RAID5 + 1 extra disk HOW

[Bill Davidsen]

roland wrote:
> On Sat, 11 Oct 2008 23:56:34 +0200, Bill Davidsen <davidsen at tmr.com> wrote:
> 
>> roland wrote:
>>> Hello,
>>>  I have a  server running on RAID5, on which I would like to add a 
>>> 4th disk, with a seperated filesystem. I need to have some workspace. 
>>> The same disk could be used to replace a failing disk.
>>>
>> This is a spare disk, and can be used only as such. I'm not sure what 
>> you had in mind by "seperated filesystem," but you can't install 
>> another drive such that Linux will use the drive in two ways at once. 
>> If that was your intent it pretty much isn't going to work, and if the 
>> spare was deployed all data on it would be lost.
>>
>>> Stuart Sears explaned me howto install this disk as a hot spare, but 
>>> it looks to me a little bit to dangereous. This server is serving 20 
>>> workstations, so I need to do this without taking to many risks. I 
>>> admit to be a novice on the matter of diskmanagement. I always do 
>>> standard installation of a filesystem.
>>>
>> Assuming you just want to add a spare, something like this works fine:
>>     mdadm --add /dev/md0 /dev/sdf1
>> Then a cat of /proc/mdstat should show the device added, running as a 
>> spare. I did that, then failed one of the existing drives, and the 
>> array rebuilt on the spare just fine, and I got a message from monitor 
>> mode saying it happened and was fixed.
> 
> I have to mention that the raid5 was build with the IBM utilities on 
> this server x3500. So I suppose this mdadm wouldn't work. Or am I wrong?
> Secondly, I need extra space only as a workarea.
> That's why I thought, maybe I can:
> - insert a 4th disk and add a filesystem, that isn't part of the raid5.
> or
> - insert a 4th and 5th disk and configure it as a raid1, to be used for 
> non critical applications
> 
> In case of failure of the raid5, I could could use the 4th disk to 
> replace the failing disk of the raid5 and let the ibm utilities rebuild 
> the raid5
> 
> Does this makes sense?
> 
Well, I understand it now. ;-)

Ask IBM how to do that, you are well out of anything related to the distribution 
and into the hardware specific area. I'd love to help you, but the only IBM RAID 
controller I have ever used as anything but a box of drives was the "ServeRAID" 
unit used in rack mounted LVD SCSI server systems. Love the controller, but it's 
pretty uncommon.

I would say you can easily use other drives outside the RAID array, adding them 
is a *very* specific issue with not only the controller but the firmware 
version. Even if I was still running servers for an ISP I would be asking IBM 
for docs before trying anything fancy. Most IBM stuff will run without the bad 
drive, send you a message, and either power down the drive or blink the light on 
the bad one, so you can pull the dead drive, plug in a spare in hot-swap 
carrier, and it will recover by magic.

Ask IBM about this.
>>
>>> So if someone could tell me howto do this in a simple way. I searched 
>>> the internet for a simple howto, but was unsuccessful. I have to do 
>>> this at a client in France, few hours driving
>>>  Could someone keep me from having a sleepless night?
>>
>> I would test the existing array before counting on it to be okay, just 
>> in case a bad spot developed.See the wiki 
>> (http://linux-raid.osdl.org). Oh, and there is a linux-raid mailing 
>> list where there are lots of us who have found problems the hard way 
>> and want to share.
>>
> 
> 
> 


-- 
Bill Davidsen <davidsen at tmr.com>
   "We have more to fear from the bungling of the incompetent than from
the machinations of the wicked."  - from Slashdot