Supermicro X7SPA L/H/HF ATOM serverboards (Level 1 Tested)

[Joe L.]

Just got my HF board finally yesterday... everything seems to be working fine, however when booting from my USB stick, somehow I have no active ethernet. I'm using the port that's combined with IPMI, and I know it works, because before I changed the boot order to boot from the USB stick, I could access the computer through the built-in webserver functionality of the board.

 

When I boot the 4.5.3 unraid server, I see the NIC lighting up during the boot phase, but when the boot is finished and the prompt appears, the lights are back off and ifconfig shows no address for the port.

 

Has anyone else seen this? What should I be looking at? I tried to change the network.cfg to disable DHCP and provide an IP, but no change, the lights on the NIC after booting are out...

 

Thanks for any hints!

 

-Christian

 

did you set the volume label on the flash drive to UNRAID?

 

What do you see when you type

ls -l /boot

 

If the flash drive is being mounted properly, identified by using the UNRAID label, you should see a listing of the flash drive.

 

If the flash drive does not get mounted, the networking config on it is not available and the networking is not enabled.

 

That is the most common reason you cannot access the built-in web-server, since the script to start it is also on the flash drive in the config folder.

[chrisb42]

Yes, I set the UNRAID label. I could easily go into /boot/config and change network.cfg with vi, so that shouldn't be the problem. I'll go over it again later today. Thanks for the tip, though.

 

-Christian

 

[Joe L.]

Yes, I set the UNRAID label. I could easily go into /boot/config and change network.cfg with vi, so that shouldn't be the problem. I'll go over it again later today. Thanks for the tip, though.

 

-Christian

 

Historically it has been the most common reason connectivity did not get established.  If you know how to use vi, then you are way more experiences then most beginning unRAID users in the Linux environment.

[chrisb42]

OK, problem was just a defective cable. Even though I saw messages in syslog about the interface being set up, the DHCP request got no response. Trying a different cable made it all work immediately!

 

Thanks all!

 

-Christian

 

[aiden]

The Bogomips was just for my own education where the processor will be if I replace one of my machines with this board. The current machine is a dual xeon lv 2.4 rig that is a firewall, proxy, IRC bot ana vmware server for ripping/transcoding. I'm just deciding if I want to give up some processing power or not. So the number is helpful in revealing the cpu is faster then an article I read which was incorrect.

 

Weebo did you ever use this board as a firewall replacement?  I'm just wondering because I am thinking about getting this to run Untangle.

[WeeboTech]

The Bogomips was just for my own education where the processor will be if I replace one of my machines with this board. The current machine is a dual xeon lv 2.4 rig that is a firewall, proxy, IRC bot ana vmware server for ripping/transcoding. I'm just deciding if I want to give up some processing power or not. So the number is helpful in revealing the cpu is faster then an article I read which was incorrect.

 

Weebo did you ever use this board as a firewall replacement?  I'm just wondering because I am thinking about getting this to run Untangle.

 

I have not yet, I'm still using my dual xeon machine until I migrate some services to a new adminstration Mobile core 2 duo machine. Just working with some of the newer OCZ vertex drives before I'm ready. Trying to eliminate spindles.

 

On the plus side I build a large unRAID server to house the static data and discovered a new way of speeding up unraid writes to parity.  Stay tuned to this and other bat channels. ;-)

[purko]

On the plus side I build a large unRAID server to house the static data and discovered a new way of speeding up unraid writes to parity.

Stay tuned to this and other bat channels. ;-)

I can't wait to hear some details about that.  Please give a hint.

 

[WeeboTech]

On the plus side I build a large unRAID server to house the static data and discovered a new way of speeding up unraid writes to parity.

Stay tuned to this and other bat channels. ;-)

I can't wait to hear some details about that.  Please give a hint.

 

 

RAID0 on the Parity drive is a real possibility now with 4.5.3 and advanced controllers.

