Thanks to the following:
sanghvi_at_proto.wilm.ge.com Arun Sanghvi
ckrieger_at_latrade.com Cliff Krieger
szgyula_at_tarkus.pha.jhu.edu Gyula Szokoly
dallaire_at_total.net Guy Dallaire
Looks like the answer is you can only lose one disk in
a RAID5 set and recover.
Responses:
----------
Whatever the number of devices that constitute your raisets, you are not
allowed to lose more than ONE disk. Make sure you have a spare drive in a
FAILEDSET so that the HSZ40 will grab it automatically if ever shit happens.
----------
Here is a quote from the StorageWorks Array Controllers guide:
If using RAID level 5, all data in the RAIDset will be lost if a second
drive fails in the same set before the first failed drive is repaired.
I have no reason to believe that it is different for software raid.
----------
Array will stop functioning if more then one drive failes.
----------
As I understand RAID-5, only *one* disk can fail without data loss.
If you mean complete disk faillure. How RAID-5 works:
You start with N drives. A data 'chunk' is divided into (N-1)
pieces, which are put on drive 1...(N-1). A special check sum is calculated,
which makes it possible to recover the data assuming *one* of these chunks are
not available. For the next data chunk, the same happens, but the drives
are 'rotated', i.e. the check sum is put on a different drive (I think drive
1, but may be (N-1) ). So in the end all drives are equal (i.e. no special
'check sum' drive exists).
Block errors are different: You can have multiple block errors on different
drives and still have no data loss *if* the errors are distributed luckily
(i.e. don't overlap).
Since SW and HW RAID-5 are doing the same, it does not matter which one
you have.
>
> Hello,
>
>
> If I have a 5 disks (rz28s), in a raid5 set,
> connected to a HSZ40, how many concurrent disk
> failures can occur and still be able to recover the data?
>
> Is the concurrent failure count different
> for hardware raid5 vs software raid5?
>
> Thanks,
>
> Les Mayeda
> Cedars-Sinai Medical Center
> mayeda_at_csmc.edu
>
Return-path: <sanghvi_at_proto.wilm.ge.com>
Received: from ns.ge.com (ns.ge.com) by CSMC.EDU (PMDF V5.0-7 #15047)
id <01IG2A50JOY894ENJ4_at_CSMC.EDU> for MAYEDA_at_CSMC.EDU; Mon,
03 Mar 1997 10:46:09 -0700 (PDT)
Received: from thomas.ge.com (thomas.ge.com [3.47.28.21])
by ns.ge.com (8.8.4/8.7.3) with ESMTP id NAA07033 for <MAYEDA_at_CSMC.EDU>; Mon,
03 Mar 1997 13:44:12 -0500 (EST)
Received: from proto.wilm.ge.com (proto.wilm.ge.com [3.43.40.202])
by thomas.ge.com (8.8.4/8.7.5) with SMTP id NAA14104 for <MAYEDA_at_CSMC.EDU>;
Mon, 03 Mar 1997 13:47:04 -0500 (EST)
Received: by proto.wilm.ge.com (5.65/DEC-Ultrix/4.3) id AA04396; Mon,
03 Mar 1997 13:45:38 -0500
Date: Mon, 03 Mar 1997 13:45:37 -0500 (EST)
From: sanghvi_at_proto.wilm.ge.com (arun sanghvi)
Subject: Re: raid5 set, how many disks can fail and still recover?
In-reply-to: <01IG290AP7EC8ZWR7J_at_CSMC.EDU>
To: MAYEDA_at_CSMC.EDU (LES MAYEDA TECH SUPPORT/ISS X6656)
Message-id: <9703031845.AA04396_at_proto.wilm.ge.com>
X-Mailer: ELM [version 2.4 PL20]
Content-type: text
Content-transfer-encoding: 7BIT
Received on Mon Mar 03 1997 - 23:01:19 NZDT