From: NeilBrown <neilb@suse.de>
To: Dylan Distasio <interzone@gmail.com>
Cc: linux-raid@vger.kernel.org
Subject: Re: RAID5 -> RAID6 conversion, please help
Date: Wed, 11 May 2011 13:29:35 +1000 [thread overview]
Message-ID: <20110511132935.0438143a@notabene.brown> (raw)
In-Reply-To: <BANLkTimYWfT_Vr7ZUnNUhOqW2JOyLHHNQQ@mail.gmail.com>
On Tue, 10 May 2011 21:04:29 -0400 Dylan Distasio <interzone@gmail.com> wrote:
> It's possible I did use different versions but I thought I had
> upgraded both of them right before the reshapes. Sorry if this is an
> elementary question, but does writing the 2nd parity block always to
> the last drive instead of rotating it increase the odds of a total
> loss of the array since the one specific drive always has the 2nd
> parity block?
No. The only possibly impact is a performance impact, and even that would be
hard to quantify.
The reason that parity is rotated is to avoid a 'hot disk'. Every update has
to write the parity block and if they are all on one disk then every write
will generate a write to that one disk.
Because of the way md implements RAID6, every write involves either a read or
a write to every device, so there is no real saving in rotating parity.
I think (but am open to being corrected) that rotating parity is only
important for RAID6 if the code implements 'subtraction' as well as
'addition' for the Q syndrome (which md doesn't) and if you have at least 5
drives, and you probably wouldn't notice until you get to 7 or more drives.
... so it might make sense to make mdadm default to converting to the -6
layout...
You can request it with "--layout=preserve".
>
> If so, do you think normalizing would be worth the risk of something
> going wrong with that operation? I'm just trying to get a feel for
> how much of a difference this makes.
Not worth the risk.
NeilBrown
>
>
> > mdadm first converts the RAID5 to RAID6 in an instant atomic operation which
> > results in the "-6" layout. It then starts a restriping process which
> > converts the layout.
> >
> > If you end up with a -6 layout then something when wrong starting the
> > restriping process.
> >
> > Maybe you used different version of mdadm? There have probably been bugs in
> > some versions..
> >
> > NeilBrown
> >
> >
> >
> --
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
next prev parent reply other threads:[~2011-05-11 3:29 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-05-10 23:15 RAID5 -> RAID6 conversion, please help Peter Kovari
2011-05-10 23:31 ` NeilBrown
2011-05-10 23:39 ` Steven Haigh
2011-05-11 0:21 ` NeilBrown
2011-05-11 0:38 ` Dylan Distasio
2011-05-11 0:47 ` NeilBrown
2011-05-11 1:04 ` Dylan Distasio
2011-05-11 3:29 ` NeilBrown [this message]
2011-05-11 0:08 ` Peter Kovari
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=20110511132935.0438143a@notabene.brown \
--to=neilb@suse.de \
--cc=interzone@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 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).