Reading SMART data when your HBA hides it

Guys running Contabo or OVHcloud dedis with LSI HBAs probably hit this. Sas2ircu and sas3ircu do not expose SMART through standard tools. Here's the workaround I pieced together after too many recieved tickets about "hidden" drive health.

First, sas2ircu/sas3ircu can dump PHY error counts:

sas2ircu 0 DISPLAY

For actual SMART, you need to bypass the HBA entirely. Two paths:

1. Direct SATA — pull the drive, slap it on a cheap SATA controller or USB dock. Obvious but annoying in colo.

2. sg_utils + pass-through — some HBAs allow SAT passthrough:

sg_turs /dev/sg0
sg_readcap /dev/sg0
smartctl -d sat+megaraid,0 -a /dev/sda

3. sas3ircu + binary parse — the DISPLAY output has raw pages. Pipe to hexdump, offset 0x2B0 usually has temperature if the backplane reports it.

I don't have their exact firmware matrix, so any leads on which OVHcloud chassis still block passthrough? Running a panel that alerts on PHY errors only, and clients hate false positives

Actually the sas2ircu Utility has a hidden Option for Enclosure status which gives you the Drive temperature directly from the backplane Expander or? I found this in the LSI Integration manual from 2014 which actually states that the SGPIODataBlock contains thermalInformation but only if the expanderFirmware supports the SFF8485Specification which actually most chineseBackplaneClones do not implement correctly because they reused the pinout for activityLEDs instead so you get blinkingLights instead of temperatureValues and then you are actually back to squareOne or?

Compoundwordsfusedtogether are actually the germanWay to deal with technicalDocumentation that never ends.

1. @rudi3 has a point about expander support
2. But my experience with this:
2. A) Contabo SM2012 chassis = works
2. B) Hetzner generic 4U = no thermal data
2. C) Leaseweb rebadge = sometimes, firmware-dependent
3. For MegaCLI alternative:
3. I) StorCLI is the modern path
3. Ii) /c0/eall/sall show all | grep -I temp
3. Iii) still not SMART, but better than nothing
4. Sg_utils path needs /dev/sg* mapping
4. A) lsscsi -g
4. B) match enclosure slot to sg device
5. Honestly for production monitoring:
5. I) PHY errors trend worse before drives die
5. II) temperature spikes correlate with 60% of my failures
5. III) combine both, stop chasing perfect SMART passthrough

Please, server does not want to be read by sas3ircu, just listen for the click of death

15 year success rate, no crying, no smart, no problem

When drive goes click-click, replace with spare from parent, hashcat approves

Hana5's method works until it doesn't. I've got 40U at RackNerd, mix of Gen 8 through Gen 11 Dell, and I stopped trusting ears after a Seagate ST8000 made zero noise then reported 0x0C power-on hours reset to zero — firmware bug, not mechanical failure.

For that chassis density, I run StorCLI on a cron, scrape /c0/eall/sall show all into rrd. Power draw on the PERC H730 is 9W vs 14W on the HBA330. Not nothing at scale.

The real fix: avoid IR firmware entirely. IT mode, passthrough, smartctl -a /dev/sdX works native. Flashing takes 20 min, saves hours later. I've pallet-shipped twelve R730xd to GreenCloudVPS colo, all IT mode before they left my bench.

RackNerd's 9W figure seems low for a PERC under load. Measured at wall?

Which expander firmware on the Leaseweb? LSI or rebadge

3 days quiet but this is still useful.

Did you ever solve this? I have the same chassis and StorCLI only shows "N/A" for temp across all slots. Wondering if it is worth flashing the expander or if the backplane itself lacks the sensor wiring.

Also re: sg_utils path. Did you need to load mpt2sas_debug or does the mapping work out of the box on newer kernels?

11 days but this is still useful — did anyone actually get sg_ses or sg_sense working through the expander on the Hetzner units

I tried mapping the sg devices per liam_funky's step 4 and the enclosure services page just returns "unsupported" on two different backplanes so I am curious if that is the same clone firmware rudi3 mentioned or if I missed a driver flag

Also

Late to this but

The ones that work have the LSI firmware string in sas2ircu list, the ones that don't show a blank vendor field or some four-letter code

I never found a way to flash them to stock without bricking the expander so I just gave up and used the PHY error method RAJ mentioned

Did anyone ever solve the temperature part on the no-name backplanes or is that still a dead end