SAFE33 (Mixed RAID0/RAID1) with the parity drive and a cache drive is a real possibility now with 4.5.3 and advanced controllers.

[ilovejedd]

RAID0 on the Parity drive is a real possibility now with 4.5.3 and advanced controllers.

SAFE33 (Mixed RAID0/RAID1) with the parity drive and a cache drive is a real possibility now with 4.5.3 and advanced controllers.

 

Nice find. I'm a bit afraid to put parity on a RAID0 array, though.

[WeeboTech]

RAID0 on the Parity drive is a real possibility now with 4.5.3 and advanced controllers.

SAFE33 (Mixed RAID0/RAID1) with the parity drive and a cache drive is a real possibility now with 4.5.3 and advanced controllers.

 

Nice find. I'm a bit afraid to put parity on a RAID0 array, though.

 

What's the worst thing that can happen? 

A drive dies, you put a new one in and re-sync parity.

Same situation if a data drive or the whole parity drive died.

 

As long as you have a spare drive on hand, or a plan, then it should not make you afraid.

Also, the areca controller beeps loudly if you loose a hard drive.

 

This is something unraid does not do yet.

Unless you are checking unRAID on schedule, a drive could go off line and you would never know.

 

 

RAID0 on parity is only useful to build a larger drive from smaller drives and to try and improve parity speed if you are doing allot of writes. Otherwise, it's probably not worth the extra cost.

 

For me the SAFE33 the carved raid0/raid1 volumes provides me with a place for parity to protect other full disk volumes and a raid1 volume for continuous file system access (remote logs, nfs file access, database, vmware systems). 

 

If I had these continuously mounted/accessing/spinning file systems running raid1 anyway, might as well carve up some large quality drives and do double duty.

[Kaygee]

Souldn't raid0 offer a good performance hike over a single disk if done right?

[BRiT]

Kaygee, it should. However in the case of using raid-0 for a parity drive inside of unRAID, it's still bound by the rotational speed and access of the data drive.

[WeeboTech]

Souldn't raid0 offer a good performance hike over a single disk if done right?

 

It does on reads. I was getting high numbers on reads.

 

I would get parity sync speed of "Estimated speed: 130,239 KB/sec"  on the outer tracks.

 

 

[Kaygee]

Raid 0 if done right should put odd blocks on disk 0, even blocks on disk1. Assuming you can match unraid request sizes (128K I seem to remember?) to the raid cards blocks size, RAID 0 should read and write phenominally well.

 

I understand unRAID performance limitations and I think a RAID 0 parity drive should help out considerably, firstly it'll cache write operations so no waiting for a spindle rotation. The "next" sector should be ready to be read (read ahead caching) as it will be on another physical drive so no waiting for the last write operation to complete prior to the read occuring.

 

Realistically you should be able to get close to a data drives write performance.

 

 

 

     

[Joe L.]

Raid 0 if done right should put odd blocks on disk 0, even blocks on disk1. Assuming you can match unraid request sizes (128K I seem to remember?) to the raid cards blocks size, RAID 0 should read and write phenominally well.

 

I understand unRAID performance limitations and I think a RAID 0 parity drive should help out considerably, firstly it'll cache write operations so no waiting for a spindle rotation. The "next" sector should be ready to be read (read ahead caching) as it will be on another physical drive so no waiting for the last write operation to complete prior to the read occuring.

 

Realistically you should be able to get close to a data drives write performance.

Since you'll be waiting for the data disk to rotate between its read and subsequent write, having a lightning fast parity drive will not provide the result you think you will be getting.

  The SLOWEST rotational speed drive will dictate "write" speed to the array.

In the case of a parity drive that was instant, you would still be waiting for the data drive to spin around between reads, writes, and subsequent reads.

[WeeboTech]

Realistically you should be able to get close to a data drives write performance.    

 

It's not quite that fast. Fastest I saw was 43MB/s.

I think BubbaQ posted that this would be the fastest we can expect at of unRAID these days.

