E200 with Linux

[Adam Hansen]

You guys are going to love this. I talked to CDW and they say the e200 is the economy controller and said that I needed to go with the P400 controller. So I plop down $500, connect the controller and run this command:

dd if=/dev/zero of=/vm/bigfile count=10 bs=500M

and get an average speed of 12.6 MB/s... this is slower than the e200. I'm about at the end of my rope.

[KastL]

Does everyone have backup battery installed?

[Daniele Palumbo]

i have it.

but this is not relevant, the real question should be: who have acceleration enabled?

i do.
(and i guess all of us have it)

bye
d.

[KastL]

You see I don't have E200 yet, but I decided to buy one. I've searched for information about this controller and found this thread. I asked local distributer about low speed problems with E200 controller in Linux and he told me that it's not a bug it's a feature. It writes slowly only without backup battery because the write cash is not protected without backup battery. And low write speed is a feature to protect data when it's been writing.
I'm planing to install RHEL 5.2

[Daniele Palumbo]

feel free to think what you want, and to belive to who want to sell you the controller.

imho, also WITHOUT the battery backup the performance described in this thread are very weak.


bye
d.

[Adam Hansen]

I have the battery backup - see my specs above. I'm going to re-run the test over the weekend hoping that the array controller ramped up the cache which speeds things up.

Do not buy the e200 if you need any write performance.

[PC Doc]

Has anyone found a solution to this 'feature' yet. We have set up a main production HP ML150 G5 and failover backup 'off the shelf' server running ubuntu 8.04.1 server with VM's of windows 2003 domain and terminal server guests.

The HP is running nice and quick Xeon quad cores, and the backup is running standard Intel Quad E1200.

We have had to resort to running the backup server as the main production server because of the lack of speed on the ML150 G5.

I NEED a solution to this, and not HP's solution of 'buy another controller and run SAS'.



Why sell this controller if its so bad and limited.

Might be looking at ringing consumer affairs as I don't understand why anyone would sell a machine marketed as a server, with transfer speeds this low.

[Adam Hansen]

Well - I have some news re: this problem.

I found another link on this site that talked about disabling the on-controller cache because leaving this feature on disabled the write cache on the respective drives. The thread said disabling this feature moved a person with an e200i controller from 8MB/s to 120 MB/s for writes.

However, after spending hours and hours and hours of time trying to sort this out, I called CDW, got an RMA for my controllers (e200i and P400) and got an Adaptec 3805 with the necessary cables. I put it in and my write performance went from 12.8MB/s to 175MB/s. I definitely recommend this card.

If you do decide to ditch the HP cards and go with the Adaptec cards, here is what you need to know:

- If you have a server with an internal drive cage, you'll need to get the mini interface to amphenol connector (2 if you have 6 drives) in order to make it work.

- If you already have a config (e.g. array) on the drives, you will want to blow the config away using the HP controller before swapping the controller out.

- Definitely update the firmware on the system mainboard and on the array controller. This is huge - we had a problem where the controller couldn't see the drives until we loaded new controller firmware.

More food for thought - Adaptec support is excellent. If you think you might need to call them, make sure you have your LSI number (comes in the box) and you know what kind of drives you have (pull one out - write down the manufacturer e.g. Seagate, model e.g. ES2 and firmware).

Hope this helps...

[PC Doc]

That sounds fantastic. Once again, why do HP sell this POS.

You said you got an RMA for the E200, and got another RAID card. Do you mean HP gave you a credit on the card, and you purchased another one through them, or they gave you money back, and you purchased the card elsewhere?

Once again, thanks for the response, and that does sound like the way to go. I researched the specs of the machine before purchasing it, and didn't find any problems on various linux forums. As the card was supported, I assumed it would be full speed, and no where near as bad as it is.

[Adam Hansen]

I buy everything through CDW, so I just called my guy, said the card was terribly slow and said that I needed an RMA. I was within the RMA window, so he just cut me an RMA and the rest is history...

[Daniele Palumbo]

I have some good news...

but first please do a check:
install slackware 12.1 (also full) and run the test above sent by me.

this is not yet a solution, but it is at least something that we can verify.

i have updated the array to last firmware at yesterday (firmware cd version 8.30), and rebuilt the array with raid 1+0 on 4 sata 160GB Seagate Barracuda ES

thanks

[PC Doc]

And....

Does it improve the write speed any.

I have managed to get the predominate resellers of HP servers in Australia to ask users what purpose they want to user the servers for if they have an E200 card in their order, so they can notify them of the 'feature' of slow write speeds.

