write performance problem with LSI Logic SAS1068

Asked by nicholas fellows

Having a bit of a problem with our web 2.0 startup.

Were using Ubuntu Server edition (on one of our DB boxes) its an HP DL145 G3 box with a
LSI Logic SAS1068 integrated controller.

Problem is our hosting providers are now saying they wont support
ubuntu server (fiesty).

Issue is - I have really really poor write performance using the above set-up our project pretty much demands good write performance. (our el cheapo dell box here is acheiving in excess of processing 200 products per second - in contrast the Dell is only able to do 20/30) -- we've done a lot of code optimisation / mysql tweakage and im happy that side of things is pretty nailed.

From lspci

02:01.0 SCSI storage controller: LSI Logic / Symbios Logic SAS1068 PCI-X Fusion-MPT SAS

I've tried a bunch of things , enabling write cacheing with sdparm and lsiutil -
Im not convinced this has neccesarily worked though.
because i get errors when i try to check ...

sdparm -6 -vv --get=WCE /dev/sda
mp_settings: page,subpage=0x8,0x0 num=1
  [0x8,0x0] pdt=0 start_byte=0x2 start_bit=2 num_bits=1 val=0 acronym: WCE
    inquiry cdb: 12 00 00 00 24 00
    /dev/sda: LSILOGIC Logical Volume 3000
    Request Sense cmd: 03 00 00 00 40 00
    request sense: requested 64 bytes but got 20 bytes
    mode sense (6) cdb: 1a 00 08 00 08 00
sg_get_mode_page_types: current values: given response length too small, offset=12 given_len=8 bd_len=8

WCE error in Caching (SBC) mode page

When I try to set write cacheing in lsiutil - it seems to think it has been successful.
with sdparm i get this.

sdparm -6 -vv --set WCE=1 /dev/sda
mp_settings: page,subpage=0x8,0x0 num=1
  pdt=0 start_byte=0x2 start_bit=2 num_bits=1 val=1 acronym: WCE
    inquiry cdb: 12 00 00 00 24 00
    /dev/sda: LSILOGIC Logical Volume 3000
    Request Sense cmd: 03 00 00 00 40 00
    request sense: requested 64 bytes but got 20 bytes
    mode sense (6) cdb: 1a 00 08 00 04 00
    mode sense (6) cdb: 1a 00 08 00 04 00
change_mode_page: page offset failed: given response length too small, offset=12 given_len=4 bd_len=8

.

Having done a bit of research over the past few days im leading in the direction of thinking that
updating the kernel driver might be an idea

 lsmod | grep mpt
mptctl 26116 0
mptsas 28040 3
mptscsih 25344 1 mptsas
mptbase 53344 3 mptctl,mptsas,mptscsih
scsi_transport_sas 29568 1 mptsas
scsi_mod 142220 7 megaraid_sas,sg,sd_mod,mptsas,mptscsih,scsi_transport_sas,libata

I've managed to get hold of the driver source code from LSI and compile it successfully/
however im really reluctant to try this on a production server and landing myself a hefty support charge.

Im wondering if anyone can shed any light onto this really frustrating problem. Or if there is a
supported reliable way to fix it ? Id rather not stray too far away from the official ubuntu package manager at this time.

Is this problem fixed in Gutsy (we are on feisty) and would moving to 64bit version help?

Kind Regards

Nick ...

Question information

Language:
English Edit question
Status:
Answered
For:
Ubuntu Edit question
Assignee:
No assignee Edit question
Last query:
Last reply:
Revision history for this message
Alan Pope 🍺🐧🐱 🦄 (popey) said :
#1

If you want support from your hosting provider would it not make sense to either switch hosting provider or switch platform (to another Linux distro they do support perhaps)?

I appreciate this doesn't solve your issue at hand, but if you need your production server to be up and running it really needs some level of support.

Revision history for this message
nicholas fellows (nick-veedow) said :
#2

Thanks for those words of wisdom ;)

Trouble is we chose the provider because when specced out the requirement they told us they could support us.
They have been fine for most things but now this fairly mission critical problem has come up they are
backing down.

Right now changing host is not an immediate solution - although after my experience with them I think its on the cards.

Either way i've posted this in the bug's section now since its probably the proper place for it.

N.

Revision history for this message
Dan Shearer (danshearer) said :
#3

> Either way i've posted this in the bug's section now since its probably the proper place for it.

Could you mark the question as 'Solved' then?

Dan Shearer

Revision history for this message
MattW (seattle) (mbw) said :
#4

In the spirit of troubleshooting & things that have save me in the past, here
are a couple of other things to try:

- Make sure you have the latest firmware on the LSI card

- Make sure you have the latest firmware on your system's motherboard

- Have you tried different SAS cables, if its an external array?

-Matt

Can you help with this problem?

Provide an answer of your own, or ask nicholas fellows for more information if necessary.

To post a message you must log in.