Important lessons learned about SAS3 drives and enclosures.

Falloutboy

Member
Oct 23, 2011
128
13
18
Okay today I believe have learned some important lessons about SAS3 drives.
I purchased 13 HGST 4TB SAS3 drives, I have a 9361-8i 1G cache and a 9361-8i 2G cache.
I am using Startech HSB4SATSASBA enclosures.

Now here is some of the fun that is to be had...
The 9361s detect the fact that these are SAS3 drives - cool I thought I was only expecting SAS2 but if I get a nibble of extra speed out of it then hey I'm happy.

This is where the fun begins and first I'll say what to do and then why.
Do NOT let your controller set the drive interface speeds as AUTO - set them to SAS2 : if you don't you may encounter the following issues:
1) Performing a chkdsk /f /r on a JBOD drives results in bad sectors being reported.
2) Performing the same chkdsk operation on the same drive shows - no bad sectors... WTF????
(2 Wastes 5 hours per check )

3) Random inexplicable sense errors when copying data to the drives.. OH Crud.
4) Drive for some reason seemed to completely vanish from the array - luckily it was only the hotswap spare.

Although I am going to physically check the connector on the inside of the enclosure tomorrow to confirm my suspicions I suspect that there is a physical difference between SAS2 and SAS3 connectors in the respect that there number of pins on the connector is larger for SAS3, I have a certified SAS3 cable and although I have to look at the connector in the HSB4SATAASBA I suspect that in the gappy piece between what you would call normal looking headers that there are 8 small pins which supply an extra data lane back to the controller per drive and this is the reason that all of the above happened.. Yet again this is to be confirmed but I will update this tomorrow once I have checked.

Since I set the physical interface on all drives to SAS2 - all the problems being reported in the logs - they are simply not happening now... Here is hoping it stays that way.
 

Aestr

Well-Known Member
Oct 22, 2014
903
299
63
Los Angeles
There shouldn't be a pinout change between SAS1/2/3. Many people here have run SAS3 drives at 12gb/s on directly connected SAS 1 and 2 backplanes such as the Supermicro A and TQ models. The extra pins on the drive your seeing are likely because it's a dual port drive although that's not unique to SAS3.

If anything if SAS2 is working for you I'd suspect that the cable or the enclosure is not consistently able to handle the signaling of 12gb/s. Since these look like spinners anyways it likely doesn't make a difference that you have to run at SAS2, but if you wanted to get back to SAS3 I'd try swapping out for high quality cables and backplanes.
 
Last edited:
  • Like
Reactions: T_Minus

cactus

Moderator
Jan 25, 2011
829
77
28
CA
I agree with @Aestr, Sounds like you have a marginal signal problem when going to the higher frequency used in SAS3. The number of pins did not change, but I personally haven't seen a SAS3 device that is not dual ported.
SAS-4.0-Drive-plug-connector-top-and-bottom-images.jpg
S14-S8 are the second port.
 

Falloutboy

Member
Oct 23, 2011
128
13
18
There shouldn't be a pinout change between SAS1/2/3. Many people here have run SAS3 drives at 12gb/s on directly connected SAS 1 and 2 backplanes such as the Supermicro A and TQ models. The extra pins on the drive your seeing are likely because it's a dual port drive although that's not unique to SAS3.

If anything if SAS2 is working for you I'd suspect that the cable or the enclosure is not consistently able to handle the signaling of 12gb/s. Since these look like spinners anyways it likely doesn't make a difference that you have to run at SAS2, but if you wanted to get back to SAS3 I'd try swapping out for high quality cables and backplanes.
I don't know if the drives are dual ported or not - I'd have to look into that I brought them for $180NZD each which for 4TB drives is dirty dirty cheap they are HGST HUS726040ALS210 AKA US7SSK400.

The cables are fan out type which connect from a 9361 to four SATA3 connections into the backplane of the enclosure and internally it has SAS connectors - you might be right about the cabling and a pure SAS enclosure may prove able to handle SAS3 at SAS3 speed - at the moment though I am happy with the fact things are up and running - hopefully reliably.

I did also purchase a cable which is to go from the 9361-8i to 4 SAS3 drives without any enclosure so I will compare the speed and reliability of that before making any final choices, in the mean time Ill check to see if the drives are dual ported. If they are and the dual porting is what it causing the issue is it possible to turn off the (dual - 2nd port) if that is what is causing the problem and return to SAS3?

Additional : Found in documentation.

SAS Dual Ported Operation
Single PHY ports (i.e. Narrow port. Wide Port NOT supported)

Ports function independently with separate firmware controls
Multiple DMA engines capable of accessing either port

Maximum outstanding credit of four per port...
 
Last edited:

Falloutboy

Member
Oct 23, 2011
128
13
18
There shouldn't be a pinout change between SAS1/2/3. Many people here have run SAS3 drives at 12gb/s on directly connected SAS 1 and 2 backplanes such as the Supermicro A and TQ models. The extra pins on the drive your seeing are likely because it's a dual port drive although that's not unique to SAS3.

If anything if SAS2 is working for you I'd suspect that the cable or the enclosure is not consistently able to handle the signaling of 12gb/s. Since these look like spinners anyways it likely doesn't make a difference that you have to run at SAS2, but if you wanted to get back to SAS3 I'd try swapping out for high quality cables and backplanes.
Would you have a link to what you consider to be high quality cables and backplanes?
 

T_Minus

Build. Break. Fix. Repeat
Feb 15, 2015
7,028
1,577
113
CA
High Quality Cables - Amphenol
High Quality Backplanes - SuperMicro, Intel and others of that caliber that sell to enterprise biz that run 10000s of them

Supermicro likely your best bet, and supermicro cables are very very likely amphenol too.
 
  • Like
Reactions: Stux