[ILUG] Booting a broken raid array

[Kevin Philp]

Conor Wynne wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Kenn Humborg wrote:
>   
>>> - partition as follows:
>>>
>>>   - /boot goes onto the compact flash card, set as the booting device
>>> in the bios
>>>   - create a slice of swap on each of the disks (say 500MB - 1GB each)
>>> at the start of the disk
>>>   - fill the rest of the disc with a software raid partition
>>>   - repeat for additional disks
>>>   - build desired raid level MD device on the partitions
>>>   - put a single LVM volume on top of that (for future expansion)
>>>       
>> If I read this right, your swap space is not on mirrored or
>> raided storage.  So, when one disk fails, the system won't
>> be able to swap data back in from that disk.  
>>
>> While this shouldn't cause a kernel crash in Linux (since
>> the kernel never swaps out its own data), it will crash
>> any user-land process that needs to swap in from that swap 
>> space.
>>     
>
> it's like when the OOM killer invokes, avoid a panic by killing off
> unimportant processes like your databases etc :-)
>
> To me that's just as bad as a panic as your going to end up rebooting
> either way.
>
>   
>> Your setup is protecting the data on the disks, but is not
>> really giving you any higher availability.
>>     
>
> Are we talking about a production environment or a home setup here?
> I assume it's home, and in which case all I would bother RAIDing
> anything other than the data. Even then I wouldn't bother wasting the
> power. I would use an offline backup like a USBHDD jobby. Only switching
> it on when required for a quick rsync, you'll find the disk lasts longer
> that way too!
>   
Its a small office setup - About 10 people - data files shared by NFS 
from the server. Currently the files are synced off-site every night 
using rsync over ssh and then once a week copied to an external hard 
drive, also held off site. So I can restore the system if it fails, 
given time. We have had to do this in the past when the server hard 
drive failed so I am quite happy our back up system works. However it 
takes time and also relies on me being there - as the only vaguely 
computer literate person in the company. As the company expands we need 
a better system.

So to make things a bit more robust I want to migrate the server to a 
RAID array for starters - I reckon hard drive failure is our most likely 
failure point. At a later stage we are thinking of having a second 
server on stand by as a backup.

Any suggestions welcome - I am no expert here.


> By all means, backup your /etc (it's generally tiny) and the data.
> How long does it take to re-install and restore data? If it's for home
> use, time is on your side. Production is a another matter.
>
>   
>> Later,
>> Kenn
>>
>>
>>     
>
> Conor.
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.7 (MingW32)
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
>
> iD8DBQFJcHIl0c4nF4jv4ukRAlW2AJ0d+4Q1cyZWvINpFFOu4iN09UkUsgCeNZk5
> SR6VqU3nH+jA31FrZzZU0C0=
> =V3v2
> -----END PGP SIGNATURE-----
>
>
>