generated a set of site-specific
  host keys, and you can copy them to the disk with:
142
143
144

      cp -p /etc/ssh/ssh_host* /mnt/etc/ssh/

145
146
147
148
  and then skip to the next bullet item.

  If you did NOT generate host keys for your MFSes, you can generate
  keys now with:
149
150
151
152
153

      ssh-keygen -t rsa1 -N "" -f /mnt/etc/ssh/ssh_host_key
      ssh-keygen -t rsa -N "" -f /mnt/etc/ssh/ssh_host_rsa_key
      ssh-keygen -t dsa -N "" -f /mnt/etc/ssh/ssh_host_dsa_key

154
155
156
157
  This installs them in the disk image, you will still have to go back and
  install these same keys in the sources for your frisbee/freebsd MFSes later
  using the updating instructions in the README file in the MFS tarball.
  So save the keys from /mnt/etc/ssh off somewhere.
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182

* /mnt/etc/emulab/{client,emulab}.pem

  These should have been created on your boss node when you did the boss
  setup.  So from your boss node do:

      sudo scp -p /usr/testbed/etc/{client,emulab}.pem pc1:/mnt/etc/emulab/

That is it for FreeBSD.  Now remount the filesystems read-only so you
can still run binaries but don't accidentally clobber anything:

      cd /
      mount -u -o ro /mnt/usr
      mount -u -o ro /mnt/var
      mount -u -o ro /mnt

and move on to updating the Linux partition on the disk.


C. Customize Linux:

Mount the Linux filesystems (recall that the FreeBSD filesystems are
still mounted on /mnt, so we use another directory):

      mkdir /mnt2
183
      mount -t ext2fs /dev/${DSK}s2 /mnt2
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223

Now you can update the necessary files as follows.
	
* /mnt2/root/.ssh/authorized_keys

  Copy over from the FreeBSD partition:

      cp -p /mnt/root/.ssh/authorized_keys /mnt2/root/.ssh/

* /mnt2/etc/shadow

  Copy in the password hash for root from the FreeBSD password file
  (/mnt/etc/master.passwd) to /mnt2/etc/shadow.  Then copy that file to
   the emulab subdirectory:

      cp -p /mnt2/etc/shadow /mnt2/etc/emulab/

* /mnt2/etc/localtime

  Copy the correct file over from /mnt2/usr/share/zoneinfo

* /mnt2/etc/ssh/ssh_host*

  Copy the host keys you created for FreeBSD above:

      cp -p /mnt/etc/ssh/ssh_host* /mnt2/etc/ssh/

* /etc/testbed/{client,emulab}.pem

  Copy over from the FreeBSD side:

      cp -p /mnt/etc/emulab/*.pem /mnt2/etc/emulab/

Now unmount the Linux filesystem:

      umount /mnt2

and fsck it for good luck.  Actually, not only good luck but also to reset
the time stamp that forces a periodic fsck:

224
      e2fsck /dev/${DSK}s2
225
226


227
228
229
230
231
232
233
D. Saving the customized image

   The "whole disk" image is the one used most frequently, but we also
   create single partition images of FreeBSD and Linux as well.
   From boss do:

       cd /usr/testbed/images
234
235
236
       sudo ssh pc1 imagezip /dev/$DSK - > FBSD410+RHL90-STD.ndz
       sudo ssh pc1 imagezip -s 1 /dev/$DSK - > FBSD410-STD.ndz
       sudo ssh pc1 imagezip -s 2 /dev/$DSK - > RHL90-STD.ndz
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308


E. Testing the image

   Now that you have saved a copy of your customization, you can test out
   the image and make sure it works.

   NOTE: you will need console access to the node at this point, either
   the VGA or the serial line depending on which version of pxeboot you
   are using.

   From your slogin session on the node do:

       reboot

   and wait for the pxeboot prompt to appear on the console as described
   earlier.  When the "Type a key..." message appears, hit space and then
   at the prompt type:

       part:1

   which tells pxeboot to boot from partition 1 (aka, the FreeBSD partition).
   The machine should proceed to boot all the way to a login prompt.  Watch
   for startup problems, like account setup or mount failures.  If all goes
   well, login to the console with the root password you set and reboot again.
   When you get the pxeboot prompt again, type space and then:

       part:2

   to boot into the Linux partition.  Again watch for problems.  If
   everything worked, skip the next step and proceed to "Installing the
   images" below.  If there was a catastrophic failure, you can reboot
   the node into the admin MFS and reload the disk either with the image
   snapshot you made or, worst case, the generic image.  If you just
   need to make some minor changes, make them and proceed with the next
   step.


F. Recreating the image

   If you need to tweak either the FreeBSD or Linux partitions, you will
   need to save the image again.  Doing this properly involves cleaning up
   anything that the Emulab node self configuration might have done.
   While you are running either FreeBSD or Linux you do the following
   from the console:

       shutdown now
       <wait for single user root prompt>
       cd /usr/local/etc/emulab
       ./prepare
       reboot

   As the node reboots, catch the pxeboot prompt and boot into the admin
   MFS.  Go back to the "Saving the customized image" step (D).


G. Installing the images

   Once you have a working image, go back to your boss and do:

       mv /proj/emulab-ops/images/*.ndz /usr/testbed/images/

   and record these initial images in the Emulab database using the
   SQL INSERT commands in setup-images.sql and setup-osids.sql.
   As the database schema occasionally changes, it is important to
   ensure that your DB matches the INSERT commands before doing them.
   The table formats are listed at the beginning of those files.



Dealing with SCSI disks (or RAID).

309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
* For completeness here are the files that need to be changed if you have
  other than IDE disks:

      FreeBSD /etc/fstab: needs "ad0" changed to appropriate type
      Linux /etc/fstab: needs "hda" changed to appropriate type
      Linux /etc/lilo.conf: ditto, plus lilo must be rerun

  Again, you should not need to mess with these files, the post-disk-loader
  script will do it for you.

* For FreeBSD, we have four different kernels that are used for various
  purposes.  Only one of them is generic.  So you will need to build
  customized kernels in order to do "link delays", delay nodes and virtual
  nodes.  The image does include our versions of all these kernels, so
  try those first and maybe save yourself some work.  These kernels support:

      disks: ad, ar, da (with ahc, ahd controllers)
      network: xl, dc, fxp, em, nge

  Copy /kernel.100HZ over to /kernel on a node and try booting it.  If
  that works and all your network interfaces were found, just create a
  new image after copying /kernel.100HZ to /kernel (copy, don't move,
  /kernel.100HZ needs to exist).
332

333
334
335
336
  Otherwise, you need to build your own kernels.  You can look at output from
  the admin MFS (aka GENERIC) kernel if necessary.  Add the necessary driver
  to the various TESTBED configs in the kernel source, and rebuild and install
  the kernels in the image.  Mail testbed-ops@emulab.net for more info.
337

338
339
340
341
* For Linux, it is even more of a PITA.  We currently don't even have a
  generic Linux kernel in the image.  So if Linux doesn't boot in the image,
  you will have to configure/build a new kernel on some external machine.
  If you don't have a Linux machine to do this with, contact us.
342

343
344
[ This file is a copy of doc/customize-image.txt in the Emulab source tree. ]
[ Last updated 11/11/04 ]