New LSI 9200 series controllers: 6Gb/s, 2.88 GB/s seq. reads, 1.87 GB/s seq. writes

[Mastakilla]

hmmm that sounds really dangerous (don't want to loose my data)

but I've got some progress in meanwhile!!

I've been trying some more stuff and at some point I tried my RAID controller without any HDs attached... and guess what:
WEBBIOS STARTS!

After that I re-attached all HDs, reloaded my BIOS settings and tried again
WebBIOS still starts!

Finally I reinstalled the new LSI firmware again and also their "the trick" works

If any of you has an idea why the *$*$ WebBIOS won't start when my fully functional and healthy RAID5 is attached, then I'm all ears

btw: Since those SATA HDs are hot pluggable, could I just open WebBIOS and then attach the HDs?

edit:
really gotta go to bed now... 5 hours left to sleep
THANKS again for the support

[trans am]

btw: Since those SATA HDs are hot pluggable, could I just open WebBIOS and then attach the HDs?

whoa. i would not recommend it. what data are you worried about on the raid set? is it just the os or do you have some important bench runs on that set?

dont start plugging in drives while the computer is on. plug and play means nothing especially on ssds and expensive raid cards. just get in the web bios and set the thing to factory defaults and destroy the set and start over from scratch. once you have it all fresh you can unplug the drives and id the cables and drives like I mentioned and do it right this time. then go in and use the auto function and create the new set. if you really want to do this right you will take each drive off the card and plug them one by one to a sata header on the board and wipe each one separately an rebuild the set but my guess is you are not going to spend the next 2 days doing this. so just erase the set you have and rebuild the set and dont kill any hardware in the process. you are walking on eggshells man and paul and i are gonna be pissed if you screw this up. You need to get in the lsi wb bios setting that allows you to reset to defaults and reboot and start over. this is the only way to make sure you dont waste anymore time. or just take your chances and see what happens. I am totally recless on cpuz and boards, mem vga but I take every precaution on this card and expensive ssds. the time it takes to get them just right is nothing in comparison to any other hardware component. call it a black art. optimizing raid cards and performance ssds/ram drives sets is the highest form of hardware tweaking possible. forget about htt and bclk and maximizing your $300 ddr3 eplida hypers. they are nothing in comparison to the frustration an time dedicated to an optimal enthusiast raid setup. You have your work cut out for you. we are glad to help from what we know but only you can make this happen in the end. go to bed. get some sleep and start over tomorrow night. if you get into this now you are jsut going to get obsessed and pissed off and you might get risk getting fired from your job. just take a deep breath at least you got in the web bios. we'll help you tomorrow.

[Mastakilla]

hehe

ok that's clear... before I had "real" data on my RAID5 I did some very heavy stresstesting of my controller / HD combo (the HDs are not officially supported)
That was including pulling SATA cables while running many benchs and apps at the same time :p

It did work very well then, but I'm not sure how WebBIOS will react on it

The RAID5 does not contain bench runs :p
It is actually almost full with data (I've spend months ripping my DVD collection on those HDs so I can stop being a disc jockey :p)

So no, re-creating the array from scratch is absolutely not an option

[Computurd]

hmmm that sounds really dangerous (don't want to loose my data)

yeah i would agree with that

no idea but we see that now you can get in so that is good news actually no idea why, but there is a guy over at OCZ forums who has the same issue with an ASUS board.

they are nothing in comparison to the frustration an time dedicated to an optimal enthusiast raid setup.

BIG plus 1 there

[trans am]

yeah i would agree with that

no idea but we see that now you can get in so that is good news actually no idea why, but there is a guy over at OCZ forums who has the same issue with an ASUS board.



