PDA

View Full Version : Q re raid on GA-K8N-ultra9 - geforce4


old man
December 22nd 05, 12:57 AM
This board has both Nvidea raid & Sil raid
If, assuming mirror, raid1, do either of these have the capability of
detecting an individual hd failure and to continue running the sys, without
having to break the raid and boot from the functioning hd?
ie The sys continues operating if an individual raid hd fails, without any
user imput - other than acknowledging the failure.
The mobo manual is not clear on this subject.

Mercury
December 22nd 05, 01:33 PM
Yes if RAID 10 (or any other real RAID) - seemlessly. That is the objective
of RAID - requndancy or discs, controllers, cabling, PUS's etc. and the
ability to continue ifr any 1 of these items fail - the degree of failure
you system cn cope with will be dependant on its config: for production I
always have dual controllers, Server PSU's , cables, RAID 10 with 1 string
of discs on 1 controller and the second on the second controller with
redundant HDD PSU's + extra fans, fan / volatge monitoing, email / pager
alerts + Hot standby disc drives + Tested spares. A lot can fail in that
config and it will still go & it will tell you of failures & recover
completely gining you the ability to hange out dead HDD's with no down time
at all.

RAID 1, 10 or 5 does this as do other forms of RAID, *NOT* RAID 0.
As soon as you get a disc falure using any RAID (raid 0 is not real raid),
you should have Either a Hot spare configured that will already be being
synchronised in the back ground maintaining a running system with only some
degradataion in performance.

Or, on lower end RAID cards the system will continue to run - this is what
real RAID is designed for. But you *must* IMO address the issue ASAP and
Know in advance for each controller what to do when a disc fails. The longer
you run RAID in a degraded mode the longer you run the risk of total loss as
you are back to** No Raid and out naked in a storm.

**often, but not always - you can have 2 discs fail in a four disc RAID 10,
3 in a 6 disc RAID 10 etc but the probability of such nice failures drops
off.

Paul wrote a few words about this the other day - have a look.

The essential points:

You *must* have a good backup reqime regardless.
Discs in RAID configs fail. RAID controllers fail, PSU's fail and kill the
lot - a 2nd disc can fail in close succession to a first so fix promptly, be
prepared for the worst case.

Rehearse faiures and know what to do in advance before you go into
production. You won't want to be learning this when an array fails.

So read around in this NG, see Paul's commencts and some of my own wafllings
& never use "RAID 0" unless you can live with complete data loss of what is
stored on such volumes (EG scratch files in video processing, large read
only databases loaded anew monthly etc).

HTH



"old man" > wrote in message
...
> This board has both Nvidea raid & Sil raid
> If, assuming mirror, raid1, do either of these have the capability of
> detecting an individual hd failure and to continue running the sys,
> without
> having to break the raid and boot from the functioning hd?
> ie The sys continues operating if an individual raid hd fails, without any
> user imput - other than acknowledging the failure.
> The mobo manual is not clear on this subject.
>
>

old man
December 22nd 05, 08:37 PM
Thanks,
I asked since I previously had raid 1 on an Sil cntrl.,Asus mobo A7N8X
delux, when 1 hd failed, the sys went down and I had to break the raid
before I could reboot, trying each disk in turn to discover which had
failed, which it did fine except 2months of data, and some installed apps
were missing!
It seems the mirror had'nt been in sync. The raid utility software had given
no warning of any problems, and had allways shown the mirror in sync!
I'm currently using an Adaptec raid cntrl card, in mirror, which I 'know'
will continue running the sys if a hd dies. I'll be adding a hot swap spare
shortly, which will automatically replace/rebuild if a hd fails. BTHW I
still have both onsite/offsite data backups.- dont think I can quite warrant
the amount of redundency you seem to have.
TY

"Mercury" > wrote in message
...
> Yes if RAID 10 (or any other real RAID) - seemlessly. That is the
objective
> of RAID - requndancy or discs, controllers, cabling, PUS's etc. and the
> ability to continue ifr any 1 of these items fail - the degree of failure
> you system cn cope with will be dependant on its config: for production I
> always have dual controllers, Server PSU's , cables, RAID 10 with 1 string
> of discs on 1 controller and the second on the second controller with
> redundant HDD PSU's + extra fans, fan / volatge monitoing, email / pager
> alerts + Hot standby disc drives + Tested spares. A lot can fail in that
> config and it will still go & it will tell you of failures & recover
> completely gining you the ability to hange out dead HDD's with no down
time
> at all.
>
> RAID 1, 10 or 5 does this as do other forms of RAID, *NOT* RAID 0.
> As soon as you get a disc falure using any RAID (raid 0 is not real raid),
> you should have Either a Hot spare configured that will already be being
> synchronised in the back ground maintaining a running system with only
some
> degradataion in performance.
>
> Or, on lower end RAID cards the system will continue to run - this is what
> real RAID is designed for. But you *must* IMO address the issue ASAP and
> Know in advance for each controller what to do when a disc fails. The
longer
> you run RAID in a degraded mode the longer you run the risk of total loss
as
> you are back to** No Raid and out naked in a storm.
>
> **often, but not always - you can have 2 discs fail in a four disc RAID
10,
> 3 in a 6 disc RAID 10 etc but the probability of such nice failures drops
> off.
>
> Paul wrote a few words about this the other day - have a look.
>
> The essential points:
>
> You *must* have a good backup reqime regardless.
> Discs in RAID configs fail. RAID controllers fail, PSU's fail and kill the
> lot - a 2nd disc can fail in close succession to a first so fix promptly,
be
> prepared for the worst case.
>
> Rehearse faiures and know what to do in advance before you go into
> production. You won't want to be learning this when an array fails.
>
> So read around in this NG, see Paul's commencts and some of my own
wafllings
> & never use "RAID 0" unless you can live with complete data loss of what
is
> stored on such volumes (EG scratch files in video processing, large read
> only databases loaded anew monthly etc).
>
> HTH
>
>
>
> "old man" > wrote in message
> ...
> > This board has both Nvidea raid & Sil raid
> > If, assuming mirror, raid1, do either of these have the capability of
> > detecting an individual hd failure and to continue running the sys,
> > without
> > having to break the raid and boot from the functioning hd?
> > ie The sys continues operating if an individual raid hd fails, without
any
> > user imput - other than acknowledging the failure.
> > The mobo manual is not clear on this subject.
> >
> >
>
>

