From: Goswin von Brederlow <goswin-v-b@web.de>
To: John Robinson <john.robinson@anonymous.org.uk>
Cc: Goswin von Brederlow <goswin-v-b@web.de>,
Linux RAID <linux-raid@vger.kernel.org>
Subject: Re: Poor write performance with write-intent bitmap?
Date: Wed, 22 Apr 2009 16:02:30 +0200 [thread overview]
Message-ID: <87skk0947d.fsf@frosties.localdomain> (raw)
In-Reply-To: <49EF108D.4070605@anonymous.org.uk> (John Robinson's message of "Wed, 22 Apr 2009 13:41:49 +0100")
John Robinson <john.robinson@anonymous.org.uk> writes:
> On 22/04/2009 10:16, Goswin von Brederlow wrote:
>> John Robinson <john.robinson@anonymous.org.uk> writes:
>>> Can't do that, my root filesystem is on the RAID-5, and part of the
>>> reason for wanting the bitmap is because the md can't be stopped while
>>> shutting down, so it was always wanting to resync at startup, which is
>>> rather tedious.
>>
>> Normal shutdown should put the raid in read-only mode as last step. At
>> least Debian does that. That way even a mounted raid will be clean
>> after reboot.
>
> Yes, I would have thought it should as well. But I've just looked at
> CentOS 5's /etc/rc.d/halt and as far as I can see it doesn't try to
> switch md devices to read-only. Of course the root filesystem has gone
> read-only but as we know that doesn't mean the device underneath it
> gets told that. In particular we know that ext3 normally opens its
> device read-write even when you're mounting the filesystem read-only
> (iirc it's so it can replay the journal).
>
> Another issue might be the LVM layer; does that need to be stopped or
> switched to read-only too?
Debian does
/sbin/vgchange -aln --ignorelockingfailure || return 2
before S60mdadm-raid, S60umountroot and S90reboot.
>> I would also suggest restructuring your system like this:
>>
>> sdX1 1GB raid1 / (+/boot)
>> sdX2 rest raid5 lvm with /usr, /var, /home, ...
>>
>> Both / and /usr can usualy be read-only preventing any filesystem
>> corruption and raid resyncs in that part of the raid.
>
> I did do this multiple partition/LV thing once upon a time, but I got
> fed up with having to resize things when one partition was full and
> others empty. The machine is primarily a fileserver and Xen host, so
> the dom0 only has 40GB of its own, and I couldn't be bothered
> splitting that up. Having said all this, your suggestion is a good
> one, it's just my preference to have it otherwise :-)
>
> Cheers,
>
> John.
I've been using a 1GB / for years and years now so that won't be a
problem. As for the rest one can also bind mount /usr, /var, /home to
/mnt/space/* respectively. I.e. have just 2 (/ and everything else)
partitions.
Esspecially for XEN hosts I find LVM verry usefull. Makes it easy to
create new logical volumes for new xen domains.
MfG
Goswin
next prev parent reply other threads:[~2009-04-22 14:02 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-04-20 17:12 Performance of a software raid 5 Johannes Segitz
2009-04-20 23:46 ` John Robinson
2009-04-21 0:10 ` Johannes Segitz
2009-04-21 0:52 ` John Robinson
2009-04-21 1:05 ` Johannes Segitz
2009-04-21 1:12 ` John Robinson
2009-04-21 1:19 ` NeilBrown
2009-04-21 2:04 ` Johannes Segitz
2009-04-21 5:46 ` Neil Brown
2009-04-21 12:40 ` Johannes Segitz
2009-04-24 13:49 ` Johannes Segitz
2009-04-26 17:03 ` Johannes Segitz
2009-04-21 18:56 ` Corey Hickey
2009-04-22 12:29 ` Bill Davidsen
2009-04-22 22:32 ` Corey Hickey
2009-04-22 9:07 ` Goswin von Brederlow
2009-04-21 0:44 ` Poor write performance with write-intent bitmap? John Robinson
2009-04-21 1:33 ` NeilBrown
2009-04-21 2:13 ` John Robinson
2009-04-21 5:50 ` Neil Brown
2009-04-21 12:05 ` John Robinson
2009-05-22 23:00 ` Redeeman
2009-04-22 9:16 ` Goswin von Brederlow
2009-04-22 12:41 ` John Robinson
2009-04-22 14:02 ` Goswin von Brederlow [this message]
2009-04-23 7:48 ` John Robinson
2009-04-22 14:21 ` Andre Noll
2009-04-23 8:04 ` John Robinson
2009-04-23 20:23 ` Goswin von Brederlow
2009-04-21 16:00 ` Bill Davidsen
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=87skk0947d.fsf@frosties.localdomain \
--to=goswin-v-b@web.de \
--cc=john.robinson@anonymous.org.uk \
--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