Why is my SAS hard drive not detected by the HBA?

1. 1.Read what the drivers logged at boot. The mpt3sas driver logs each LSI/Broadcom controller it finds, with its firmware version and capabilities, and the sd driver logs every disk it attaches.

   dmesg | grep -i -E 'mpt3sas|mpt2sas|sd '

   Readingmpt3sas lines with 'FWVersion(..)' and 'Protocol=(Initiator,Target), Capabilities=(...)' mean the driver found the controller. 'Raid' inside Capabilities appears only when the firmware reports Integrated RAID capability, which tells you the card runs IR rather than IT firmware. For each disk the sd driver logs '... N-byte logical blocks: (...)' and 'Attached SCSI disk'; no such line for the new drive means the OS never got a disk from the controller.
2. 2.List the SCSI hosts. This answers whether the card itself is in play before you look for the drive.

   lsscsi -H

   ReadingThe HBA or RAID card should appear as a host with driver mpt3sas or megaraid_sas. If it is not listed, the problem is the card, the slot or the driver, not the drive.
3. 3.List every device with its transport and generic name.

   lsscsi -g -t

   ReadingOne line per device with its /dev/sdX and /dev/sgN names; a SAS drive shows 'sas:' and its SAS (naa-5) address. If the SAS drive is missing while other disks on the same HBA appear, the problem is that drive's slot, cable or power, or the drive.
4. 4.Match physical drives to block devices by serial number, so you know which one is missing.

   lsblk -o NAME,HCTL,TRAN,SIZE,MODEL,SERIAL

   ReadingTRAN is the transport ('sas', 'sata' and so on); HCTL is Host:Channel:Target:Lun. Compare SERIAL with the label on the drive.

1. 1.On an LSI/Broadcom SAS3 HBA (SAS3008 and the other chips the SAS3IRCU guide lists), list the controllers, then the devices on the first one. sas2ircu is the SAS2 counterpart; in the TrueNAS thread below it was run on a 9207.

   sas3ircu list && sas3ircu 0 display

   Readingdisplay lists each physical device the HBA sees with Enclosure #, Slot #, SAS Address, State, Model Number, Serial No, Protocol (SAS or SATA) and Drive Type (SAS_HDD, SATA_HDD, SAS_SSD, SATA_SSD). State 'Ready (RDY)' means usable as a normal disk, 'Available (AVL)' might not be ready, 'Missing (MIS)' is removed or not responding, 'Failed (FLD)' has failed. A drive absent from this list is not visible to the controller at all, which points at power, cable, backplane or the drive rather than the operating system: go to p3 and p4. In one TrueNAS thread, sas2ircu list on a card passed through to a VM returned 'SAS2IRCU: MPTLib2 Error 1' and no fix was confirmed.
2. 2.On a MegaRAID controller, use StorCLI instead. Start with the controller summary.

   storcli /c0 show

   ReadingSummarises the controller with its drive groups, virtual drives, physical drives and enclosures. It confirms the controller is up and which drives it lists.
3. 3.List every enclosure, slot and drive the MegaRAID controller sees.

   storcli /c0/eall/sall show

   ReadingA drive listed here but missing from lsblk is a controller configuration matter, not a cable fault: see the JBOD step below. Drives attached without an enclosure are addressed by controller and slot only, as /c0/sN.
4. 4.Check the PHYs on the MegaRAID controller when the drive is not listed at all.

   storcli /c0/pall show

   ReadingShows basic PHY layer information for each controller phy. A phy with no attached device on the port the drive is cabled to points at the cable, the backplane or power.
5. 5.On a Dell PowerEdge, the iDRAC can list the physical disks from its own command line over SSH.

   racadm storage get pdisks

   ReadingLists the physical disks the iDRAC reports for the server's storage controllers. A drive missing here is not seen by the controller. The Dell KB we read shows the output only as an image, so we do not interpret its fields here.
6. 6.If the controller lists the drive and the OS does not, the card is presenting arrays, not disks. Seagate notes that many SAS RAID controllers require a drive to be assigned to an array. The choices are a virtual drive, JBOD on controllers that support it, or an HBA or IT-mode card (see IT mode explained below).
7. 7.This writes the controller's configuration and can cost you data. Run it only on a drive that holds nothing you need and belongs to no existing array; we have not found a Broadcom statement of what happens to data on a drive switched to JBOD, so treat it as data-destroying. It sets one drive to JBOD so the OS can see it.
   Destructive. This command destroys what is on the device and cannot be undone. Confirm the device path twice and have a backup of anything you still need before running it.

   storcli /c0/e<E>/s<S> set jbod

   ReadingStorCLI's reference says JBOD is 'Not supported by all controllers'. Some controllers need JBOD enabled controller-wide first with storcli /c0 set jbod=on, which also writes configuration. If no enclosure is used, address the drive as /c0/sN.