If the latest firmware fixes or improves dramatically the write speed, I will be updating ASAP.

[kairu0]

I've confirmed that the writing speed is pathetic still on the latest firmware 1.80, using both Debian Unstable and CentOS 5.2. System load is >6 when performing a large copy. :-(

Compared to the onboard SATA, this "upgrade" is really not performing well...

[Marco Stefanetti]

Hi all, I had the same problems, an apparently slow E200, but lately It's getting better and my tests are not so bad:

CentosOs 5.2 64bit

uname -a
Linux myname 2.6.18-92.el5 #1 SMP Tue Jun 10 18:51:06 EDT 2008 x86_64 x86_64 x86_64 GNU/Linux

hpasmcli> show server
System : ProLiant DL180 G5

hpaducli -f adu
ADU Version 8.10.4.0
Diagnostic Module Version 4.9-85
Time Generated Thursday December 18, 2008 1:45:57AM
Device Summary: Smart Array E200 in slot ATTR_VALUE_SLOT_UNKNOWN
Consolidated Error Report:
Report for Smart Array E200 in slot ATTR_VALUE_SLOT_UNKNOWN
Smart Array E200 in slot ATTR_VALUE_SLOT_UNKNOWN : Identify Controller
Configured Logical Drives 2 (0x02)
Configuration Signature 0xa00872cb
RAM Firmware Revision 1.78
ROM Firmware Revision 1.78
Hardware Revision 0x00

2x HP SATA 250GB Raid 1+0
1x Maxtor 250GB no raid

Cache Ratio 50% 50%

Array Configuration Utility 8.10.2.0 : Physical Drive write cache ENABLED (really dangerous???)

Raid disks:
dd if=/dev/zero of=/root/bigfile count=10 bs=500M
10+0 records in
10+0 records out
5242880000 bytes (5.2 GB) copied, 74.0209 seconds, 70.8 MB/s

Single disk:
dd if=/dev/zero of=/backup/bigfile count=10 bs=500M
10+0 records in
10+0 records out
5242880000 bytes (5.2 GB) copied, 86.7074 seconds, 60.5 MB/s


I tried many time on different disk partition and got between 55 and 80 MB/s.
I had all your same problems, I think the difference was enabling "physical drive write cache"


any improvement appreciated, bye

[Daniele Palumbo]

i have all kind of cache enables, and i guess that most of users here tested it or phoned at HP and they told us to do that.

the simplest test is to turn on the machine, run a dd, and then make login throug ssh from another machine.

it should be REALLY slow.

[PC Doc]

Question for Marco Stefanetti.

How exactly did you enable physical drive write caching. I have the SATA drives as supplied by HP for the hot swap bay. I had heard it was possible to turn the cache back on, but don't know how to do it.

I have 500 Gb drives installed and have run the dd if= of= test for 500M and 2000M files in raid 0, 0+1 and 5 modes on the E200, then removed the card and ran the same tests running mdadm raid 1. I got better results without the E200 installed.

At no stage have I been able to get above 50M/sec except with single non hardware or software raid. If I can get your speeds running with the E200 in raid 0+1 I will be ecstatic.

The unit is a HP ML150 with 8GB RAM running Ubuntu 8.04 64bit with windows 2003 Term server in VM for 20 users. When outlook 2007 is running for 3 clients, hardware interupts get up around 40%.

From my research, I feel it is because of the poor write speed to the Virtual drives causing the problem. I made the adjustments per http://blog.dhampir.no/content/performance-tuning-vmware-server

These changes have improved performance a lot, but the system is still not performing as well as it could running a native windows 2003 term server.

Once again, could you please advise how you enable physical drive write caching.

Thank you.

[Marco Stefanetti]

Hi, I got some improvements removing hp drivers and upgrading to kernel 2.6.18-92.1.22.el5,
cache ratio 75% write, 25% read

This is my last test:

dd if=/dev/zero of=/root/bigfile count=10 bs=500M
10+0 records in
10+0 records out
5242880000 bytes (5.2 GB) copied, 46.3108 seconds, 113 MB/s


Daniele Palumbo:

Yes, I have 4 active virtual machines and if I run the dd command, all became slower, and the shell is very slow, but working

Pc Doc:

I installed from hp site the following pakages:

cpqacuxe-8.10-2.noarch.rpm
cpq_cciss-3.6.18-23.rhel5.x86_64.rpm REMOVED!!
hpacucli-8.10-2.noarch.rpm
hpadu-8.10-4.noarch.rpm
hp-health-8.1.1-14.rhel5.x86_64.rpm
hpsmh-2.1.7-168.linux.x86_64.rpm
hp-smh-templates-8.1.1-19.noarch.rpm
hp-snmp-agents-8.1.1-22.rhel5.x86_64.rpm

I did't find a command line tool to enable write cache, I go to this url: https://localhost:2381
and then "HP Array configuration utility", common tasks, advanced features, Physical Write Drive Cache


bye

[cleepdar]

Enabling the Drive Write Cache from hpacucli is well documented.