Mercury
December 23rd 05, 02:14 AM
Dual everything on servers... including dual UPS & auto start generator.
Desktop dev systems, I use good PSU & RAID 1 & **UPS**.

Sounds like you got in a pickle with your SIL controller either through a
bios / controller firmware out of date dependancy, or ummmm you don't reboot
enough, or something nasty has happened. I can't explain the unknown :), but
I have heard of odd things like that before.

When I referred to Paul, I was also meaning the Asus NG too - seems you
guessed.

Sorry about the typos...

BOL

- Tim





"old man" > wrote in message
...
> Thanks,
> I asked since I previously had raid 1 on an Sil cntrl.,Asus mobo A7N8X
> delux, when 1 hd failed, the sys went down and I had to break the raid
> before I could reboot, trying each disk in turn to discover which had
> failed, which it did fine except 2months of data, and some installed apps
> were missing!
> It seems the mirror had'nt been in sync. The raid utility software had
> given
> no warning of any problems, and had allways shown the mirror in sync!
> I'm currently using an Adaptec raid cntrl card, in mirror, which I 'know'
> will continue running the sys if a hd dies. I'll be adding a hot swap
> spare
> shortly, which will automatically replace/rebuild if a hd fails. BTHW I
> still have both onsite/offsite data backups.- dont think I can quite
> warrant
> the amount of redundency you seem to have.
> TY
>
> "Mercury" > wrote in message
> ...
>> Yes if RAID 10 (or any other real RAID) - seemlessly. That is the
> objective
>> of RAID - requndancy or discs, controllers, cabling, PUS's etc. and the
>> ability to continue ifr any 1 of these items fail - the degree of failure
>> you system cn cope with will be dependant on its config: for production I
>> always have dual controllers, Server PSU's , cables, RAID 10 with 1
>> string
>> of discs on 1 controller and the second on the second controller with
>> redundant HDD PSU's + extra fans, fan / volatge monitoing, email / pager
>> alerts + Hot standby disc drives + Tested spares. A lot can fail in that
>> config and it will still go & it will tell you of failures & recover
>> completely gining you the ability to hange out dead HDD's with no down
> time
>> at all.
>>
>> RAID 1, 10 or 5 does this as do other forms of RAID, *NOT* RAID 0.
>> As soon as you get a disc falure using any RAID (raid 0 is not real
>> raid),
>> you should have Either a Hot spare configured that will already be being
>> synchronised in the back ground maintaining a running system with only
> some
>> degradataion in performance.
>>
>> Or, on lower end RAID cards the system will continue to run - this is
>> what
>> real RAID is designed for. But you *must* IMO address the issue ASAP and
>> Know in advance for each controller what to do when a disc fails. The
> longer
>> you run RAID in a degraded mode the longer you run the risk of total loss
> as
>> you are back to** No Raid and out naked in a storm.
>>
>> **often, but not always - you can have 2 discs fail in a four disc RAID
> 10,
>> 3 in a 6 disc RAID 10 etc but the probability of such nice failures drops
>> off.
>>
>> Paul wrote a few words about this the other day - have a look.
>>
>> The essential points:
>>
>> You *must* have a good backup reqime regardless.
>> Discs in RAID configs fail. RAID controllers fail, PSU's fail and kill
>> the
>> lot - a 2nd disc can fail in close succession to a first so fix promptly,
> be
>> prepared for the worst case.
>>
>> Rehearse faiures and know what to do in advance before you go into
>> production. You won't want to be learning this when an array fails.
>>
>> So read around in this NG, see Paul's commencts and some of my own
> wafllings
>> & never use "RAID 0" unless you can live with complete data loss of what
> is
>> stored on such volumes (EG scratch files in video processing, large read
>> only databases loaded anew monthly etc).
>>
>> HTH
>>
>>
>>
>> "old man" > wrote in message
>> ...
>> > This board has both Nvidea raid & Sil raid
>> > If, assuming mirror, raid1, do either of these have the capability of
>> > detecting an individual hd failure and to continue running the sys,
>> > without
>> > having to break the raid and boot from the functioning hd?
>> > ie The sys continues operating if an individual raid hd fails, without
> any
>> > user imput - other than acknowledging the failure.
>> > The mobo manual is not clear on this subject.
>> >
>> >
>>
>>
>
>