Drive LEDs not working with HP DL380 drive cage and H221 (SAS2308) HBA

[fuzzy.art5728]

First-time poster here; been lurking and learning for a while. I'm building a DIY disk shelf using an HP DL380 G8 2.5" SFF drive cage and I can't seem to get the drive LEDs to work. I took inspiration from another post here (Converting an HP DL380e Gen8 14xLFF Server to a Disk Shelf). I was curious if someone might be able to point me in the right direction or simply confirm that the drive indicators aren't going to work with my configuration. This will ultimately become a TrueNAS array, so I understand that the management LEDs probably won't be functional without a hardware RAID controller--just looking for activity/power indicators here.

Like the LFF drive cage in the link above, the SFF version I'm using only has connectors for data and power (SFF-8087 and Molex Mini-Fit Jr.). The backplane is connected to the host computer with an HP H221 HBA. I chose this adapter because I suspected that the drive indicators are a proprietary HP thing and wanted to stick with HP firmware to keep the components simiar. What I've built here is essentially the same thing as if you installed an H220* in a DL380 G8 server and connected the hot-swap array to it. (*Note: I am using an H221, which is equivalent to the LSI 9205/9207-8e--external connectors. The H220 is the 8i variant.)

Here's what I've tried (drive LEDs don't work):

• Original HP P11 firmware on the card
• Flashed HP P15 firmware (the latest available here).
• Cross-flashed to H220 P15 firmware in case internal/external connectors matter
• Cross-flashed to H220i P15 firmware because, why not?

In researching this problem, I found a thread on the TrueNAS forums describing the opposite problem--drive LEDs that stop working with an H220 after installing generic LSI P20 firmware. This isn't surprising since the card is no longer running HP's firmware, but the thread does show that somebody at least had an H22x card with working indicators on a DL380 G8. However, the drive cage in this example is the 3.5" LFF version--same as the first link in my post--and I'm wondering if that backplane is somehow different than the one I'm using. Compared to the photo of my SFF backplane below, it definitely contains more components including some sort of controller under a heatsink. More on this below.

Other info that might be helpful:

• The physical and logical drive bays are swapped, so drive 1 on the HBA is slot 8 in the drive cage. This is apparently a common issue and I've read where this mismatch causes the wrong LEDs to light up. For testing, I have drives installed in slots 4 and 5 (logical 5 and 4, respectively) just to be sure I'm actually seeing something illuminate even if the ID numbers are wrong. No lights.
• Running locate using sas2ircu produces a "SEP write request failed" error. I suspect this is because the SFF backplane doesn't contain a storage enclosure processor (SEP); i.e., probably that controller chip in the LFF backplane. If I don't have a SEP, how the heck does the DL380 G8 operate the drive indicators in a "real" DL380? Does this maybe require a hardware RAID controller like the P420 on the server's motherboard?

Thanks for reading if you've gotten this far. I did a lot of testing, but maybe I'm missing something. Any input or suggestions would be really helpful. The drive indicators aren't a dealbreaker, but it'd be nice to have that functionality if it's a simple fix.

 

[gb00s]

Have you verified that the LED lights are not turned off for the drives? To activate them do sdparm --clear=RLM /dev/sdx.

 

[fuzzy.art5728]

Have you verified that the LED lights are not turned off for the drives? To activate them do sdparm --clear=RLM /dev/sdx.

No luck there. I receive a MODE SENSE command: Illegal request error. Maybe this has to do with SAS vs. SATA drives? The drives are Samsung enterprise SATA SSDs.

 

[gregsachs]

I wonder if your SFF-8087 doesn't have the sideband connection; since they look like only 4 bundles.

50cm Mini-SAS Cable SFF-8087 To SFF-8087 - SAS Cables | Cables | StarTech.com

StarTech.com 50cm Internal Mini-SAS Cable SFF-8087 To SFF-8087 & Sideband - Serial Attached SCSI (SAS) internal cable - with Sidebands - 4-Lane - 36 pin 4i Mini MultiLane - 36 pin 4i Mini MultiLane - 50 cm

www.startech.com

https://tools.molex.com/pdm_docs/sd/795762112_sd.pdf

 

[fuzzy.art5728]

I wonder if your SFF-8087 doesn't have the sideband connection; since they look like only 4 bundles.

Oh, you might be onto something there. The H221 HBA has SFF-8088 external Mini SAS outputs. I'm using SFF-8088 to SFF-8087 cables. I don't think the external connector spec has pins for the sideband. Would the sideband be used to transmit the drive activity data from the HBA to the backplane?

I was actually looking at a pinout chart earlier and was wondering how the activity indicators could possibly work with only RX and TX lines on the Mini SAS connector. I chalked it up to the HP firmware doing something with the procotol. I'll feel kinda silly if this isn't working because there aren't enough wires in the cable...

 

[gregsachs]

This is just a guess, but i'm sure that the 12x3.5" backplane has an expander, under the headstink, so would not necessarily need sideband. The 8x 2.5" backplane looks dumb, so probably needs sideband to get the leds to work.

 

[fuzzy.art5728]

This is just a guess, but i'm sure that the 12x3.5" backplane has an expander, under the headstink, so would not necessarily need sideband. The 8x 2.5" backplane looks dumb, so probably needs sideband to get the leds to work.

That seems plausible. The folks in that first thread who have adapted the LFF drive cage all mostly seem to be using SFF-8087 internal cables, though. It would be interesting to hear if anyone has tried the LFF cage with SFF-8088 cables and still has working drive indicators.

The DL380 G8 SFF backplane PCB is indeed "dumb" and the onboard ICs are mostly just local regulation and switching with a few EEPROMs. I took apart a spare G8 drive caddy and discovered that there's actually a PIC microcontroller on the flat flex PCB near the LEDs (picture below). So, yeah, I suspect that the sideband is used to interface directly with each drive caddy.

Just to compare, I briefly considered picking up a DL380 G7 drive cage because the older G7 uses the basic HP caddies with only power and activity LEDs via light pipes. The PCB, though, suggests the same situation as with the G8. There are PIC microcontrollers on the backplane (and even what appears to be programming/debugging headers marked "PIC A" and "PIC B"!). I bet this backplane would also not work without the sideband signals even though the LED indication is far simpler.

It's too bad none of this is conclusive. I'm not sure if I'll buy a cheap H220 or H222 with SFF-8087 internal connectors just to confirm the sideband hypothesis. At the end of the day, the drive LEDs aren't really that important, and probably the only way to get them to work is to run an '8087 cable from the inside of my host computer, out the back, and to the drive cage. That's kinda hokey, especially since that type of cable doesn't have extra shielding and isn't designed for use outside of an enclosure/case.