[SunHELP] [50% solved, but need help ; -)]Sol10 - fsck problem / on SunSparc-AXMP

[Michael Karl]

Hi all,

thank you for your tips and tricks.

My plan now:
1. metadetach all submirrors ... so the "mirror-disk" is out of work.
2. fsck on all partions of the "mirror-disk", because metadetach has 
made them unclean.
3. mount the / of the "mirror-disk" and edit /etc/vfstab and change all 
entries /dev/md/dsk/dx to /dev/dsk/cxtxdxsx. At last I uncomment or 
delete in the /etc/system the entry
* Begin MDD root info (do not edit)
rootdev:/pseudo/md at 0:0,0,blk
* End MDD root info (do not edit)
4. I don't make any changes on the "first-disk" (it's still working); 
reboot and boot "mirror-disk" and test it.
5. If the "mirror-disk" is booting and working well, I reboot the system 
and with starting the "first-disk".
6. I'll install the recommended patch-cluster ... if this is working, I 
metattach the "mirror-disk"-partitions.
7. If not, first I have to analyse the errors ... and if I should find 
no error-solutions, I boot the "mirror-disk", have a look on the 
md-devices ... normally all md-devices still exists ... change the 
/etc/vfstab except the root-entry, make metaroot and lockfs ... reboot 
the "mirror-disk" and metattach the "first-disk"-partitions.

BTW. I'm thinking, that I need the way back, because the actual Solaris 
10 10/08 boot-install-cd let the machine crash with panic during the 
boot-process.

Any comments ?

Thank you all

Michael

Steve Sandau schrieb:
> listmail wrote:
>> I was speaking in context of my own setups and habits.  I prefer to 
>> clear out the OS mirror configs. to make the drives truly free 
>> standing for which either or could be booted freely without disksuite 
>> worrying about the other disk's state, along with other flexibility 
>> that it brings.  In your case you don't want to much since having a 
>> stripe.
>>
>> * Anyway *, Perhaps in your case you may want to physically 
>> disconnect c1t0d0 to leave it in its current state and then boot 
>> c1t1d0s0 to do the cluster test? (installing the boot block and/or do 
>> config changes from single mode cdrom on it if needed).
>>
>> Michael Karl wrote:
>>> Hello,
>>>
>>> http://docs.sun.com/app/docs/doc/817-2530/6mi6gg883?a=view
>>>
>>> 1. I didn't make metaroot and metaclear to the mirrors on root ... 
>>> I'm asking why ?
>>> 2. If I metadetach all submirrors of the boot-disk, I'm thinking 
>>> that I must be able to boot directly the "first-boot"-disk and the 
>>> "mirror-boot"-disk via ok-prompt ... but now I could only boot the 
>>> "first-boot"-disk without "1." well ... but the 
>>> "mirror-boot"-disk-root is not clean for booting ... may be it's not 
>>> in sync.
>>> 3. I don't remember, how I solved this problem in the past. I'm 
>>> thinking: metadetach the submirror, upgrading with patch-cluster the 
>>> "first-disk", looking for working of the "first-disk", metattach the 
>>> submirrors if everything is working ... if not, I boot the 
>>> metadetached mirror and make with metattach the sync to the 
>>> "first-disk" to the original state.
>>>
>>> Where is my mistake now ?
>
> I think the advice to clear all metadata information is probably a 
> good one.
>
> Once the box is set to boot off a mirrored set, there is an entry put 
> in the /etc/system file to that effect. The box will boot off either 
> disk but will mount the root metadevice, which means it will use 
> whichever disk is still part of the root mirrored pair for the root 
> filesystem.
>
> To clean out the metadata information, mount each of the root mirrors 
> (one at a time) and remove all metadata references in /etc/system and 
> in /etc/vfstab. The /etc/system line (md at 0 or something like that) 
> needs to be removed, and in /etc/vfstab all the metadevices (md0, 
> etc.) need to be replaced with c0t0d0s0 or whatever the appropriate 
> slice is on that disk. So, on the first disk you'll use c0t0d0s0 and 
> when editing the vfstab on the second disk you'll use c0t1d0s0 (or 
> whatever numbers are appropriate for your setup).
>
> Then you wil be able to boot cleanly to either disk, check its status 
> and patches, and be able to work with whichever disk you choose 
> without affecting the other disk.
>
> When you are done you will need to completely remirror the disks of 
> course, including running the metaroot command to put the information 
> back in /etc/system.
>
> Steve
> _______________________________________________
> SunHELP maillist  -  SunHELP at sunhelp.org
> http://www.sunhelp.org/mailman/listinfo/sunhelp