Software RAID5 - parity regeneration - disks order

[rs]

Hi,

I have a fairly big software RAID5 based on 16 disks. Probably because
of some sort of the power failure (it happened during the night, first
system reported network card failure and then collapsed itself) the
motherboard died, and one of the RAID's disk as well. Since this is a
quite old machine I couldn't get the same motherboard, I bought
similar one but different model and brand. I connected all the
components and rebuilt the operating system, without any problems.
After connecting all the RAID's disks system reported four of them as
brand new with unallocated space. I Spent some time to rebuild the
headers of all these 4 disks, and started to regenerate the parity of
the RAID. Unfortunately after the regeneration completed and RAID
reported as healthy, I cannot access many directories on it. I can go
inside them in DOS shell, but they are empty. System reports the space
occupied by files and empty space correctly as it was before the
crash. After I restart the computer system starts regenerating the
RAID again and again.

What could happen was I connected 4 disks in the wrong order. Two
disks from IDE1 to IDE2 and vice versa. They must have been wrongly
noted on one motherboard's diagram or the person who assembled the
computer before marked cables the wrong way. At this point it doesn't
really matter, but if that was the case, doesn't the regenerating the
RAID screwed up the previous parity information and the game is
basically over, or should I fight using some third party software or
tools with current disks setup? Or maybe I should change the order of
these 4 hard disks, update their header's information again and start
regenerating the RAID all over? Or maybe there are other options I
cannot think of?

I will very greatly appreciate any valuable hint or information. <rs>

 

[Dan Adams]

You really need to use a hardware RAID solution.

 

[Toshi1873]

You really need to use a hardware RAID solution.

Not necessarily. There are pros/cons to both hardware RAID and software
RAID.

A software RAID can be moved from system to system and is about as non-
proprietary as you can get. If you use mdadm, you don't even need a
configuration file for the RAID array (although it's good for
documentation reasons) because mdadm will reassemble drives into their
arrays automatically. You don't even have to worry about drive ordering
or what physical interface you've plugged the drive into.

The downside is that you spend CPU cycles doing the work of a hardware
RAID card and you'll (probably?) have more traffic to/from the devices
because the CPU is calculating parity. Hardware RAID does better in
this regard (as long as it's "true" hardware RAID and not software-
assisted hardware RAID).

(There's a lot more to the hardware vs software RAID argument then this
incomplete list.)

I've used both. Typically we end up using hardware RAID for the Windows
servers and software RAID for the Linux servers. HW RAID is great for
whole-disc raids while SW RAID lets us do some interesting things.

 

[rs]

You really need to use a hardware RAID solution.

thanks for the answer but it won't help me much, will it? on another
hand do you know how much would cost hardware solution for 3.6TB RAID?
<rs>

 

[Folkert Rienstra]

Not necessarily. There are pros/cons to both hardware RAID and software
RAID.

A software RAID can be moved from system to system and is about as non-
proprietary as you can get. If you use mdadm, you don't even need a
configuration file for the RAID array (although it's good for
documentation reasons) because mdadm will reassemble drives into their
arrays automatically. You don't even have to worry about drive ordering
or what physical interface you've plugged the drive into.

The downside is that you spend CPU cycles doing the work of a hardware
RAID card and you'll (probably?) have more traffic to/from the devices
because the CPU is calculating parity.

Hardware RAID does better in this regard

(as long as it's "true" hardware RAID and not software-assisted hardware RAID).

Interresting. Do you have examples for 'true' hardware RAID?
Or did you mean Hardware (assisted) Raid versus Firmware (assisted) Raid

(There's a lot more to the hardware vs software RAID argument then this
incomplete list.)

I've used both.

Typically we end up using hardware RAID for the Windows servers

Which ones, the software assisted ones or the 'true' hardware ones?

 

[Curious George]

thanks for the answer but it won't help me much, will it? on another
hand do you know how much would cost hardware solution for 3.6TB RAID?
<rs>

Could be as little as the price of your current system plus the price
of a good card.

 

[dannysdailys]

rswrote
Hi

I have a fairly big software RAID5 based on 16 disks. Probabl becaus
of some sort of the power failure (it happened during the night firs
system reported network card failure and then collapsed itself) th
motherboard died, and one of the RAID's disk as well. Since this i

quite old machine I couldn't get the same motherboard, I bough
similar one but different model and brand. I connected all th
components and rebuilt the operating system, without any problems
After connecting all the RAID's disks system reported four of the a
brand new with unallocated space. I Spent some time to rebuild th
headers of all these 4 disks, and started to regenerate the parit o
the RAID. Unfortunately after the regeneration completed and RAI
reported as healthy, I cannot access many directories on it. I ca g
inside them in DOS shell, but they are empty. System reports th spac
occupied by files and empty space correctly as it was before th
crash. After I restart the computer system starts regenerating th
RAID again and again.

What could happen was I connected 4 disks in the wrong order. Tw
disks from IDE1 to IDE2 and vice versa. They must have been wrongl
noted on one motherboard's diagram or the person who assembled th
computer before marked cables the wrong way. At this point i doesn'
really matter, but if that was the case, doesn't the regeneratin th
RAID scr**ed up the previous parity information and the game i
basically over, or should I fight using some third party softwar o
tools with current disks setup? Or maybe I should change the orde o
these 4 hard disks, update their header's information again an star
regenerating the RAID all over? Or maybe there are other options
cannot think of

I will very greatly appreciate any valuable hint or information <rs

I'm just curious; did you say you had a power failure on a 16 dis
RAID array? You have no power redundancy? Did you say you let
surge take out the motherboard of a 16 disk array?

I couldn't have heard that right. Right

 

[Curious George]

I'm just curious; did you say you had a power failure on a 16 disk
RAID array? You have no power redundancy? Did you say you let a
surge take out the motherboard of a 16 disk array??

I couldn't have heard that right. Right?

& he has no backup, but the data is important enough to warrant the
time expenditure & stress of trying to fix this. IIRC drive order
doesn't matter for M$/Veritas Raid5 so the problem is a bit deeper and
won't come from a quick fix. Sounds to me like either he can't really
afford a 16 disk array or it can't afford his planning. Something to
consider as money & time are spent trying to get this buckets of bolts
off the ground again...