zpetar wrote:
> > If the OS supports the RAID controller, you should just see a single
> > big disk. That's because the card is supposed to handle all of the
> > RAID functionality by itself and just present a single array to the
> > OS. But, alas, I don't think Seawolf supports the HPT370. (Roswell2
> > seemed to detect the HPT370 on my Abit KG7-RAID boards but I didn't
> > exercise it.)
>
> kewl ... but how to ''force'' seawolf to see one big disk ... ;) ...
> new kernel ?
>
> _______________________________________________
> Seawolf-list mailing list
> [EMAIL PROTECTED]
> https://listman.redhat.com/mailman/listinfo/seawolf-list
Hi
I have a few raid systems.
The device is usaly something like /dev/ida/c0d0 .
To create the file system use :
fdisk /dev/ida/c0d0
Then create a primary partion starting at "1" {the default} and ending on
the last cylinder {the next default}.
Change the partion type to the filesystem you want to use.
Write the configuration.
Format the new partition, likely :
mke2fs /dev/ida/c0d0p1
NOTE:
It is unwise to create single huge filesystems. You need to make a
compromize between the maximum number of inodes and the size of inodes.
Another problem is "fsck" when the drive needs to be checked it will take
ages to check a huge drive. You can usualy create 14 [1-3 primary ,5-15
logical] partitions. Using a journaling file system optomized to your
requirements of running performance and recovery speed can make a big
difference as well. Ext3 is able to run as fast as Ext2 or with less
recovery time at a cost of some speed. Having directories with over 1000
files makes file access slower as well. We have many users so for instance
this is similar to how we provide home directories with private web space.
/home/location/type/u/s/username
Where location represents the municipality, type indicates staff/user, and
username represents the users name.
By seperating the usre home directories into double hashed directories,
the number of users per directory is less.
Seperating each location, we use different partions to allow more
available inodes.
The type sepertaion is purly administative.
Our mail spool is similarly hashed, makeing mail access faster. Rather
than 5000 mailboxes in one directory when access the file usualy less than
50 entries now have to be sorted. The heirchy is easy to manage, because
their are 26*2 = 52 directories in the first hash and 26 * 2 + 10 + 4 = 66
directories in the second hash, make directory listing managable. The load
required for users to check for new messages has been greatly reduced as
well as the load required to find the file to deliver new mail.
If you want to know how to do this contact me. [guy(at)incentre(dot)net]
Guy
PS I have modified procmail and Eudora's "Qualcomm POP server" qpopper to
make this work.
_______________________________________________
Seawolf-list mailing list
[EMAIL PROTECTED]
https://listman.redhat.com/mailman/listinfo/seawolf-list
