From: Trevor Osatchuk (fybar27@gmail.com)
Date: Fri Mar 09 2007 - 13:03:38 EST
Thanks to Steve VanDevender and Dirk Kleinhesselink. As I had thought
the reported disk size by HP is probably using 1000 instead of 1024
for k. The more important issue of reported partition size and usable
space was simply explained that newfs reserves 10% of disk space for
root writes so that no matter how much space is used by non-priveleged
users root will always have space available. The reasons for this are
described by Steve:
Partly this is because the file allocation algorithms for UFS-like
filesystems can better avoid fragmentation if they can depend on
having some free space available, and partly so that root will
continue to be able to write files even if the space available to
non-privileged users is full.
The man page for tunefs says that setting this value to zero via newfs
and the -m option can have the impact of lowering the throughput by a
factor of 3. Steve also suggested that for large disks a smaller
percentage of 5 or even 2 might be sufficient. Since this would
primarily be used for defrag I can see the logic that something like a
1GB would easily be enough to defrag the largest drive. I will play
with this value and run some tests to see if lowering it too much will
cause any lag.
Thanks again boys!
Original post follows:
I have a mirrorset on two HSG80's. I have it configured as folllows:
HSG BOTTOM>sho dev
Name Type Port Targ Lun Used by
------------------------------------------------------------------------------
DISK10000 disk 1 0 0 MIRR1
DISK10100 disk 1 1 0
DISK10200 disk 1 2 0
DISK20000 disk 2 0 0 MIRR1
DISK20100 disk 2 1 0
DISK20200 disk 2 2 0
HSG BOTTOM>sho mirr
Name Storageset Uses Used by
------------------------------------------------------------------------------
MIRR1 mirrorset DISK10000 D0
DISK20000
The disks in question are Model: BF0728A4BA disks from HP. They have
72.8GB of space on them. I did the following on my Tru64 4.0F server
to mount this raid:
$> disklabel -wr /dev/rrz16a
$> disklabel /dev/rrz16a
# /dev/rrz16a:
type: SCSI
disk: HSG80
label:
flags:
bytes/sector: 512
sectors/track: 255
tracks/cylinder: 4
sectors/cylinder: 1020
cylinders: 8369
sectors/unit: 142229253
rpm: 3600
interleave: 1
trackskew: 7
cylinderskew: 26
headswitch: 0 # milliseconds
track-to-track seek: 0 # milliseconds
drivedata: 0
8 partitions:
# size offset fstype [fsize bsize cpg] #
NOTE: values n ot exact
a: 131072 0 unused 0 0 #
(Cyl. 0 - 1 28*)
b: 262144 131072 unused 0 0 #
(Cyl. 128*- 3 85*)
c: 142229253 0 unused 0 0 #
(Cyl. 0 - 1 39440*)
d: 0 0 unused 0 0 #
(Cyl. 0 - - 1)
e: 0 0 unused 0 0 #
(Cyl. 0 - - 1)
f: 0 0 unused 0 0 #
(Cyl. 0 - - 1)
g: 70918018 393216 unused 0 0 #
(Cyl. 385*- 6 9912*)
h: 70918019 71311234 unused 0 0 #
(Cyl. 69912*- 139440*)
I then did a newfs on partition 'c' to use the whole disk and mounted it:
$>df -k
Filesystem Size Used Avail Use% Mounted on
/dev/rz16c 68883207 1 61994885 1% /mnt
That shows roughly 66G gigs for size and only 59G available. A far
cry from 72.8GB as labeled. So, is it my disklabel that is screwed or
maybe a config on the HSG? Or, is it that HP reports 72.8GB using
1000 1024K blocks as opposed to 1024 1024K blocks? Also, why is the
partition size shown as 66G but only had 59G available.
Thanks,
fybar
This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:50:34 EDT