See the document "Configuring Arrays on HP Smart Array
Controllers - Reference Guide"

The command should look something like this:

hpacucli ctrl slot=0 modify dwc=enable

[PC Doc]

Thank you very very much..

We have been battling with this machine for a couple of months now, we were relying on the backup machine for a while as its throughput was greater than the HP.

Final question, are you using SATA2 drives, or SATA1. It appears the drives that come with the machine are only SATA1, or at least limited to 1.5Gb/s. Is this correct, or are they 3.0Gb/s drives that are limited by the card?

Pity HP 2nd level techs couldn't give me this information.

[alexgva]

I've tried what Marco suggested but no luck with me (raid 1+0, 2 x 160GB seagate baracuda 7200, firmware HPGD).
I installed CentOS(64) hoping to reach the 70 MB/s and more but nothing to do... max 33 MB/s
The only difference I can see is the E200 firmware. Mine is 1.80..... Marco Stefanetti is 1.78.
Previoulsy I had tried SLES 10x(32) and the best I could get was 31MB/s (write cache enabled...even 100% write)
With W2K03 I was able to reach 35MB/s when I disabled the windows file paging (write cache enabled, 75% write).......


Now I'm wondering if I should downgrade the firmware if this is possible..

Marco, If you had a roadmap I could try to reproduce and validate your install. This could perhaps help other users.

Any suggestion is welcome.

Alex

[Marco Stefanetti]

*** Cleepdar

thanks, It works:

hpacucli ctrl sn=PA6C90P9SWK1IY modify dwc=enable

but can't use or set slot parameter.

hpacucli ctrl sn=PA6C90P9SWK1IY show

Smart Array E200 in Slot ATTR_VALUE_SLOT_UNKNOWN
Bus Interface: PCI
Slot: Slot Unknown
Serial Number: PA6C90P9SWK1IY
Cache Serial Number: P9A3A0B9SWK8D2
RAID 6 (ADG) Status: Disabled
Controller Status: OK
Chassis Slot:
Hardware Revision: Rev A
Firmware Version: 1.78
Rebuild Priority: Medium
Expand Priority: Medium
Surface Scan Delay: 15 sec
Cache Board Present: True
Cache Status: OK
Accelerator Ratio: 25% Read / 75% Write
Drive Write Cache: Enabled
Total Cache Size: 128 MB
Battery Pack Count: 1
Battery Status: OK
SATA NCQ Supported: False


*** PC Doc

raid 1+0, two identical disks

Physical Drive at Port 1I : Box 1 : Bay 1, Internal
Controller Smart Array E200
Bus Interface PCI Express
Controller Location Unknown
Location Port 1I : Box 1 : Bay 1, Internal
Size 250 GB
Type SATA
PHY Count 1
Transfer Speed 1.5 Gbps
Firmware Version HPG2
Model ATA GB0250C8045
Status OK
S.M.A.R.T. Predictive Failure Events No

*** alexgva

I'm sorry, no roadmap. I installed centOs normally, then, when I got into the slow transfer rate, I began to install and unistall drivers and packages many times.
I can give you any detail of my actual configuration, but, by now, I will not reinstall!!

[Lukas Hubschmid]

Hello.

In this thread, it's only a problem about the write performance ...

How is the read performance of the E200/128 BBWC? Any experience?

Greez
Lukas

[sirrus]

I have this issue too.

E200 + 128MB BWCC

Latest 1.84B firmware

Io Scheduler: cfq (even tried deadline and noop with almost no change)

Very poor write performance (10MB/s) and read performance of 54MB/s), IO Wait 100% and up to 4s (seconds not ms!!!!) write queue delay.

Enabling DWC results in 54MB/s read and 45MB/sec write

The same harddrives as mdraid 1 connected with the HP ML110 onboard SATA - 90MB/s read and 80MB/s write performance....

I will remove my e200 and migrate to mdraid!

By the way - I've installed two new 2TB harddrives. Onbord performance 130MB/s read and  120MB/s write.

On the e200 is poor: 64MB/s read and 50 MB/s write.

To be onest - this is disgusting!