Hi All,

Trying to help a friend un-frig a situation where they're getting very poor
vendor support for a system that runs on AIX 4.3 on an F50 with SSA disks. I
have some old AIX limited experience and no with SSA.

Problem: two disks failed, initially replaced disks took on new hdisk numbers
since the old config entries weren't removed. Have a backup tape if new disks
can be put back into the vg with the right LV's, etc.

They had one disk fail on Monday. It was identified as pdisk4/hdisk5. That got
pulled and the new drive came in as hdisk 6. Then, Tuesday, after IBM replace
the bad drive,  a second disk failed that day, and it got replaced and came in
as hdisk7. I think it was hdisk2 but don't know the pdisk number or if that
matters.

The vendor who is supposed to dial in and support the machine doesn't have a
clue as to how to instruct their customer to do the failed disk replacements so
they can do a restore.

I have minimal knowledge of the system configuration but it shouldn't be that
hard right? LOL. hdisk0 is rootvg. hdisk1-5 were allocated to another vg say
vg2. The two failed drives were hdisk 5 (now hdisk 6) and hdisk 2 (now hdisk7)

Looking at the system on Wednesday, someone's removed hdisk 2 & 5 so instead of
showing up as defined, they aren't there. lsdev -Ccdisk just shows hdisks
0,1,3,4,6,7 as all being available.

Does anyone have suggestions on how to get them back up? With SSA, I don't know
if  it matters if the new disks are 6 and 7. I think they can just be mapped
with SMIT? Like mapping hdisk6 back to pdisk4?

I've forgotten so much. If they wanted to delete 6 and 7 so they come back in
as 2 and 5, I'm guessing shutdown -m to get into single user mode and then
rmdev -R hdisk6 and then the same for hdisk7? Rebooting the system would then
fill the "holes" from 0-5?

I have no idea if an exportvg was done or if they've lost the "easy" way to get
back the LV config for that vg by deleting the two failed disks.

I'm short on time so I'm wondering how lost this cause is or if there's some
steps that the customer can take to recover their system with the back tape,
and some manipulation of the AIX system, without knowing what LV's were needed
or raw data spaces or whatever SYSBASE was using.

Too many fingers in the pie so any help to get them back up is greatly
appreciated.

TIA,
Mark

--rm the dollars, convert number names to numerals and rm the parens
-- hot(two)new(one) (dash) getnews(one) at  y a h o o . c o  m

On 2004-08-19, Mark Masters <FakeEmail@yahoo.com> wrote:

<snip>

> I have minimal knowledge of the system configuration but it shouldn't be that
> hard right? LOL. hdisk0 is rootvg. hdisk1-5 were allocated to another vg say
> vg2. The two failed drives were hdisk 5 (now hdisk 6) and hdisk 2 (now hdisk7)
>
> Looking at the system on Wednesday, someone's removed hdisk 2 & 5 so instead of
> showing up as defined, they aren't there. lsdev -Ccdisk just shows hdisks
> 0,1,3,4,6,7 as all being available.
>
> Does anyone have suggestions on how to get them back up? With SSA, I don't know
> if  it matters if the new disks are 6 and 7. I think they can just be mapped
> with SMIT? Like mapping hdisk6 back to pdisk4?
>
> I've forgotten so much. If they wanted to delete 6 and 7 so they come back in
> as 2 and 5, I'm guessing shutdown -m to get into single user mode and then
> rmdev -R hdisk6 and then the same for hdisk7? Rebooting the system would then
> fill the "holes" from 0-5?

Being particularly anal about my SSA numbering I always rmdev the [hp]disks 
before swapping in the new disks.  You can still delete the entries in the 
ODM w/rmdev.  If you're unconcerned with which replacement drives get what 
[pd]disk number you can just cfgmgr the system after deleting them and they'll
take the lowest available number for each.  If you want specific slots in
the drawers to be specific numbers you can manually add them provided you have
the connection IDs (cfgmgr, lsattr them, rmdev them, map/mkdev as desired).

BTW:  for defined drives that aren't being used you don't have to go into 
single-user mode or reboot.  If you had to it would be pointless to have
hot-swappable drives, eh?

In short:  lsattr your new drives, write down the adapter(s), disk connection
ID (connwhere_shad), and the enclosure ID.  rmdev them (both hdisk and pdisk),
rmdev the old unused [ph]disk slots, and mkdev p & hdisks, mapping as 
desired, and using the -l arg to set the name to the desired number.

> I have no idea if an exportvg was done or if they've lost the "easy" way to get
> back the LV config for that vg by deleting the two failed disks.

Of course, I haven't actually tried this since we're using RAID
configurations with all of our SSA.  On mirrored system disks I have had to
force a break of a mirror, remove the dead drives, add the new ones (even
if they're the same name) and rebuild the mirrored set.  I don't believe 
you can just "rsync" the mirrored copies on freshly-added drives.  Setting
up the mirror in the first place does some initialisation on the drives in
question, and that data will have been lost.

-- 
	--Arthur Corliss
	  Bolverk's Lair -- http://arthur.corlissfamily.org/
	  Digital Mages -- http://www.digitalmages.com/
	  "Live Free or Die, the Only Way to Live" -- NH State Motto

On Thu, 19 Aug 2004 20:24:35 -0000, Arthur Corliss
<acorliss@bifrost.nevaeh-linux.org> wrote:

>On 2004-08-19, Mark Masters <FakeEmail@yahoo.com> wrote:
>
><snip>
>
<snip>

Good info.

Thanks,
Mark