[CentOS-docs] Updated How to Setup a Software RAID on CentOS 5

Wed Apr 29 17:56:21 UTC 2009
Ned Slider <ned at unixmail.co.uk>

Ed Heron wrote:
> From: "Phil Schaffner", Tuesday, April 28, 2009 5:36 PM
> 
>> I have attempted to address all comments:
>>
>> http://wiki.centos.org/HowTos/SoftwareRAIDonCentOS5
>>
>> Give me your best shot! :-D
>>
>> Phil
> 
>   There's always going to be an argument about whether to put /boot and swap 
> on RAID.  It's all about performance most of the time being slightly better 
> versus stability in the event of device failure.

I can't think of a good argument for not having /boot on the raid1. 
Presumably performance isn't an issue as the contents of /boot gets read 
once at system boot (other than maybe fast booting performance, but 
that's not really going to be an issue for a server running raid1, right?)

SWAP on raid0 is a bad idea as one drive failure may cause the system to 
fail to boot. I guess SWAP on raid1 is the safe option. If SWAP 
performance is critical, then maybe two independent non-raid partitions, 
one at the start of each drive, and set to the same priority would be a 
better solution (effectively giving stripped raid0 performance). 
Presumably then a drive failure wouldn't prevent booting but would 
result in a warning that one of the SWAPs was unavailable (assuming the 
system could function fine with the size of the remaining SWAP)?

Hopefully someone can verify my logic here.

> 
>   What's the disaster recovery plan here?
> 
>   Obviously, if the second drive fails, there's no issue.  Standard removal 
> and eventual addition of replacement device(s).
> 
>   If the first drive fails, are we hoping the computer will boot off the 
> second drive or are we moving the second drive to the first interface?
> 

If grub is present on the mbr of both drives, then the system will 
remain functional if *either* drive fails without any further 
intervention. I would think this is the ideal.

>   Is it outside the scope of this document to describe and test disaster 
> recovery?  I think it is.  I'm just making a note to suggest a further 
> complimentary page at some time in the future...  (though, at this time, I 
> am NOT volunteering to write it)  Is there a wiki page todo list somewhere?
> 

I would be in favour of extending the current page to include 
testing/recovery information in the event of a drive failure. If the 
information is not specific to only raid1, then maybe a separate page is 
warranted.