SAS cabling and power

[rslocalhost]

As recommended in another thread, I bought an LSI SAS9211-8I card to burn-in test my SAS drives in a non-production machine. This card has two internal x4 SFF8087 Mini-SAS connectors. So I looked around for a cable (it doesn't come with any) that converts the SFF8087 to a SAS connector (29 pin). I found this cable on amazon.

After flashing the card to JBOD/IT mode, the cable arrived. I plugged in the 2 SAS drives I wanted to test, and they weren't found. A quick bit of investigating found they weren't spinning (platter drives, not SSD). So no power. I checked the connections, trying unplugging and replugging, even tried the other unused arms. Still no power. For kicks, since I can plug a SATA drive into a SAS connector, I tried that. The SATA drive spun.

I looked at the drives, and the SAS drives do need slight more amps then the SATA. So I figured since the system I was using was a Dell desktop, and Dell power supplies aren't the greatest IMHO, I moved to one machine that had a $100 power supply (not Dell, Antec). Same results.

So, is the cable bad or is there something else I'm missing? Cable looks in great shape, so don't see how that could be bad, but short of another answer...

 

[Chris Moore]

What kind of drives?
You didn't specify.

Sent from my SAMSUNG-SGH-I537 using Tapatalk

 

[rslocalhost]

The SAS drives are 6TB 12GB/s HGST HUS726060AL5210 and HUS72606AL4214. (Yes, I understand the card is only 6GB/s, but this is for burn in testing, not production use.)

 

[Chris Moore]

The SAS drives are 6TB 12GB/s HGST HUS726060AL5210 and HUS72606AL4214. (Yes, I understand the card is only 6GB/s, but this is for burn in testing, not production use.)

I believe those drives have a new non-standard way of powering them completely off, using some of the interface connector pins. It's some new Hitachi thing.
You have to tape pin 3 of the SATA power connection to get them to spin.

Sent from my SAMSUNG-SGH-I537 using Tapatalk

 

[Chris Moore]

Let me know if it works for you.

Sent from my SAMSUNG-SGH-I537 using Tapatalk

 

[rslocalhost]

TL,DR: That was the problem.

According to a pdf by HGST, the power on disable isn't a problem for SAS:

Q: If I am using SAS HDDs, should I be concerned?
A: No. This feature is available for use on newer SAS chassis. Older SAS
chassis will ignore this feature. Specifically, SAS HDDs never defined an
alternate usage to P3 (Pin 3), and therefore legacy systems will have it
as a “no-connect”...

but then they go to say later:

Q: When was this feature introduced on SAS HDDs?
A: With the introduction of 12G SAS, a new SAS standard, SAS-3,
redefined P3 (Pin 3) from “3.3V Power” to “POWER DISABLE”, i.e.
“Reset”...

HUH!!! Which is it?!?!

It turns out to be the second answer is correct and the first is wrong. I found a really old SAS drive in the server room (320 GB). When I used a SATA plug from my power supply, it spun up. The new drive did not. Then I moved to a molex plug and used a molex to SATA adapter, both drives spun up. BTW, they did have to be plugged into the controller too. Just power alone didn't work for SAS (but did for SATA). Probably a SAS "feature. We did the all the tests with the controller.

So thank you Chris. I doubt I would have figured this out on my own.

 

[nightshade00013]

Yep, this is something I posted about not long ago.

https://forums.freenas.org/index.php?threads/pwdis-feature-in-new-sata-drives-to-watch-for.60174/

More and more drives will start to come with it. What I am wondering is the older backplanes that predate the spec and likely provide 3.3v all the time. Would be nice it they just provided a jumper to turn the feature on or off.

 

[Chris Moore]

Yep, this is something I posted about not long ago.

https://forums.freenas.org/index.php?threads/pwdis-feature-in-new-sata-drives-to-watch-for.60174/

More and more drives will start to come with it. What I am wondering is the older backplanes that predate the spec and likely provide 3.3v all the time. Would be nice it they just provided a jumper to turn the feature on or off.

I can't speak for all hardware by any means, but the Supermicro and Chenbro SAS backplanes that I have used only have 5v and 12v going into them to begin with, so if they are supplying 3.3v to the drive I would be very surprised. There are some smaller enclosures that have SATA power inputs that may pass that 3.3v line on to the drive. I can see this being an issue that crops up time to time going forward because many people will be cabling their drives up in smaller installations where there is no backplane.

 

[nightshade00013]

I think the chenbro only has 12V going to it. But by any means stepping voltage down is VERY easy to do. PSU's supply 3.3 5 and 12 volts but we use voltages well below that regularly on motherboards for cpu's and ram. I wouldn't make the assumption that the backplanes can not do the same thing. If the spec's at the point of purchase specified 3.3v to be supplied I would err on the side of caution and figure it is being supplied until it's proven it's not.

 

[Chris Moore]

I think the chenbro only has 12V going to it. But by any means stepping voltage down is VERY easy to do. PSU's supply 3.3 5 and 12 volts but we use voltages well below that regularly on motherboards for cpu's and ram. I wouldn't make the assumption that the backplanes can not do the same thing. If the spec's at the point of purchase specified 3.3v to be supplied I would err on the side of caution and figure it is being supplied until it's proven it's not.

Some have 5v. I have one that is 12v only.

 

[nightshade00013]

Right. OK, I figured the only chenbro was the same as the one I ended up with which is 12V only

 

[Stux]

AFAIK no 3.5” disks use 3.3V power. I think you could just cut the orange cable.

After all molex peripheral connectors don’t have the orange cable anyway.

 

[Chris Moore]

AFAIK no 3.5” disks use 3.3V power. I think you could just cut the orange cable.

After all molex peripheral connectors don’t have the orange cable anyway.

If I recall correctly, the 3.3v lead was introduced in the power supply an effort to get drive manufacturers to support lower power states. The problem was, because the 4-pin Molex connectors do not provide 3.3 V power, Molex to SATA adapters provide only 5 V and 12 V power and leave the 3.3 V lines unconnected. This precludes the use of such adapters with drives that require 3.3 V power. Understanding this, drive manufacturers have largely left the 3.3 V power lines unused. The idea that HGST is working with and I think this is being introduced as a new "Standard"... They are implementing the ability to reset a drive by hitting the 3.3 V power lead instead of having to actually walk around to the server and pull the drive out and put it back in. This would require hardware support, that doesn't exist in older hardware.

 

[Chris Moore]

I was doing some reading today and came across this document on the HP website that discusses SAS cabling for dual domain / dual path and how expander architecture is better than direct wired, etc. I thought might be beneficial or interesting:
https://support.hpe.com/hpsc/doc/public/display?docId=c01451157

 

[nightshade00013]

Very interesting. I wonder if it's only HP expander backplanes that support the dual path connection to SAS controllers. I was actually wondering if it was possible to do something like this with the Chenbro case since each backplane has the ability to do an output as well as an input and can chain to either drives directly or expanders. Otherwise maybe it would be possible to add a second SAS controller to a system to interface with each backplane. Doesn't really mean a whole lot for a home system for high availability but allowing a backplane to use the fastest path is always a plus.