From: Brad Campbell <brad@wasp.net.au>
To: Ryan Wagoner <rswagoner@gmail.com>
Cc: RAID Linux <linux-raid@vger.kernel.org>
Subject: Re: Basic RAID5/6 reshape question
Date: Fri, 12 Dec 2008 23:17:49 +0400 [thread overview]
Message-ID: <4942B8DD.8090809@wasp.net.au> (raw)
In-Reply-To: <7d86ddb90812111706t14e7fbdar6bf31a22d510a6b6@mail.gmail.com>
Ryan Wagoner wrote:
> It takes 3 hours to resync my 3 x 1 TB drive RAID 5 array at a rate of
> 85MB/s. It shouldn't take 20 hours to write 3 x 1 TB drives. How fast
> is it progressing? You might consider a PCI-X or PCI Express SATA or
> SAS controller card. I have my drives connected to an Supermicro LSI
> 1068 PCI Express SAS card, which performs great.
>
I'm using 2 x Marvell 7042 based PCI-E x4 controllers and a SIL 3xxx (I don't recall but its PCI-E
x1 and 2 ports). Bandwidth is not an issue.
A resync appears to be a completely different kettle of fish to a reshape. With re-shape it appears
to write status data to the raid superblocks periodically. This appears to thrash the heads from the
current position to the end of the disk and back. At the start of the reshape I was running about
15-20MB/s to each drive. As the reshape progressed to the end of the disks I was looking at closer
to 40MB/s. So obviously the final figure was less than the initial 20 hours, but a hell of a lot
slower than a straight resync.
This blind stab at reshape behaviour is absolutely pure conjecture based on a gross guess at what
was happening, however it does seem to bear out scrutiny.
> The other option would be to create 2 RAID 5 arrays of 5 drives each.
> Create the first array and use LVM with the filesystem of your choice.
> Add the second array later and expand the LVM volume.
Mmm thanks, but no.. to use a car analogy, "I'd like something I can fix with a screwdriver and a
hammer". LVM is overcomplicating things way past what I require. In addition, I really like RAID-6.
Over the last 4 years with my 15 drive RAID-6 I have had two double drive failures. Both times one
drive died in a not so nice fashion, and one developed a grown defect during the rebuild. Raid-5
just does not cut the mustard with me anymore.
Regards,
Brad
--
Dolphins are so intelligent that within a few weeks they can
train Americans to stand at the edge of the pool and throw them
fish.
next prev parent reply other threads:[~2008-12-12 19:17 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-12-11 9:55 Basic RAID5/6 reshape question Brad Campbell
2008-12-12 1:06 ` Ryan Wagoner
2008-12-12 19:17 ` Brad Campbell [this message]
2008-12-12 20:27 ` Justin Piszcz
2008-12-12 3:52 ` Roger Heflin
2008-12-15 22:24 ` Neil Brown
2008-12-15 22:20 ` Neil Brown
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=4942B8DD.8090809@wasp.net.au \
--to=brad@wasp.net.au \
--cc=linux-raid@vger.kernel.org \
--cc=rswagoner@gmail.com \
/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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).