BIG plus 1 there

paul its true. you me and steve and tilt spent last newyears eve tearing our hair out on these things. when i came into this game I knew nothing and went through hell and back. between us lsi should be sending us cards on a monthly basis. there is no way their engineers have spent a a fraction of their time on the 92xx series as we have. As for our buddy I am completely out of ideas. he shot himself in the foot by ripping all those dvds to this raid set. I look at these sets as raid0 only like the majority of us on these cards. I think of them as disposable raid sets because there is no end to optimization on them and then a new board comes into the picture for example the last oct and nov hw pcmark contests. THere is no way in hell i would put anything important on a raid set like this. thats what my 1tb F3's are for. I feel bad for the guy and hope he can get his data off this set to a big storage drive but im very pessimistic. I had my entire star trek next generation archive on a f1 a while back and lost it because I had that disk plugged in during a windows install. I spent the next month ripping them all over from netflix and learned the hard way. I have never tried raid 5 but If i were you I would get all the best disks on a raid0 set off the lsi and have your data standalone on a 1-2TB F3 or similar. if you want to get in this segment of hardware frustration you will not want any crucial data in the equation. in order to get your sweet spots you will have gone through several stripe settings, several reformats and new sets. you have no idea what you just got yourself into man. I think at the end he is gonna need to redo the whole set and get back in the netflix cue history.

[eagle101]

nothing but problems LSI, screw you LSI they lied to me

Hi all,

I just discovered that I can't start WebBIOS anymore.

I do not know since when / what this is, since I have only used WebBIOS during the initial setup of the RAID5

A few weeks ago I have updated my LSI's firmware, driver and all LSI software to the latest version (still the latest now). MSM and the RAID5 kept on working great...

What happens is:
* I boot my PC
* I press CTRL-H when it sais I should press CTRL-H for WebBIOS
* It quickly shows the message "WebBIOS will be executed after the POST is over"
* It proceeds to boot but never enters WebBIOS

First it just proceeded by booting Windows, then I tried changing my Gigabyte onboard controllers from AHCI to IDE (the AHCI screen of these controllers came AFTER my LSI boot screen), then I disconnected my SSD (OS drive), then I disabled all onboard controllers except ICH (which had nothing attached) and finally I tried all possible combinations of this... Nothing helps...
If the OS drive is attached, it simply boots the OS... If the OS drive isn't attached, it simply stops booting after the POST (blinking cursor, no error at all)
I tried starting the pre-boot CLI, which works (but it also doesn't start post-POST, like the WebBIOS)

Does anyone have an idea how I can "fix" WebBIOS?

edit1:
Only found this similar issue:
http://www.fixya.com/support/t175482...web_bios_start
Their solution is to remove HDs from the onboard controller, but I already tried that and it didn't help for me...

I just also tried the following:
* Removed DVDROM and SSD (so that there are no devices attached to the onboard controller)
* Removed the 2 new HDs from my LSI RAID controller (so that only the 4 HDs are attached that I initially setup using WebBIOS)
* Loaded fail save settings in the BIOS (although I didn't overclock yet, this should reset all my "optimizations" to the most default BIOS settings)

The above also did NOT help

Anyone still has ideas?

edit2:
I just had a very long call with LSI support (friendly helpful people btw)

He thought it had to do with insufficient memory to load WebBIOS (a server mobo has more memory or something?) and suggested to disable as much as possible in the MOBO BIOS.
So I loaded fail safe settings and then disabled anything I could (including disconnecting as much hardware as possible), but it didn't help either...

Then he suggested to re-install the old firmware of my RAID Controller, since the new firmware has a lot of new advanced features that I'm not using (related to SSD), it might help free up some memory...
I told him I'll get back to him once I tried that.

So I just tried it, again including disabling / detaching as much as possible and this time I also tried loading "optimized default" first...
Again this didn't help...

So I'm stuck once more...

actually the only thing that is different now compared to when I used WebBIOS to initially setup the RAID5 is the BIOS of the MOBO I Think (F7 vs F6)

I'm not sure if that is so easy to roll back, since I do would like to keep my current F7 settings and I'm afraid rolling back will erase those...

So does anyone have an idea what is going on here?
Does anyone understand which memory the LSI support guy suggests is insufficient? Which memory is being used to load such post-POST BIOS applications? How do there post-POST (but pre-WhatEverIsSelectedInTheBIOSAsBootDevice) BIOS applications actually work? Is there a name for such applications (so I can google for it to find out more about it)?

Thanks!

[Mastakilla]

something bright just popped up in my mind...

My Gigabyte UD7 has a normal PC BIOS.
My RAID5 array is 1 large partition using GPT instead of MBR.
Only an EFI BIOS, and not a normal PC BIOS, can boot from a GPT...

