Re: Choosing data file size for a multi TB database? (just the bit about 90% utilization)

On 8/30/05, Mark W. Farnham <mwf@xxxxxxxx> wrote:
> 
> <snip>
> 
> your OS vendor. HP I know is not happy if you exceed 90% utilization on
> a mount point.
> <snip>
> 
> I'm not sure what you mean by "not happy" but the default default inode 
> plus
> bad spot allocation on many UNIX descendents is still 10%.
> 
> Most badspot mapping is done in the hardware now, and 10% has become a
> ridiculous amount of storage to dedicate to information about the
> name, dlm, size, etc., starting point, and various other metadata about 
> the
> list of files on a mount point.
> 
> This is especially true if a mount point is dedicated to one or a modest
> number of files, as should be the case on mount points holding
> Oracle data files.
> 
> So, IF the unhappiness you're referring to is this 10% reservation, then 
> by
> all means build your mountpoints for Oracle data files with much less
> allocated to inode space. The exact syntax may vary by platform, so I'm 
> not
> even going to start. If you converse with your Hardware/OS vendor, be
> certain they understand the question or you'll just get back "Oh, you
> shouldn't monkey with that" or some other such nonsense. 10% is a whole 
> lot
> to toss down the drain for inode storage.
> 
> Then again, you might be writing about something entirely different.
> 
> Regards,
> 
> mwf
> 
> 
> 
Hijacking this thread towards the South Pole ...

Using extfs3 on Linux, its not unusual for 5% of a volume to be allocated in 
reserve.
I dug this one up from the last time I setup a leenucks box ...
If you don't specify largefile ... I believe it creates an inode for every 2 
blocks.

usual disclaimer of most people don't use cooked filesystems ... use 
ASM/raw/3rd party volume manager ... this was fine for a dev box.


[root@somebox etc]# mkfs.ext3 -b 4096 -L /u08 -T largefile4 -v /dev/sdd2
mke2fs 1.32 (09-Nov-2002)
Filesystem label=/u08
OS type: Linux
Block size=4096 (log=2)
Fragment size=4096 (log=2)
4096 inodes, 4168867 blocks
208443 blocks (5.00%) reserved for the super user
First data block=0
128 block groups
32768 blocks per group, 32768 fragments per group
32 inodes per group
Superblock backups stored on blocks:
32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208,
4096000

Writing inode tables: done
Creating journal (8192 blocks): done
Writing superblocks and filesystem accounting information: done

This filesystem will be automatically checked every 26 mounts or
180 days, whichever comes first. Use tune2fs -c or -i to override.
[root@nycdora0009 etc]# mount -v -t ext3 /dev/sdd2 /u08
/dev/sdd2 on /u08 type ext3 (rw)


mke2fs supports the switch "-m"
-m reserved-blocks-percentage

choose wisely.

hth.

Paul

Other related posts: