YAAC - Yet Another Atlas Copy

[c3]

Yup, not sure how many of these there are, but I am building one too! Thanks John

 

Like others, I see the need to consolidate machines and I already have a few ESX hosts and a couple iSCSI storage arrays. So, this is the first step at combining those.

 

YAAC, the first, will host an unRAID guest using a pass-thru M1015 controller to support up to 8 drives. Other guests will include XBMC with centralize mySQL db. YAAC, as a host will join my existing ESX cluster and potentially use the iSCSI storage.

 

Case: Norco 4244 - not sure the version, it has 2 USB up front, and 80mm fans.

MB: SM X9SCM-F

CPU: E3-1230

Heat Sink: Intel OEM

Memory: Kingston 4x4GB ECC

Power Supply: Seasonic X750

SATA Controller: M1015

Boot/ESX datastore: 750GB Seagate 7200rpm

unRAID drives: (5) 2TB Seagate LP 5900rpm

iSCSI drives: (7) 2TB Hitachi 7200rpm

 

UPDATE 12-13-11: Third motherboard is now running ESX, full RAM (total 4x4GB), using the reverse breakout for ESX, and SFF8087 cables.

With the Seasonic X750 and leftover cables from prior Seasonic X builds, the 1-to-7 is not needed. I just run (4) of the 3 molex cables up to the backplanes.

 

I would like to get fan speed control. Anyone doing done that for the 4224? My reason is power, not noise. 6 fans at full speed 24x365 is just wasted energy.

I join the club of wondering what those (4) switches on each backplane are for?

[c3]

I started collecting parts a while back, but with the arrival of an open-box X9SCM, building began...

 

First motherboard came in with visible CPU socket damage http://c3images.posterous.com/bent-pin#

Second motherboard came in with visible CPU socket damage http://c3images.posterous.com/bent-pin2#

NewEgg RMA'd them both, hope they bin them and not try to sell on.

 

UPDATE 12/10/11: A third motherboard is now running ESX !

 

[nia]

Great to have more at the Atlas party 

 

Second motherboard came in with visible CPU socket damage http://c3images.posterous.com/bent-pin2#

 

But I boldly when ahead and fiddled with motherboard #2. After a brief session in very bright sunlight and a magnifying glass, I was satisfied enough to install the E3-1230. I only have 2x4GB memory, so I would not be able to detect the problem seen by nia. I have ordered 2 more so I can get that tested.

The pins bent on your board are different than the one that was bent on mine related to the memory controller. Can't say if you have other lurking problems from those pins, but luckily you have another board coming to remove any doubt as I read your post. That being said, my unbending appears to have been sucessful, as I have had no issues since. So you could very well have a successful run with board #2 as well 

 

Enjoy your building efforts - may you have a smooth run at it...

[c3]

Newegg is awesome, they RMA'd motherboard#1 and motherboard#2. I have requested repair estimate from SuperMicro, rumor is $75.

 

Any hints as to a super test for motherboards? (ya, ESX, unraid vm, and parity check...)

[Johnm]

what a run of bad luck with those Mobo's.

 

I know you had pm'd me about the first board.

 

I would like to point out that I revived 3 of these boards open box and had not 1 issue.

 

The first board i Believe was listed as "asian sample board" instead of "open box" and came in and X8 box.

the rest came looking brand new and worked perfectly.

 

newegg is usually a pleasure to work with and will bend over backwards for the customer.

 

Yaac... LoLz ..

 

Good luck whit the rest of your build.

 

actually, are you planning using an iSCSI target for you datastore are are you going to keep it all in one?

[c3]

Well, I am investigating having SuperMicro replace the sockets. I could attempt self repair, but it would always be a nag when odd things happen.

 

I had hoped that the open-boxes were due to marginal power supplies, as they do require the real stuff, or just people not wanting to deal with roll your own ESX LAN drivers. But careless installers, passing the buck is bad karma, shame on you. My only disappointment with NewEgg is the claim to have tested functional. They are accepting the returns (if SuperMicro wont fix). I have asked to talk with RMA testing about the process to avoid this in future.