Could this have something to do with my issue?
Is it possible that, if the controller detects a filesystem on its HDs, that it tries to start WebBIOS via that filesystem?
And when no filesystem is detected it must be using some other method (cause I can start WebBIOS if no HDs are attached)

when I arrive home this evening, I'll try starting WebBIOS with only the 2 new HDs attached (cause these HDs are not configured / partitioned yet).
If that works, the chance seems higher that the GPT is related to my issue...

and once more, if anyone has suggestions to solve my issue: please let me know

[Mastakilla]

WebBIOS also works when only attaching the unconfigured / unpartitioned HDs

I tried them in both ports, so it also isn't a problem with the ports...

In my eyes, this makes it more likely that the GPT is being the issue here...

[Computurd]

interesting, very interesting.
you gonna try EFI bios?
have you tried using the MSM?

[Mastakilla]

Gigabyte does not have EFI BIOS as far as I know...

Called LSI again yesterday:
IT was another guy this time, he still kept on claiming it was related to the firmware of LSI being too big and the CMOS on my mobo being too small.

He suggested me to go back to even an older firmware.

I don't believe that will help, but I'll try it later today just to make sure
If this is really a bug, then I want LSI to be convinced about this as well, so they can work on it...

The guy also explained me how to do it using MSM (p 285, chapter 8.7 Changing a Virtual Drive Configuration). It is called completely different as when you do it with WebBIOS (a more generic name), which is why I couldn't find it...

Anyway, I still prefer doing it with WebBIOS (no risk for blue screens)

Also I heart one of my 6 HDs clicking last few days (not the idle clicking, but the clicking of death). I suppose it is one of the new HDs, so I first have to figure out which one, so I can replace it...)

[Mastakilla]

hmm

flashing to the even older firmware has bricked my controller
no error occurred during the flashing, but still it fails to get past the firmware loading

I have to RMA now

[Tiltevros]

do u have PEG enable in PCIE settings?

[Mastakilla]

I do not have PEG settings for my Gigabyte mobo, but if "Init Display First" is the same, it is set for PCI-e x16 1, which is where my videocard is in...
So that does seem correct to me

What issue are you trying to solve by checking that?
The broken firmware or the WebBIOS issue?

[tived]

I am new here, but reading this below, is just amazing. I have to say its overwhelming that quality of good sound advise here. You are are truly a great bunch.

I take my hat off for you trans am!

Sorry for going OT here, but i think this is special!

Henrik
A Dane Down under

whoa. i would not recommend it. what data are you worried about on the raid set? is it just the os or do you have some important bench runs on that set?

dont start plugging in drives while the computer is on. plug and play means nothing especially on ssds and expensive raid cards. just get in the web bios and set the thing to factory defaults and destroy the set and start over from scratch. once you have it all fresh you can unplug the drives and id the cables and drives like I mentioned and do it right this time. then go in and use the auto function and create the new set. if you really want to do this right you will take each drive off the card and plug them one by one to a sata header on the board and wipe each one separately an rebuild the set but my guess is you are not going to spend the next 2 days doing this. so just erase the set you have and rebuild the set and dont kill any hardware in the process. you are walking on eggshells man and paul and i are gonna be pissed if you screw this up. You need to get in the lsi wb bios setting that allows you to reset to defaults and reboot and start over. this is the only way to make sure you dont waste anymore time. or just take your chances and see what happens. I am totally recless on cpuz and boards, mem vga but I take every precaution on this card and expensive ssds. the time it takes to get them just right is nothing in comparison to any other hardware component. call it a black art. optimizing raid cards and performance ssds/ram drives sets is the highest form of hardware tweaking possible. forget about htt and bclk and maximizing your $300 ddr3 eplida hypers. they are nothing in comparison to the frustration an time dedicated to an optimal enthusiast raid setup. You have your work cut out for you. we are glad to help from what we know but only you can make this happen in the end. go to bed. get some sleep and start over tomorrow night. if you get into this now you are jsut going to get obsessed and pissed off and you might get risk getting fired from your job. just take a deep breath at least you got in the web bios. we'll help you tomorrow.

[Tiltevros]

I do not have PEG settings for my Gigabyte mobo, but if "Init Display First" is the same, it is set for PCI-e x16 1, which is where my videocard is in...
So that does seem correct to me

What issue are you trying to solve by checking that?
The broken firmware or the WebBIOS issue?

controllers cant boot when u have PEG on. maybe this is ur problem.
server motherboard dont have that option thats why u havent heard any problems there..

try to set smaller size to packets size for PCIE and disable the PEG option
and post back.

thnx

[sethk]

Hello,
I've been searching through this thread trying to find out what I need. I'm looking to build a development use database server and need very high random IOPS (and occasionally seq for log use). While the data is important, I trust our backup procedure and if the database is down for a day or two while we recover day-old data that's fine (I may enable log shipping to get more fine grained data recovery). This is not a production use server, it just has several developers banging on it (with backup database servers on their PCs).

