dragonhawk at gmail.com
Mon Jan 8 10:34:55 CST 2007
On 1/8/07, Steve Thompson <smt at vgersoft.com> wrote:
> (1) There will be a second PE2900 + Perc 5/E in the same rack. The Perc
> card will have nothing connected to it. If I use hardware RAID-1 on the
> first Perc, how easy is in in practice to replicate the configuration to
> the Perc in the second PE2900 if the first PE2900 or first Perc bites the
PERC's support a concept called "Configuration On Disk" (COD). Any
disk in a RAID set has all of the information about that RAID set
written to it. If you change cards, or even if the NVRAM on the card
fails or is reset, you can load the COD info and all should be well.
> (2) Am I better advised to set up all the drives as single-disk JBOD's and
> use software RAID-1?
Should I fish or cut bait?
Some people favor one, some the other. There are arguments either
way. Some of it depends on the circumstances. It is mostly personal
> I'm personally more comfortable with this since I have used software RAID
> successfully for several years, and I can easily manage the above-mentioned
> failure scenario.
Well, *that* particular failure scenario is well-handled.
The major argument against hardware RAID is that one needs the
controller. If, for some reason, you wanted to move the disk box to a
server *without* a Dell PERC (or an AMI/LSI MegaRAID controller, which
they are based on), you're sunk. Given that you already have a full
spare system in your plan, that seems unlikely.
In general, I find hardware RAID easier to setup, and more robust in
the face of disk failures. With software RAID, there is a period
between power on and OS start that RAID is not available. With
hardware RAID, everything down to the MBR is protected, and appears as
one logical unit. I like that.
> I have no Perc's at present
> and so have no feel for how they work in the real world.
I've never had serious trouble with them. My only complaint was
with the PERC/3i, which was a performance dog (slow).
More information about the Linux-PowerEdge