> Thanks for your reply. This list is amazing. I got responses from 3 people
> in
> about 20 minutes!
>
> Nobody really commented on my other uncertainty, though. How big of a
> partition do I need for the journal file? If the journal file is 10000
> blocks, how
> does that translate into MBs? And do I want a larger journal file? I'm
> mostly
> going to be storing HUGE (i.e., 50 MB to 2 GB) video files on the RAID.
> The RAID
> is 600 GBs. So, I won't have more than a couple of thousand files. How do
> you
> calculate how big a journal partition to make?
IIRC I was asking the same myself until I realized that the size of the
log is limited and quite small anyway. I think the size was limited to
128M or something with a filesystem size of some hundred megs. From my
experience the partition type doesn't matter at all.
Performance was much better with old XFS code on software RAID 5 when
using external log, every other configuration didn't give me much more
performance, YMMW.
Simon
>
> And does it matter what kind of "flag" I put on the partition with fdisk?
> Is
> 83 (Linux) fine? I don't think I need "fp" (RAID Autostart). I don't want
> to
> start my RAIDS manually anyway.
>
> Finally, should I really see a significant speed up in I/O with the
> journal
> on a separate drive? Is it only during writing that I'll see the speed up?
>
> Thanks.
>
>
> In a message dated 10/29/2003 4:20:33 PM Eastern Standard Time,
> nathans@xxxxxxx writes:
> On Wed, Oct 29, 2003 at 03:10:20PM -0500, AndyLiebman@xxxxxxx wrote:
>> ...
>> The command mkfs.xfs -f -l logdev=/dev/hdb3,size=10000b -b size=4096
>> /dev/md3 is accepted by the system. It looks as if I formatted the RAID
> with XFS and
>> put the journal on in a separate place.
>>
>> But when I try to mount the RAID as I did above, I get the following
>> error:
>> mount: wrong fs type, bad option, bad superblock on /dev/md3
>> or too many mounted file systems.
>>
>> I believe I also tried something like mount -t xfs logdev=dev/hdb3
>> /dev/md3.
>> I think I also tried mount -t xfs logdev=dev/hdb3,size=10000b
>> /dev/md3
>> Does anybody have any ideas about what the trouble is?
>
> You need to use the -o option to mount(8), and use the full
> device paths -- i.e.
>
> mount -t xfs -ologdev=/dev/hdb3 /dev/md3
>
> There is no size= mount option to XFS (mount(8) has the list),
> that information is obtained from the superblock.
>
> cheers.
>
> --
> Nathan
>
>
>
|