 

All my ESX hosts have access to the iSCSI storage, YAAC will too. I have a dual port 1Gbe so four ports on 3 vSwitches, management gets redundant. The whole setup is complicated, but I'll add to the description as I go.

[Johnm]

All my ESX hosts have access to the iSCSI storage, YAAC will too. I have a dual port 1Gbe so four ports on 3 vSwitches, management gets redundant. The whole setup is complicated, but I'll add to the description as I go.

 

Nice. i was very tempted to go this route and might still in the future. esecially since the X9boards can connect the NIC straight to the target natively.

 

My ultimate choice with atlas was "this box is to cut down the number of servers I have and go greener".

With that, i decided that my iSCSI target did not fit that profile (16 drive raid6).

 

maybe one day I'll consider a 4 drive target that runs on an Atom.

[c3]

Third motherboard arrived, no visible damage, ESX installed. Awaiting additional RAM and cables. Top post has details.

To-Do: flash M1015

[savestheday]

Hi C3,

I copied John's fans too and they come with fan controller cables that kick the speed down.  I had Coolermaster PFM fans in there and had the BIOS control them but they would spin up in very consistent waves so it I think the temp stuff is off in ESXi (did this in unRAID direct too).  I have (3) 120mm Noctua's running at full speed and (2) 80mm Noctua's in the back.  You can hear them but they're not too loud.  In fact, my home theater is in the basement about 15-20 ft from my ESXi server and it is barely audible. 

[c3]

I am ordering the 120mm fan plate and some Delta QFR1212GHE-PWM fans. No idea what the outcome will be. I expect to have a bit of work to do to get them operating properly, but at least they will lower the power demand when able.

 

Without installing a Supermicro CIM, ESX has no chance at properly controlling the fans. And the Supermicro CIM would be for Supermicro chassis fans. So, I would not recommend using it.

[savestheday]

I am ordering the 120mm fan plate and some Delta QFR1212GHE-PWM fans. No idea what the outcome will be. I expect to have a bit of work to do to get them operating properly, but at least they will lower the power demand when able.

 

Without installing a Supermicro CIM, ESX has no chance at properly controlling the fans. And the Supermicro CIM would be for Supermicro chassis fans. So, I would not recommend using it.

 

ESXi is based on Linux no?  Linux certainly has fan control.  My Supermicro mobo has PFM connectors and settings in the BIOS to control fans, ESXi (nor unRAID) does it properly tho.  I'd expect them to run at a low speed and spin up when needed but that didn't happen.

[c3]

ESXi is based on Linux no?  Linux certainly has fan control.  My Supermicro mobo has PFM connectors and settings in the BIOS to control fans, ESXi (nor unRAID) does it properly tho.  I'd expect them to run at a low speed and spin up when needed but that didn't happen.

 

Nope, common misconception, but ESX is not based on Linux, nor is it open sourced, and a few thousand more things... In short ESX is not KVM, the one in Linux. The ESX Console (v1-4) was Linux based. The console is not the hypervisor.

 

Sure the motherboard has connectors, but since you're not in a supermicro chassis, which fan is connected to what? On my X9SCM, which is the CPU fan? and on yours? So, what is the fan control to do when CPU reads overtemp? Spin them all up? Which sensors for heat should be used? These are things the CIM does, and without it, do not expect good results. Using the wrong CIM could have negative effects.

 

[c3]

I got a response from Supermicro on the fan power available on the X9SCM.  Supermicro hardware designer says, 5A is the recommended limit for PCIe 12V and motherboard fans combined (60W). M1015 and the like are around 10W each. My powerful fans might push things.

[Johnm]

correct.

 

If you get a supermicro case for this board (and most current SM Boards), if you plug the supermicro fans to the boards headers, the systems become underpowered and wont boot.

 

I have my CPU fan and PWM exhaust fans on my motherboard and my fanwall fans on a molex connector.

 

Thank you for reminding the forum users of this limitation.

 

[c3]

I was suspicious when I read the manual on the fan headers. It says 2.5A per header, which is huge. With 5 fan headers, those alone are 7.5A. So, I asked. Since PCIe load can be considerable, that 2.5A is kind of fake.