raid and power off savety

raid and power off savety

[Schuett Thomas EXT]

Hello,

I am sorry for having to ask a question you might rate very stupid,
but I really want to know it: 

If you think through system crash scenarios, what types of chrashes 
are you are thinking of? Do you only consider harddisk faults, or 
do you consider power failures as well?

Or more extreme: Does RAID gives you savety on HD HW failures, but may
make things worse for power failures (hard power switch off)?

Or more academic: Does raid have some journaling of its dirty cache 
(and/or whatever would need to be journaled) in a non volatile way,
or isn't there any need for it for any tricky reason?

thomas

Re: raid and power off savety

[Luca Berra]

On Tue, Mar 22, 2005 at 12:29:57PM +0100, Schuett Thomas EXT wrote:
Define RAID,
I will answer for RAID = MD driver, but hw raid solutions might have
different answers (eg battery backed up cache)

>If you think through system crash scenarios, what types of chrashes 
>are you are thinking of? Do you only consider harddisk faults, or 
>do you consider power failures as well?
md protects against disk failures

>Or more extreme: Does RAID gives you savety on HD HW failures, but may
>make things worse for power failures (hard power switch off)?
a power failure _might_ make things worse

>Or more academic: Does raid have some journaling of its dirty cache 
>(and/or whatever would need to be journaled) in a non volatile way,
>or isn't there any need for it for any tricky reason?
there isn't one for lack of a non-volatile storage for dirty cache

-- 
Luca Berra -- bluca@comedia.it
        Communication Media & Services S.r.l.
 /"\
 \ /     ASCII RIBBON CAMPAIGN
  X        AGAINST HTML MAIL
 / \

Re: raid and power off savety

[Gordon Henderson]

On Tue, 22 Mar 2005, Schuett Thomas EXT wrote:

> Hello,
>
> I am sorry for having to ask a question you might rate very stupid,
> but I really want to know it:
>
> If you think through system crash scenarios, what types of chrashes
> are you are thinking of? Do you only consider harddisk faults, or
> do you consider power failures as well?

I consider power failures too. In theory, you really ought to have UPS
systems, *and* have them connected to the server(s) so they can monitor it
and graceful shutdown, but with the best will in the world this isn't
always possible.

Tonight, I've had a Linux based router power cycled 3 times in a
thunderstorm )-:

reboot   system boot  2.4.23           Wed Mar 23 20:23          (00:59)
reboot   system boot  2.4.23           Wed Mar 23 20:02          (01:20)
reboot   system boot  2.4.23           Wed Mar 23 19:57          (01:25)


It's located in a open-sided barn in a farm the middle of no-where (part
of a community WiFi broadband network) It has 2 x 80GB drives (it runs a
squid cache amongst other stuff) which are all RAID-1 partitions. (and
ext3)

I used to have a UPS on it, but that fried over a year ago in another
thunderstorm. Since then we've not had the money to buy a new one to
replace it, so we've just suffered the odd outage and hoped for the best.

> Or more extreme: Does RAID gives you savety on HD HW failures, but may
> make things worse for power failures (hard power switch off)?

So-far, I've not had a problem with a hard power off - on this or other
routers in this network, or on other servers I have with s/w RAID. I'm not
saying I'll never have a problem, but maybe so-far I've just been lucky.

Gordon