Recomended Sas HBA for supermicro sc846

[ktsiv]

Recomended Sas HBA

Dear all,

On a supermicro sc846, with the BPN-SAS2-846EL1 backplane, what controller will you recomend using
for as fast as possible (with the mentioned hardware), performance? Disks will be SAS2, 4 TB, 12 G.

 

[PhiloEpisteme]

This LSI-9211- 8i flashed to IT mode ought to do the trick. It supports SAS2 and the expander chip supports 8 lanes. You should be fine running cables from that HBA to pri_j0 and pri_j1 on that backplane and have 8 sas2 lanes. Here is a tread talking about this as well.

 

[ktsiv]

Thank you for your answer, the system has Adaptec ASR-71605 and LSI SAS9271 included in price. Any use for them?

 

[Ericloewe]

Nope, but they might be with the effort to sell on eBay.

 

[ktsiv]

Ok I might have found some Dell PERC H310 refurb in Europe, must now find how to flash them in IT mode...
Otherwise only solution is ebay and have to wait almost a month...

 

[Ericloewe]

The Resources section has something about the Dell cards, possibly in the large resource about flashing.

 

[PhiloEpisteme]

Ok I might have found some Dell PERC H310 refurb in Europe, must now find how to flash them in IT mode...
Otherwise only solution is ebay and have to wait almost a month...

I can't say why it would take a month, but I've purchased several things from the seller I linked above. They are very knowledgeable and responsive to messages. If you have a specific need/time crunch they might be able to help you out.

 

[ktsiv]

ok thanks Do we know what to expect performancewise? on the ethernet side?

 

[Ericloewe]

Depends on what's in the server...

 

[ktsiv]

24 x 4 tb SAS 7200rpm, 128 GB ram, 2 x8 core xeon @ 2ghz...

 

[Ericloewe]

I'd say at least a decent portion of 10 GbE, depending on the specifics.

 

[ktsiv]

Thank you, for you kind help.

One last thing is there any performance gain on the specific system by choosing better than 2008 chipset?
eg 2308?

 

[Ericloewe]

The 2308 is faster and supports more disks, but neither is a realistic concern.
The 3008 is SAS3, which makes sense if you'll be using expanders (assuming they're also SAS3) and SSDs. With spinning rust, it's a wash.

 

[TheNetStriker]

I've just bought an Supermicro SC846 and I installed an Fujitsu D2607-A21 SAS2 HBA (IT mode Firmware 20.00.07.00) controller. Everything works fine, but smartctl show's that my 8 Seagate ST8000NE001 SATA disks are only running with 1.5 Gb/s.

I only had one port connected to pri_j0 so I just tried connecting the second port of the controller to pri_j1 on the backplane, but that does not seem to make a difference in smartctl.

Is it normal that the disks only run with 1.5 Gb/s on this configuration or am I doing something wrong? I'm just curious if I'm loosing performance because something is connected or configured wrong.

 

[Ericloewe]

Is it normal that the disks only run with 1.5 Gb/s on this configuration or am I doing something wrong? I'm just curious if I'm loosing performance because something is connected or configured wrong.

What backplane do you have? Might this be a bad cable?

 

[TheNetStriker]

What backplane do you have? Might this be a bad cable?

It is a SAS-846EL1 Rev 1.1 backplane. I currently have two different cables connected. One that was delivered with the server (connected to pri_j0) and the one I just installed. (I took the second cable from an HP Proliant DL380 G7) The one that was delivered with the server looks a bit cheap. (it does not have a cable sleeve)

I also have a second cable from the Proliant server. I will try to replace the cable that came with the server also and see if this helps.

 

[TheNetStriker]

It is a SAS-846EL1 Rev 1.1 backplane. I currently have two different cables connected. One that was delivered with the server (connected to pri_j0) and the one I just installed. (I took the second cable from an HP Proliant DL380 G7) The one that was delivered with the server looks a bit cheap. (it does not have a cable sleeve)

I also have a second cable from the Proliant server. I will try to replace the cable that came with the server also and see if this helps.

I've just replaced the first cable, but the drives are still at 1.5 Gb/s. Is there something else I could do?

 

[Ericloewe]

So, you have a SAS1 expander backplane… It should support 3 Gb/s connections, but first-generation LSI hardware is not up to the same standard as later models. It may be just a simple bug that could have been fixed, but all the SAS1 and SAS2 stuff has been discontinued for a while.
You can try to update the expander firmware, maybe the latest one does have a fix.

If not, you might be able to upgrade the backplane with either a passive (-A model) backplane or a more recent expander backplane. Be sure to research which models will fit your specific chassis.

 

[TheNetStriker]

So, you have a SAS1 expander backplane… It should support 3 Gb/s connections, but first-generation LSI hardware is not up to the same standard as later models. It may be just a simple bug that could have been fixed, but all the SAS1 and SAS2 stuff has been discontinued for a while.
You can try to update the expander firmware, maybe the latest one does have a fix.

If not, you might be able to upgrade the backplane with either a passive (-A model) backplane or a more recent expander backplane. Be sure to research which models will fit your specific chassis.

Ok thanks for the hint. Sadly I now have a bigger problem. After I've replaced the first cable I switched back to the old cable. After that the pool was in degraded state and TrueNAS tried to resilver over night. During the resilvering I noticed that the drives where now running at 3 Gb/s, so maybe this value just increases when there is more activity in the Pool?

Today the pool broke completely. I noticed that the Raw_Read_Error_Rate of all my drives is now really high. (Over 180 Million on every of the 8 drives) The drives are all brand new Ironwolf Pro, so it is really unlikely that all of the drives just failed at the same time. My guess is that the old cable broke when I reconnected it and this caused the read errors.

I have a Backup of all the files, but now I'm unable to build a new Pool because of the bad smart values. I'm currently running a long smart test to see if this helps. Is there any way to recover from this situation?

 

[Ericloewe]

Are you sure that value isn’t always crazy? At least older Seagates encoded two separate values in there, in some funky encoding.
Besides, that’s not a value that depends on host communication, it should be strictly internal to the drive.