xfs
[Top] [All Lists]

Re: XFS and RAID

To: Raymond <support@xxxxxxxxxxxxxxxxxxxx>, linux-xfs@xxxxxxxxxxx
Subject: Re: XFS and RAID
From: Seth Mos <knuffie@xxxxxxxxx>
Date: Fri, 24 May 2002 10:23:22 +0200
In-reply-to: <auto-000017217178@xxxxxxxxxxxxxxxxxxxxxxxxx>
Sender: owner-linux-xfs@xxxxxxxxxxx
At 23:50 23-5-2002 -0700, Raymond wrote:
Have installed Suse 8.0 on an Intel L440GX system. Controller is an AMI
MegaRAID 1500.series 467.

Seagate 18g drives are configured as follows:
        Channel A - 2 disks at RAID1
        Channel B - 5 disks at RAID5
        One disk as hot spare.

Boot partition is an ex2; var, home and root are XFS. Home is on B channel,
everything else is on A.

If you are using ext2 for the boot partition ext3 on that partition is a logical step.

Have not installed experimental kernel RPM with ACL patch.

If you are using anything later then 2.4.18 you should also install the newer userspace >2.0

Any controller tweaks or caveats I should be aware of?

The controller will work fine although it's raid 5 write/read speed is not high. (not related to XFS) If you have the battery backup option you can switch on write back caching which will help the write performance greatly (all raid levels).

Software raid5 will give you a much better performance. And if you make it with an external log onto the raid1(hardware) device it will perform a few hundred percent better then the hardware raid5 of the controller.

If you are planning to handle large files the default mkfs options will suit it just fine. If you plan to have lot's of small files and a modify a lot (mail/news spool). It would the be better to make the fs with a larger log "mkfs.xfs -l size=32768b /dev/foo" and mount it with more buffers "mount -o logbufs=8"

This also means recovery takes longer and you could potentially lose more data if the server goes down but it is a trade off that you can make.

Cheers

--
Seth
It might just be your lucky day, if you only knew.


<Prev in Thread] Current Thread [Next in Thread>