Dear Christoph Hellwig <hch@xxxxxxxxxxxxx>,
Thank you for your infomation.
I found a reason for this problem. This problem is happend by lack of memory.
"xfs_check" and "xfs_repaire" need too much memory. In our case (160GByte full
disk),
"xfs_check" and "xfs_repaire" request 100-400MByte memory.
So, do you have any idea to reduce memory for "xfs_check" and "xfs_repaire" ?.
Now, i am using xfs-cmd 2.3.5 version. More recent version is same condition.
Also, is it possible to caluculate max memory ?
Another quesion,
I want to improve xfs performance (read,warite,delete). But, at this time,i can
not
kernel version(2.4.17),but it is possible to change code. Do you have any ideas
to improve performance ?
>As a simple addition: you could try to get newer XFS userspace from
>oss.sgi.com - but 2.4.17 was before XFS 1.1 and thus has a slightly different
>user <-> kernelspace api so it will probably not work. I still think it's
>probably your kernel corrupting the filesystem and not a userspace issues,
>though. Especially as powerpc is big endian and unsigned char - XFS was
>written for such an enviroment (IRIX/mips) first but for the linux port
>that combination has been broken a few times.
>
Shigenori Motooka
MegaChips Corporation
SYSTEM BUSINESS UNIT
4-1-6,Miyahara,Yodogawa-ku,Osaka,532-0003 JAPAN
Phone:+81-6-6399-2865
Fax :+81-6-6399-6204
E-mail:motooka.shigenori@xxxxxxxxxxxxxxx
rom owner-linux-xfs@xxxxxxxxxxx Wed Mar 3 17:36:26 2004
Received: with ECARTIS (v1.0.0; list linux-xfs); Wed, 03 Mar 2004 17:36:30
-0800 (PST)
Received: from gawab.com (www.gawab.com [204.97.230.36])
by oss.sgi.com (8.12.10/8.12.9) with SMTP id i241aPKO031899
for <linux-xfs@xxxxxxxxxxx>; Wed, 3 Mar 2004 17:36:26 -0800
Received: (qmail 49529 invoked by uid 1004); 4 Mar 2004 01:41:15 -0000
Received: from unknown (HELO ramy) (ramy@xxxxxxxxxxxxxx)
by gawab.com with SMTP; 4 Mar 2004 01:41:15 -0000
From: "Ramy M. Hassan" <ramy@xxxxxxxxx>
To: <linux-xfs@xxxxxxxxxxx>
Subject:
Date: Thu, 4 Mar 2004 03:35:57 +0200
Message-ID: <000301c40189$0dde0d20$ba10a8c0@ramy>
MIME-Version: 1.0
X-Priority: 3 (Normal)
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook, Build 10.0.2627
X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1165
Importance: Normal
Content-Disposition: inline
Content-Type: text/plain
Content-Transfer-Encoding: 7bit
Content-length: 939
X-archive-position: 2329
X-ecartis-version: Ecartis v1.0.0
Sender: linux-xfs-bounce@xxxxxxxxxxx
Errors-to: linux-xfs-bounce@xxxxxxxxxxx
X-original-sender: ramy@xxxxxxxxx
Precedence: bulk
X-list: linux-xfs
Hi,
I have about 1,000,000 directories currently stored on an ext3
partition. I want to move data to xfs. Due to ext3 limitations I apply a
directory hashing algorithm to place those 1M directories into other
subdirectories so that the maximum number of inodes in a directory does
not exceed 32K ( ext3 limit ).
I know that xfs does not have this limitation , and I read that it uses
balanced trees to organize directories, but is it better in xfs to have
all those 1M directories in a single flat level ( I never need to ls or
readdir them all ) , or stick to balance them in three or four levels ?
I also wonder if xfs performance is good for too many small files or
not. I know it is the best for large files, but this is not my case ( my
avarage file size is about 140k ) . If xfs is not the best choice for me
, can you please suggest me a more suitable alternative.
Best regards
Ramy M. Hassan
[[HTML alternate version deleted]]
|