 

Without unRAID the drives were 80MB/s for the samsung spinpoint F3 on writes and 120-127MB/s Seagate 1.5TB 7200 RPM 32MB cache.

 

Where the samsung excels is in read speed. I was getting 149MB/s reads. while the Seagate was 120MB/s.

 

Next I'll put the data drive and the raid0 parity on caching controllers to help alleviate the rotational delays.

[sosdk]

If someone could just make this work as a Parity drive then parity speeds should be fast enough for even the fastest HD's....

 

http://www.ocztechnology.com/aboutocz/press/2010/369

 

Speed claims of up to 1.4GBit/Sec....that should be enough 

[WeeboTech]

If someone could just make this work as a Parity drive then parity speeds should be fast enough for even the fastest HD's....

 

http://www.ocztechnology.com/aboutocz/press/2010/369

 

Speed claims of up to 1.4GBit/Sec....that should be enough 

 

You are still limited to the read/write speed of the magnetic media you are using for the data.

[sosdk]

That is correct but with this card the limiting factor isnt the speed of the parity disk any more.

 

And if you got enough $ use cards as data disks.....cost would be massive but speed would be fast enough.

[WeeboTech]

That is correct but with this card the limiting factor isnt the speed of the parity disk any more.

 

And if you got enough $ use cards as data disks.....cost would be massive but speed would be fast enough.

 

There comes a point of diminishing returns. I do not think unRAID will achieve the speeds to warrant the cost of these cards.

I think BubbaQ tested unRAID with two SSD's and the throughput found was around 40MB/s.

 

If someone was all about speed, they would go with a raid0,raid5 or raid6 environment.

 

I've always seen unRAID about cost effective use of spindles.

 

 

 

 

[Joe L.]

That is correct but with this card the limiting factor isnt the speed of the parity disk any more.

It never was.  It was always limited by the slowest rotating drive involved.  If that was a data disk, then the data disk limited the speed to write to the array. If the slowest rotating disk was the parity disk, then it was the limiting factor.  The rotational speed of the slowest disk involved mattered.  The other could be instant, and the write speed would not change.

And if you got enough $ use cards as data disks.....cost would be massive but speed would be fast enough.

True...  expensive, but very very fast.  Then you would find other limiting factors in the SATA chipsets used, or the bus used as discovered in the link posted a few back in this thread.

[sosdk]

True you would problable run into other limiting factors if you build an array with these drives. And yes it would most likely to be much cheaper to use Raid 5 or 6.

 

I'm more than fine with the 30-40 mb/s write I get from my array and still have the security with 1 parity disk.

[NeilS]

I've just built an unRaid server using one of these boards and am very pleased with the results but with one exception: when performing a cold boot, the board spends around 60s "waiting for initialisation" before searching for any boot devices. This doesn't happen from a warm boot. I have set the fast boot option in the bios and USB as the first boot device. Once the init is complete the board detects the USB & HDDs fine and no further delays are experienced.

 

Configuration is Kingston flash drive, 2GB ram (kingston value as previously stated compatible in this thread) & 2 * WD 2TB EARS drives which have pins 7&8 jumpered & Seasonic 330 80plus PSU.

 

It runs virtually silently and power consumption is excellent at about 21w with the drives spun down & idle, 34w performing a parity check, and I think peak of about 50w on bootup. I have yet to try enabling the headless setting in the bios but that could result in lower still power draw.

 

Everything else is fine, just the delay on boot. Has anyone else experienced this?

 

Regards,

 

Neil

[purko]

when performing a cold boot, the board spends around 60s "waiting for initialisation" before searching for any boot devices.

It's probably looking for floppy.   

 

[Guest wolfdv]

 

 

Everything else is fine, just the delay on boot. Has anyone else experienced this?

 

 

 

I have not experienced this.. my X7SPA-H boots normally with no delays

 

maybe check in the bios, and disable ports/devices that arent being used, set drives to ahci, that sorta deal.