[megaraid] HP LH3 netraid performance issue

[MAtteo HCE Valsasna]

hi all,

I apologize for not being stricly in topic, but I hope I can catch
people with HW similar to mine here.

any references to more appropiate forums/resources are welcome, also on
private mail.

I have an HP LH3 server with its integrated netraid card, and 3 HP 18GB
disks in a hw raid 5 volume. I am currently using the sym53c8xx driver
and the megaraid (both statically compiled) from a standard kernel on a
RH6.2 machine.

# cat /proc/scsi/sym53c8xx/0
General information:
Chip sym53c895, device id 0xc, revision id 0x1
IO port address 0xe800, IRQ number 10
Using memory mapped IO at virtual address 0x90800800
Synchronous period factor 10, max commands per lun 32

My issues are about the megaraid driver:

I just discovered about the -DHP switch in the megaraid.c driver (shame
on me for not reading the sources!). That should get rid of the strange
characters the driver showed at boot:

megaraid: [D :B ] detected 1 logical drives

I hope it will also let the driver show something when I type
# cat /proc/scsi/megaraid/1

(now it prints nothing)

I am also concerned about the performance of the megaraid volume (the
following was evaluated on an almost idle machine):
# hdparm -t /dev/sda

/dev/sda:
Timing buffered disk reads: 64 MB in 3.27 seconds = 19.57 MB/sec

with a 3-disk raid 5 set I expected something more.

will I get better performance by simply recompiling the kernel with the
-DHP switch?

at boot, I get the following messages

messages.1.gz:Oct 4 13:39:25 mull kernel: sym53c8xx: at PCI bus 1,
device 7, function 0
messages.1.gz:Oct 4 13:39:25 mull kernel: sym53c8xx: 53c895 detected
with Symbios NVRAM
messages.1.gz:Oct 4 13:39:25 mull kernel: sym53c895-0: rev=0x01,
base=0xfebff800, io_port=0xe800, irq=10
messages.1.gz:Oct 4 13:39:25 mull kernel: sym53c895-0: Symbios format
NVRAM, ID 7, Fast-40, Parity Checking
messages.1.gz:Oct 4 13:39:25 mull kernel: sym53c895-0: initial
SCNTL3/DMODE/DCNTL/CTEST3/4/5 = (hex) 07/4e/a0/01/00/24
messages.1.gz:Oct 4 13:39:25 mull kernel: sym53c895-0: final
SCNTL3/DMODE/DCNTL/CTEST3/4/5 = (hex) 07/4e/80/00/08/24
messages.1.gz:Oct 4 13:39:25 mull kernel: sym53c895-0: on-chip RAM at
0xfebfe000
messages.1.gz:Oct 4 13:39:25 mull kernel: sym53c895-0: resetting,
command processing suspended for 2 seconds
messages.1.gz:Oct 4 13:39:25 mull kernel: sym53c895-0: restart (scsi
reset).
messages.1.gz:Oct 4 13:39:25 mull kernel: sym53c895-0: enabling clock
multiplier
messages.1.gz:Oct 4 13:39:25 mull kernel: sym53c895-0: Downloading SCSI
SCRIPTS.
messages.1.gz:Oct 4 13:39:25 mull kernel: scsi0 : sym53c8xx - version
1.3g
messages.1.gz:Oct 4 13:39:25 mull kernel: sym53c895-0: command
processing resumed

this message appears in syslog the night after a reboot, at the time
when I run the backup script.

messages.1.gz:Oct 5 01:00:00 mull kernel: sym53c895-0-<3,*>: FAST-10
SCSI 10.0 MB/s (100 ns, offset 31)

Could it mean that for some reason my scsi channel is running at 10MB/s?

Or is it related to the st driver being loaded? it does not appear when
I insmod && rmmod st.o.

thanks for any help

MAtteo

--
MAtteo Valsasna - matteo....@uninsubria.it
Network and UNIX administrator
SIC - Universita` degli Studi dell'insubria (sede di Como)
voice ++39-031-238.9720 fax ++39-031-238.9709
----------------------------------------------------------------
Oxymoron: Unachievable Accomplishments.

[MAtteo HCE Valsasna]

I reply to my own message

I compiled the kernel with the -DHP switch (added to CFLAGS in the
Makefile)

the only effect was to print the correct firmware and version numbers at
boot.

nothing is printed when I cat /proc/scsi/megaraid/1

no improvement in performance (seems the drives are stuck at 10MB/s)

MAtteo

The cat that ate the ball of yarn....had mittens!