From: Bill Davidsen <davidsen@tmr.com>
To: Carlos Mennens <carloswill@gmail.com>
Cc: Mdadm <linux-raid@vger.kernel.org>
Subject: Re: RAID1 On 3 Drives
Date: Sat, 20 Mar 2010 12:08:40 -0400 [thread overview]
Message-ID: <4BA4F308.8010303@tmr.com> (raw)
In-Reply-To: <d80f793f1003150530y1b229f43s1f140f91ab74f3e8@mail.gmail.com>
Carlos Mennens wrote:
> On Sat, Mar 13, 2010 at 6:50 PM, Bill Davidsen <davidsen@tmr.com> wrote:
>
>> Having a spare on raid-1 is fairly pointless, it hurts performance and buys
>> you nothing. Having one more copy of the data built and ready serves you
>> better.
>>
>
> Can you explain this as I find this interesting. How does having a
> /boot partition on 3 drives with 1 spare hurt performance? Are you
> saying that I would get better drive performance if I had all 4 disk
> partitions active members of my RAID1 /boot? I just don't understand
> how the 4th disk doing nothing but acting as a spare would hinder
> performance.
>
>
In RAID-1 any drive which is a member of the array is already a spare,
they are all the same. So if a drive failed you would just stop using
it. Having a spare makes no sense, since the spare would be rebuilt
using the same data it would have contained as a member of the array.
Note that while the "spare" is being rebuilt the array will be quite
busy and response will be poor, not an issue if /boot is the only thing
on the drives, but more of an issue if the array just uses partitions
and the thrashing of a spare rebuild slows other arrays on the same drives.
Finally, RAID-1 supposedly will read from an unbusy drive if there are
multiple reads against the array, allowing overlapping of seek and
transfer. That's only an issue for /boot if you are trying for a ten
second boot time, but it applies to all RAID-1 arrays.
> Secondly, if the above statement also applies to my / partition? Would
> you suggest using all 4 drives as active partitions in a RAID5 array
> too? If I have a 3 disk RAID5 and one hot spare, do you think I would
> get less performance value for my configuration?
>
No, because the drives are not all the same, so you can't keep a copy of
just the drive ready to fail next. For RAID-[456&10] you want at least
one spare for reliability and fast return to full performance.
Ask again if I didn't give enough detail.
--
Bill Davidsen <davidsen@tmr.com>
"We can't solve today's problems by using the same thinking we
used in creating them." - Einstein
next prev parent reply other threads:[~2010-03-20 16:08 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-03-13 22:30 RAID1 On 3 Drives Joachim Otahal
2010-03-13 23:18 ` Carlos Mennens
2010-03-13 23:50 ` Bill Davidsen
2010-03-15 12:30 ` Carlos Mennens
2010-03-15 12:48 ` Majed B.
2010-03-15 14:47 ` Carlos Mennens
2010-03-15 16:05 ` Majed B.
2010-03-16 1:46 ` Keld Simonsen
2010-03-20 16:08 ` Bill Davidsen [this message]
-- strict thread matches above, loose matches on Subject: below --
2010-03-13 21:06 Carlos Mennens
2010-03-13 21:21 ` Majed B.
2010-03-13 21:24 ` Carlos Mennens
2010-03-14 21:09 ` Luca Berra
2010-03-13 21:45 ` Neil Brown
2010-03-13 21:26 ` Neil Brown
2010-03-13 21:31 ` Majed B.
2010-03-13 21:50 ` Neil Brown
2010-03-14 3:30 ` Majed B.
2010-03-14 1:22 ` thib
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4BA4F308.8010303@tmr.com \
--to=davidsen@tmr.com \
--cc=carloswill@gmail.com \
--cc=linux-raid@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.