SAS3 & Sata3 Drives Mixed on a SAS2 Backplane (BPN-SAS2-836EL1)

[Oda Nobunaga]

I have a SuperMicro 3U/16 BAY chassis with SAS2 expander backplane (BPN-SAS2-836EL1), and LSI 9211-8i HBA which I plan to flash into IT mode. I also have an existing NAS that runs on a SuperMicro X11 Board and Xeon E3-1230 v5, with a series of Sata drives as storage. (4 x 4TB WD Red & 2 x 4TB WD Gold).

I am also currently seeing 6TB Seagate Exos 7E8 Enterprise SAS 12.0Gb/s drives for around $150 a piece, new, which is right in my budget.

My plan is transfer the main guts of my existing NAS (which more or less acts as a media-storage-box for the local network) into the chassis, reconfigured to host all the drives as JBOD on the backplane through the HBA, with additional SAS2/SAS3 (whichever is available cheaper at sufficient quality/reliability) drives added as SnapRaid parity drives. I then intend to build a second "primary" server using entirely Exos SAS3 drives which will format the drives using ZFS, and once that is up the unit made from my existing NAS will serve as a backup/overflow server.

My thinking was that for the backup server: Sata drives are cheaper to replace in bulk when they inevitably fail, while the SAS drives ought to be more resilient to silent corruption and failures as a result of total throughput, making them more likely to survive full recovery to another drive in the event of a failure. I had planned on loading four Sata drives to every SAS parity drive, such that it would eventually be twelve 4TB Sata storage drives, three 6TB SAS parity drives, and a lone "risky" drive without parity, neatly using up the 16 bays on the chassis. (A PCIE-NVME and 2.5" SSDs at the rear of the chassis would offer fast file transfer.)

But my research online now suggests that there are potential issues with mixing Sata and SAS drives on a SAS2 Expander backplane?

Reports seem contradictory; some say it should be fine past SAS1, others that it risky even on a SAS3 backplane, others that it is problematic only under certain circumstances, most of which are easy enough to mitigate...

So, what's the deal? Can anyone offer me some clarification as to the general state of mixing SAS and Sata on a single expander backplane? Does anyone know if my particular backplane offers issues in this regard? Should I instead keep one system straight Sata, the other straight SAS? I do not fancy replacing all existing drives at once..

 

[BLinux]

hi Oda Nobunaga, Akechi Mitsuhide here...

I can't remember exactly which Seagate model HDD it was, but I discovered there are SAS3 Seagate HDDs that are very buggy; specifically, they WILL NOT negotiate down to SAS-2 or earlier protocols and will ONLY work with SAS-3 equipment. These were 4TB and 8TB SAS-3 HDDs from Seagate. So, before you buy any Seagate SAS-3 HDDs, do your research on this issue first. I was helping someone with this problem for several days until I discovered that Seagate's SAS-3 implementation doesn't negotiate well with SAS-2 equipment in a direct attached scenario - so I suspect it would be even worse in a SAS-2 expander setup.

I don't have specific information on mixing SATA and SAS. However, I do know (and there's a thread i started around here somewhere), that the Supermicro backplane's behavior is different for SATA and SAS hard drives. The way the activity LEDs work is different, and that might be cause for concern, because you might think a HDD bay shows something is faulty, when it is not as the SATA and SAS HDDs will look different. I've also observed differences in LED behavior due to differing firmware, especially if it has a OEM firmware from Sun/Oracle, or HP. It is cosmetic, but may affect operations.

 

[i386]

Using sata & sas on the same expander? (Should be) no Problem (see @BLinux post about seagate sas hdds).
Using sata & sas in the same logical volume? Maybe.

(Nearline-)SAS (and enterprise sata) use the fail fast approach while consumer/prosumer drives try to handle a failure before reporting it to the host.

 

[Oda Nobunaga]

hi Oda Nobunaga, Akechi Mitsuhide here...

WHAAAAA? YOU DARE SHOW YOUR HEAD AGAIN AFTER SUCH A BETRAYAL AS YOUR REBELLION?

I can't remember exactly which Seagate model HDD it was, but I discovered there are SAS3 Seagate HDDs that are very buggy; specifically, they WILL NOT negotiate down to SAS-2 or earlier protocols and will ONLY work with SAS-3 equipment. These were 4TB and 8TB SAS-3 HDDs from Seagate. So, before you buy any Seagate SAS-3 HDDs, do your research on this issue first. I was helping someone with this problem for several days until I discovered that Seagate's SAS-3 implementation doesn't negotiate well with SAS-2 equipment in a direct attached scenario - so I suspect it would be even worse in a SAS-2 expander setup.

This is good to know! I'm glad I came across it before sinking money into drives! I'll definitely have to check out compatibility further before any purchases.

I don't have specific information on mixing SATA and SAS. However, I do know (and there's a thread i started around here somewhere), that the Supermicro backplane's behavior is different for SATA and SAS hard drives. The way the activity LEDs work is different, and that might be cause for concern, because you might think a HDD bay shows something is faulty, when it is not as the SATA and SAS HDDs will look different. I've also observed differences in LED behavior due to differing firmware, especially if it has a OEM firmware from Sun/Oracle, or HP. It is cosmetic, but may affect operations.

Yes, this I had come across. It's good to know for administering such a system, but doesn't seem so much a cause for concern as long as you are aware of it. (And aren't in a massive scale production environment where different employees will check each system and thus might not know.)

 

[Oda Nobunaga]

Using sata & sas on the same expander? (Should be) no Problem (see @BLinux post about seagate sas hdds).
Using sata & sas in the same logical volume? Maybe.

(Nearline-)SAS (and enterprise sata) use the fail fast approach while consumer/prosumer drives try to handle a failure before reporting it to the host.

Yes, I could see that being problematic.