Linux RAID subsystem development
 help / color / mirror / Atom feed
From: Werner Fischer <devlists@wefi.net>
To: "Tyler J. Wagner" <tyler@tolaris.com>
Cc: linux-raid@vger.kernel.org
Subject: Re: alignment of metadata 1.2 and mkfs.ext4
Date: Wed, 06 Jul 2011 10:35:13 +0200	[thread overview]
Message-ID: <1309941313.2211.16.camel@werner-t410> (raw)
In-Reply-To: <1309936276.11661.8.camel@baal>

Hi all,

like Tyler I'm also asking me those questions.
Reason is that newer Solid-State Drives (like the Intel 320 Series) have
Flash-Chips with page size of 8KiB (the former SSDs have mostly 4KiB
pages). So in the best case, all data written to such an SSD should be
at least 8KiB aligned - which I think is not the case for MD-RAID with
metadata 1.2 superblocks - because of the "4K from the beginning of the
device":
https://raid.wiki.kernel.org/index.php/RAID_superblock_formats#Sub-versions_of_the_version-1_superblock
When I understand the wiki page correctly, the metadata 1.2 superblock
has a size of 1024 Bytes (256 Bytes + 2 Bytes/device for a maximum of
384 devices), right?

When I simply put an Ext4 file system with a block size of 4KiB on the
RAID, there is no problem as there will always be two Ext4 blocks
residing in a single 8 KiB page of the SSD.

My questions:
1) But when I use LVM2 with a PE size of 4 MiB on top of a RAID, I will 
   not be perfectly aligned any more, right?
2) Are there any plans that e.g. the first 1MiB will be reserved for 
   metadata (like fdisk does when creating new partitions), so that 
   there won't be any alignment issues in the future?
     (btw: haven the metadata at the end of the device like in the 
     metadata format 0.90 or 1.0 would also not interfere the alignment 
     - is there a special reason why the metadata has been put to the 
     beginning of the device?)

best regards,
Werner

On Wed, 2011-07-06 at 08:11 +0100, Tyler J. Wagner wrote:
> Hi all (well, Neil, really),
> 
> The the following, Neil states that that is no issue with alignment of
> metadata 1.2 and mkfs.ext4 on 4K drives:
> 
> http://www.issociate.de/board/post/507176/filesystem_on_mdadm_raid_from_unpartioned_block_devices.html
> 
> Sorry to ask again, but I just want to clarify. Metadata 1.2 puts the
> data 4K from the start of the device/partition. Does mkfs put the start
> of the filesystem at 64K from the start? IE, will the following align,
> without any need for options?
> 
> (create partitions starting 2048 512-byte sectors from start = 1MB)
> mdadm --create /dev/md0 -v --raid-devices=4 --level=raid10
> --metadata=1.2 /dev/sda1 /dev/sdb1 /dev/sdc1 /dev/sdd1
> mkfs.ext4 /dev/md0
> 
> And, slightly off-topic, it's not necessary to define stride and stripe
> any longer?
> mkfs.ext4 -E stride=16,stripe-width=32 /dev/md0
> 
> Regards,
> Tyler
> 
> 

-- 
: Werner Fischer
: Technology Specialist
: Thomas-Krenn.AG | The server-experts
: http://www.thomas-krenn.com | http://www.thomas-krenn.com/wiki


  reply	other threads:[~2011-07-06  8:35 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2011-07-06  7:11 alignment of metadata 1.2 and mkfs.ext4 Tyler J. Wagner
2011-07-06  8:35 ` Werner Fischer [this message]
2011-07-06  9:20 ` Roman Mamedov
2011-07-06  9:52   ` Tyler J. Wagner
2011-07-06 10:02     ` Roman Mamedov
2011-07-06 13:40       ` Tyler J. Wagner
2011-07-06 10:03     ` Robin Hill
  -- strict thread matches above, loose matches on Subject: below --
2011-07-06  7:10 Tyler J. Wagner

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=1309941313.2211.16.camel@werner-t410 \
    --to=devlists@wefi.net \
    --cc=linux-raid@vger.kernel.org \
    --cc=tyler@tolaris.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