1. 1.Check what the drive is plugged into. Seagate states SATA drives may be plugged into SAS controllers, but SAS drives cannot be plugged into SATA controllers, so motherboard SATA ports and SATA-only cards will not run a SAS drive. The SAS drive connector, SFF-8482, carries two 7-contact signal sets and a 15-contact power set, and a SAS receptacle accepts both a SATA-style plug and a full SAS plug, which is why the fit runs one way.
2. 2.Power the system off before you touch a cable. The SFF-8482 specification states 'Hot plugging of cables is not supported by this interface.' Reseat both ends; in one serverbuilds.net thread the advice for drives lost after a cable swap was to check the SFF-8087 orientation and seating.
3. 3.Check the cable type for a drive cabled straight to the card. The controller end is the multi-lane connector (SFF-8643 on Broadcom's 93xx and 94xx cards; Supermicro's backplane guide connects a SAS2 HBA with SFF-8087) and the drive end must be SFF-8482. Broadcom's cable guide lists CBL-SFF8643-SAS8482SB-06M for this, next to SFF-8643-to-SATA and SFF-8643-to-SFF-8087 backplane cables.
4. 4.Check the breakout direction when a breakout feeds a backplane. Broadcom labels each cable end 'To Controller' and 'To Drives/Backplane'. Supermicro's answer to a user who saw no drives going from four discrete controller ports into an SFF-8087 backplane connector was that the cable must be a cross-over type (its part CBL-0116L). Sellers call the two kinds forward and reverse breakouts; we have not found a primary document that defines both terms, and posters in one Unraid thread say the two look the same.
5. 5.Check that the bay has a host lane. On a direct-attach backplane each bay has its own lane: Supermicro's BPN-SAS3-826A serves twelve bays from three Mini SAS HD connectors, four bays each, so a bay whose connector is not cabled to a host port has no link. An expander backplane such as the BPN-SAS3-826EL1 reaches all drives from fewer host connections. Move the drive to a bay or cable where another drive is detected; if it appears there, the original bay, lane or cable is the fault.

1. 1.Check whether the drive spins up at all. Western Digital's power-disable brief gives 'The HDD will not spin up / power up' as the symptom of a drive held in reset by pin P3.
2. 2.Know which drives this applies to. WD's brief says SAS-3 redefined P3 from 3.3V power to POWER DISABLE, SATA added the feature in revision 3.3 (published 2 February 2016), and a power source that holds 3.3V on P3 keeps a drive with the feature in hard reset. The shucking guide linked below covers the pin itself.
3. 3.Treat SAS-specific reports as forum evidence. WD says SAS drives are not affected in older SAS chassis, which leave P3 unconnected. A PSU's SATA power plug is not a SAS chassis, though: in an Unraid thread and a serverbuilds.net thread, SAS drives on SFF-8482 breakouts powered from the PSU's SATA plugs stayed invisible until 3.3V was taken off P3, and a reply in the second thread said such cables pass 3.3V through from the PSU.
4. 4.Apply WD's workaround: power the drive through a Molex-to-SATA adapter, which removes power from P3. WD says plugging a drive with or without the feature into either kind of system does not damage it. The Unraid thread's drives appeared once powered this way; other posters taped the pin or cut the 3.3V wire, which are forum reports, not documented fixes.

1. 1.Read the drive's identity, health and logs. Use the /dev/sgN name from lsscsi -g; -d scsi stops smartctl sending ATA commands to a SCSI device. Behind a MegaRAID controller use smartctl -a -d megaraid,N /dev/sda (or /dev/bus/0), where N (0 to 127) is which disk on the controller to read (smartctl(8)); the man page says this interface also works for Dell PERC controllers.

   smartctl -a -d scsi /dev/sgN

   ReadingFor SCSI devices -a equals -H -i -A -l error -l selftest: identity, health, and the error and self-test logs.
2. 2.Check the rate each SAS link negotiated. Leave ',reset' off the option while diagnosing, because it clears the counters.

   smartctl -l sasphy /dev/sgN

   Reading'negotiated logical link rate: phy enabled; 6 Gbps' (or 12, 3 or 1.5 Gbps) means the link is up at that rate. 'phy disabled', 'phy enabled; speed negotiation failed' or 'phy enabled; unsupported phy attached' point at the link: cable, backplane or port. A dual-port SAS drive reports two ports. Broadcom's 9300-16i (SAS3008) guide lists SAS link rates of 3, 6 and 12Gb/s, so a 6Gb/s SAS drive on that card is a supported combination.
3. 3.Confirm the drive answers commands and note its model and firmware.

   sg_inq /dev/sgN

   ReadingPrints vendor, product and firmware revision from the standard INQUIRY data. sg_inq -i /dev/sgN prints the device identification VPD page (0x83).
4. 4.If one drive is listed twice, check the backplane. A dual-port SAS drive on a dual-expander backplane such as Supermicro's BPN-SAS3-826EL2 is reached through two expanders, and SAS3IRCU shows a dual-port SAS drive as a duplicate device. With two HBAs, Supermicro says MPIO software is required for failover.

Broadcom's 9300-16i guide lists SAS link rates of 3, 6 and 12Gb/s for that SAS3008 card, so a 6Gb/s SAS drive on it is a supported combination. We have not found a Broadcom document stating a 6Gb/s card's support for 12Gb/s drives. In one serverbuilds.net thread, 12Gb drives were invisible where 6Gb drives worked, and the fix there was taking 3.3V off the power pin, not anything to do with link rate. smartctl -l sasphy shows the rate a link actually negotiated.

Two documented explanations fit. The port may be SATA-only: Seagate states SATA drives may be plugged into SAS controllers but SAS drives cannot be plugged into SATA controllers. Or the SAS drive may have the power-disable feature and be held in reset by 3.3V on P3, which WD's brief describes for drives with that feature. In two threads (Level1Techs, TrueNAS) a SATA test drive on the same cabling appeared while the SAS drives did not; in a third (Unraid) a SATA drive on other cables did.

We have not found an SFF or vendor document that says the connector key blocks it. The SFF-8482 text we read describes the other direction: a SAS receptacle accepts both a plug with one signal set plus power and a plug with both signal sets. Seagate's statement is that a SAS drive on a SATA controller will not work.