On 3/08/10 10:20 PM, Emily Grettel wrote:
Thanks for the reply James,
> If it were my system, I'd export the ZFS volumes containing my data,
> reinstall on the new motherboard, and then reimport ZFS.
I was thinking that too, but unfortunately I've created quite a few
zones and there are quite a few users on the system.
Redoing the entire server will take a week :(
Thanks though, I shall try driver-discuss too!
The essential problem is that your new motherboard will have
different paths to each device.
As James mentioned, you could change the first line of
/etc/path_to_inst, or.....
here's the _unsupported_ totally ugly hack way of getting a
new motherboard up and running.
Before you start, BE VERY GRATEFUL you're running ZFS. (I'll
explain why a little later).
* touch /reconfigure
* poweroff
* replace motherboard
* turn system on
* do whatever bios futzing is needed in order to find your
primary boot device
* at the grub boot menu, select your desired BE, navigate to
the kernel$ line and hit 'e'
* go to the end of this line, and hit 'a' (to add), then add
" -arvs" (ie, a space, then -arvs) and hit escape
* hit 'b' to boot
* Unless you're prompted for where /etc/path_to_inst is,
hit enter each time you're prompted during the boot process.
* When you're asked for a username for single-user mode, type
root and enter your root password.
* Run these operations to test:
format < /dev/null
zpool status -v
zpool import -a
zfs list
dladm show-link
dladm show-ether
The format test will print out the device paths for the
devices which the kernel has probed. Note these for later.
The zpool status -xv test will show you the paths to each
vdev in your pools.
The zpool import -a test will attempt to import as many
pools as can be found. This should work seamlessly, and
you should then see all your datasets in the zfs list test.
The dladm tests will show you what NICs you have installed.
Note the instance numbers - they almost certainly will have
changed from what you have configured with /etc/hostname.$nic$inst.
Change the /etc/hostname.... file to reflect the new instance
number(s).
Also, if you are running a graphics head on this system, and
you've got a customised /etc/X11/xorg.conf, make sure you
check the BusID settings to make sure that they're correct.
Use the /usr/bin/scanpci utility for this.
Now, why should be grateful for ZFS? Because ZFS uses the
cXtYdZ number as a fallback for detecting and opening
devices. What it uses as a primary method is the device id,
or devid. This is closely related to the GUID aka Globally
Unique IDentifier. If you want more info about devids and
guids, you can review a presentation I wrote about them a
while back:
http://www.slideshare.net/JamesCMcPherson/what-is-a-guid
James C. McPherson
--
Oracle
http://www.jmcp.homeunix.com/blog
_______________________________________________
install-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/install-discuss
