Sorry for the delay in replying :-(
Corin Hartland-Swann ([EMAIL PROTECTED]) wrote on 22 February 2001 16:36:
>On Thu, 22 Feb 2001, Carlos Carvalho wrote:
>> What happens if I have a raid5 array and, instead of the usual
>> situation of one disk failing, I have the opposite, all disks stop
>> except one?
>
>Unless you are using a two-disk RAID-5 array (which is really a RAID-1
>Mirror), if all disks except one fail, the array will fail.
>
>> This may seem strange, but may happen when you have one power supply
>> feeding all disks but one, and another power supply feeding the last
>> disk. If the power supply serving most disks fails, there will be one
>> single disk running. At reboot what happens????
>
>The array will not start, because one disk is not enough to reconstruct
>the contents of the array.
Yes, my doubt is about how the kernel will behave. It'd be enough for
me to remove the UNfailed disk and force its re-creation from the
other ones. Problem is how to do this. Maybe I can just pull its plug
and let the array start in degraded mode. The other ones will be
coherent, so it'll start. Then shutdown, put the other one back and
start again.
>If you are trying to remove the power supply as
>a point of failure, you probably want to use RAID-10 instead of RAID-5
It wastes too much disk. I can't afford a 50% loss.
In fact my goal is to have each element in the raid5 be a raid0, each
one using a separate scsi board and cables. I didn't mention it to try
to focus on the startup issue.
>RAID-10 arrays also give much better read/write/seek performance than the
>RAID-5 array.
Raid1 is better in writing, not read/seek. I expect a raid5-0 will
be the same compared to a raid1-0 as a raid5 is to raid1. I must trade
write efficiency for di$k $pace...
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to [EMAIL PROTECTED]
