Golden Image install: disk space errors

[thom_14]

Ok, here's a strange one. I capture a golden image, then turn around and install the same system with the same image, and it bombs - the error message is that I don't have enough disk space! I have the latest cumulative PAx patch, the latest Ignite bits, etc. I've spent some time hunting around the 'net to no avail, so I've come here.

Here's the relevant snippet from the log, and I've attached the relevant config files.

***

* NFS mounting clients directory.
* Using client directory: /var/opt/ignite/clients/0x001083CF1725
* Checking configuration for consistency...
ERROR: The total volume sizes in disk group: "vg00" exceed the available
capacity by 872Mb. For more information run "instl_dbg -D
-d".
ERROR: The size specified for volume: "/" (200MB) is not large enough to
accommodate the amount of software to be loaded into it (233MB). For
more information run "instl_dbg -D -d".
ERROR: There is not enough disk space to allocate for volume: "/usr" (short
by 636Mb). For more information run "instl_dbg -D -d".
ERROR: There is not enough disk space to allocate for volume: "/var" (short
by 236Mb). For more information run "instl_dbg -D -d".
WARNING: Free space (27352KB) in "/var" where /var/adm/crash is located is less
than 25% of system memory (1835008KB). This will likely result in an
inability to capture a single selective compressed dump in the event
of a system crash. Additional space will be required to uncompress the
dump in order to analyze it and to save multiple dumps. You can
increase the size of the "/var" volume on the "File System" tab when
using the advanced user interface.
WARNING: The disk at: 10/0/15/1.6.0 (SEAGATE_ST39236LC) appears to contain a
file system and boot area. Continuing the installation will destroy
any existing data on this disk.

Because of the error(s) above, the user-interface must be
used to correct them... Non-interactive install canceled.

[Michael Roberts_3]

Your log file indicates you are using Ignite version B.5.2 which has known issues and is not the latest version (C.6.1).

There is information at:
http://www.docs.hp.com/en/IUX/workaround5-4.html

Indicating the problems with versions B.5.[0-3]

[thom_14]

Cool, I'll upgrade Ignite and try again; I'd been told by the local admin that we were already at the latest rev, and I was too dumb to verify it myself.

Thanks!

[Kent Ostby]

thorn -- shame on you for believing a co-worker! :-)

I've also seen this problem when someone trys to run this on an older system without going into interactive mode and selecting one of the environments.

[thom_14]

I'm still getting the error! I installed the newest Ignite bits, and created a new image, then tried to install the same system with it. Any other ideas?

>>>

======= 07/27/05 15:47:25 EDT HP-UX Installation Initialization. (Wed Jul 27
15:47:25 EDT 2005)
@(#) Ignite-UX Revision C.6.1.44
@(#) install/init (opt) $Revision: 10.303 $

(snip)

* Processed 70% of archive
* Processed 80% of archive
* Processed 90% of archive
pax_iux: root/users/root/src/perl-5.8.5/libperl.a : No such file or directory
pax_iux: root/users/root/src/perl-5.8.5/miniperl : No space left on device
pax_iux: root/users/root/src/perl-5.8.5/opmini.o : No space left on device
pax_iux: root/users/root/src/perl-5.8.5/extra.pods : No space left on device
pax_iux: root/users/root/src/perl-5.8.5/perlmain.o : No space left on device
pax_iux: root/users/root/src/perl-5.8.5/perl : No space left on device

[Michael Roberts_3]

I'm suspicious that the impact statements in the config aren't correct. On the Ignite server can you run:
/opt/ignite/bin/instl_dbg -v -d -D /var/opt/ignite/clients/0x001083CF1725

this will print out the disk allocation for the client based on the config and the hw.info file.

[thom_14]

instl_dbg -v -d -D /var/opt/ignite/clients/0x001083CF1725
======= Disk and Volume allocation ======
Mount Size Usage Disk / impact amount
Directory (Mb) Group (MB / %) shy (Kb)
----------------------------------------------------------------------
/stand 128 HFS vg00 42 33% 0
(swap_dump) 2500 SWAP_DUMP vg00 0 0% 0
/ 6048 VxFS vg00 5540 91% 0
----------------------------------------------------------------------
Size unallocated from group: vg00: 8680Mb
* Expected physical allocation of volumes on disks:
* 10/0/15/1.6.0 (/dev/dsk/c3t6d0) (vg00)
* 3997 - 135069 KB /stand
* 135069 - 2695069 KB primary
* 2695069 - 8888221 KB /
* 0 KB unallocated
* 10/0/15/1.5.0 (/dev/dsk/c3t5d0) (vg00)
* 8888320 KB unallocated

And the impacts section of the cfg file:

impacts = "/" 20Kb
impacts = "/.sw" 5Kb
impacts = "/dev" 14Kb
impacts = "/etc" 43954Kb
impacts = "/home" 4Kb
impacts = "/opt" 3196390Kb
impacts = "/root" 142761Kb
impacts = "/sbin" 46692Kb
impacts = "/stand" 43606Kb
impacts = "/tmp" 1416Kb
impacts = "/usr" 1396182Kb
impacts = "/var" 154751Kb


I'm still not clear on the role of the impacts section. I know how to generate the info, and I dutifully plugged it in, but what does IG do with it? In this case, I just have one big / file system; what effect do the other filesystems mentioned in the impacts section have on the vg creation? Is it my monolitic root file system that's screwing things up?

[Michael Roberts_3]

The impacts section allows Ignite to check that the software selected to install will fit on the mount points/file systems that are configured. Since you can change the layout of the target system it is divided by top level directories since those are likely mount points if you change the configuration. Also the imapacts are additive in the case of multiple software sources, if you install from two SD depots and an archive the impacts of the all of the sources on say /var are added up.

The error messages indicate that the config is asking for separate mount points for at least /,/usr, and /var . So it appears that your desire to have one large '/' is not what is actually happening.

Your config /var/opt/ignite/data/Rel_B.11.11/GoldFinger_cfg
contains a 32 bit selection and a 64 bit selection and they are not configured the same. Which one are you using?
"golden image - 32 bit OS"
OR
"golden image - 64 bit OS"

The 32 bit selection indicates it is a gzip'd tar file "archive_type = gzip tar" and the 64 bit selection indicates gzip'd cpio "archive_type = gzip cpio". If you are using the 64 bit selection you should make sure the archive_type is correct for your archive. The default behavior of make_sys_image is to create gzip'd tar archive.

vg00 has two disks and the first one 10/0/15/1.6.0 is 100% allocated. Since '/' must be created contiguous it cannot span to the second disk in the volume group, same with /stand and primary swap. So if you allowed /var to be a separate mount point (good idea any way so run away logs do not fill up root) it looks like that would give you enough space.

( sorry it took so long to get back to you)

[thom_14]

Ok, I re-installed the box with a traditional /var, /opt etc layout, re-grabbed the golden image, then re-laid it down - the problem went away.

So, I guess I need to stick to traditional file system layouts.

Thanks!