With all that said, is a LSI 9200 series card + 3 x 120GB sandforce drives a good idea for this use (at most there will be 4 drives)? Which card should I buy? Do I need fastpath? Will RAID 5 kill the IOPS and should I use RAID 0 instead? Suggestions and feedback are very welcome.

[mbreslin]

Short of doing a bazillion transactions/second raid 5 should be fine. My advice would be to get the 9260 without fastpath and then get fastpath if iops aren't high enough for your needs.

The 1880 is apparently pretty awesome if you don't have budget constraints.

If you are budget minded I could sell you a 9211-8i for pretty cheap. The sequentials aren't amazing and no ability to change stripe size but it does crank out a butload of iops.

[sethk]

Thanks for the feedback, I decided to get a 9260 (no fastpath yet).

Good tip regarding the Areca - I have older Arecas in other servers, but given the low drive limit for this server, it seemed overkill for this application. It certainly would be a good choice for a storage server.

[Mastakilla]

I just got my replacement LSI 9260-8i board.
Restoring my old RAID5 on the new controller was pretty easy. I didn't need to do anything, just plug in my HDs and my RAID5 worked on the new controller.

The new controller contained the very old 12.0.1-0081 (2.0.03-0725) firmware.

WebBIOS works with the new controller and this old firmware...

I now just flashed the controller to the new firmware 12.11.0-0016 (2.110.03-0967) and again WebBIOS stops working

So it is clearly not a hardware issue but an issue with the new firmware / WebBIOS / mobo-controller compability.

The issue only occurs when I attach my RAID5 of 4x 2TB containing a GPT partition table of 1 single big NTFS partition (I think GPT may be the trigger here)

Is anyone else having this same problem?

Please tell me:
mobo + BIOS
LSI firmware
RAID config
Partition type + config

Thanks in advance!

[Mastakilla]

controllers cant boot when u have PEG on. maybe this is ur problem.
server motherboard dont have that option thats why u havent heard any problems there..

try to set smaller size to packets size for PCIE and disable the PEG option
and post back.

thnx

Thanks for the answer Tiltevros, but I don't really get it...

My controller has no problem booting (actually I boot from the SSD and my RAID5 is an unbootable GPT partition) .The problem is that WebBIOS does not start
Also I don't have a server mobo. I have a consumer mobo. And I can't find any option that is named PEG...

Also I couldn't find anywhere the option to set a smaller size to packets for PCIE.

Any idea where to look next?

[F@32]

Just pulled a trigger on used 9260-4i for USD 195. Christmas came early hehe.

[Computurd]

cant beat that! will be fun

[F@32]

ok so I logged into OCAU forum this morning & there's a fellow selling an IBM M1015, which he claims is a LSI 9240-8i.. he wants $70.
google seems to confirm the claim
either way.. I've bought it.. newegg has the card for 270
I'm not sure how much they sell for here in australia, but for vague comparisons sake, a LSI 9260-8i on newegg is ~500, here in oz $700
so its seems i got myself a bargain
I dont know too much about these cards.. they dont appear to have any onboard cache, correct?
what XOR do they have? how many mhz?
is this.. 'fastpath key' a recommended addition?
cables are cheap enough on ebay..
any help, me would like, TIA!

Hell of a deal, mate. 9240 is pimped out 9211, even though it has no cache but it allows different stripe sizes. When 9211 starts hitting a wall at 64K> files sizes sequential reads, 9240 will blast right past it.

[Computurd]

excellent. this will be very interesting to find optimal stripe, as from my testing im sure that the 64k isnt it. still fast though
im very interested to see how this handles at larger/smaller sizes. that would be a great controller for me to pick up, thinking of splitting my array into two rigs....great thing about the 9211 was that you get fastpath speeds at such a cheap price!
dollar for dollar that 9240 could be hard to beat! what is the ETA on it? and if you dont mind, where did you pick it up?

[flamenko]

Contacted the company....hoping to grab the 9200 4port for review