VMware Cloud Community
wreedMH
Hot Shot
Hot Shot
Jump to solution

How to identify disks with LSI HBA

Hello,

I have a drive that has failed but I cannot tell which slot it is in. I have the LSI 9207 HBA so the blink LED feature does not work.

How do you guys go about figuring this out?

1 Solution

Accepted Solutions
srodenburg
Expert
Expert
Jump to solution

You are not going to like my answer...

I use the exact same cards. All nodes have direct card-to-cable-to-single slot connections (no SAS-expander in between). Port 0 on each cable is always connected to slot 0 in the chassis. Port 1 goes to slot 1 and so on and so on.

So you would think that if you look in vCenter and see the vmhba0:C0:T3:L0 mapping that VMware creates from it, you would be able to deduct which disk is in which slot. Right?  Wrong !

This mapping works "most of the time" but it get's mixed up every now and then. You cannot rely on it.

I have spend a LOT of time on this topic and I never got a reliable, consistent mapping of chassis-slots to VMware's notation (C0:T2:L0 etc.) system.

The only way to find out which disk is where, is to install the SMIS and LSI/Avago MSM (as described above) and that is the ONLY place you should look. There, you can use the NAA.xxxx names that you see vCenter and look for that number (without the NAA part) in the MSM and it will tell you which controller-port it is connected to (0 to 7 on a LSI 9207). If you then mapped the controller ports on the cables consistently to backplane slots (like I did), you can reliably find the disk. Those NAA numbers in vCenter are actually the unique SAS addresses of a disks (much like a MAC address of a NIC).

Mind you, i'm not talking about backplanes with expanders where a "bundled cable" is used to connect both end. I'm talking about a 4-port SAS connector going to a fan-out cable where each "port" (they are called P0, P1 etc. in the Supermicro world) goes directly into a single-disk backplane SAS connector.

Things like "blinking" are mostly just sending IO's to a disk in a time-pattern (twice a second, or whatever the vendor choose) but that really does not help much if the disk is busy. When the disk is dead, well good luck sending IO's to that disk...

The only reason why on many machines, finding a disk through a visual indicator in a GUI works, is because they use a SAS-Expander which is hooked up via a "SCSI Enclosure Services (SES)" cable and specially written firmware for the whole combination. So messing around like this on Cisco UCS or a HPE ProLiant is not needed. But if you go the Supermicro route, or any other self-built system, then you have to "roll your own" in regards to disk identification.

This is the price you pay for using "DIY".

View solution in original post

Reply
0 Kudos
15 Replies
vijayrana968
Virtuoso
Virtuoso
Jump to solution

Are you able to see in hardware status tab of ESXI host in vSphere client or web client ?

Reply
0 Kudos
TheBobkin
Champion
Champion
Jump to solution

Hello,

Do these hosts have iDRAC an connection? You should be able to get this information from there.

See what information is available on hardware status as  vijayrana968 advised, steps here:

http://www.perthorn.com/vsan-operations-local-disk-identification/

Not sure which SSACLI or other utility works with Dell controllers pre-PERC9 (I doubt PERCCLI will work), try StorCLI - have a look on the Dell portal for these servers/controllers, potentially a utility is bundled with other tools as seen with HPE tools.

https://blog.hostonnet.com/megaraid

Hope this helps.

Bob

wreedMH
Hot Shot
Hot Shot
Jump to solution

Nothing is displayed under hardware status.

The servers do have iDRACs, but do not report any information on the storage since the servers have LSI 9207 HBAs and not a Dell branded controller. Smiley Sad

Any other ideas?

Reply
0 Kudos
vijayrana968
Virtuoso
Virtuoso
Jump to solution

May I know the controller make and model ?

Reply
0 Kudos
wreedMH
Hot Shot
Hot Shot
Jump to solution

LSI 9207-8i

Reply
0 Kudos
vijayrana968
Virtuoso
Virtuoso
Jump to solution

I don't see any CIM or SMIS provider module for this controller on Broadcom site, Once you install one of these module, you will see information on ESXI. These modules are responsible to report status of controller and disks in Hardware tab of ESXI.

Reply
0 Kudos
TheBobkin
Champion
Champion
Jump to solution

Hello wreedMH,

As I was saying - best bet is likely a vendor-specific storage utility (comparable to SSACLI/PERCCLI), from reading into it MSM seems to be the nearest thing available for LSI controllers.

Need SMIS and MegaRAID Storage Manager (MSM), here are the steps for configuring these:

https://serenity-networks.com/how-to-install-lsi-megaraid-storage-manager-msm-on-vmware-esxi-5-5/

https://www.vladan.fr/install-smis-provider-lsiavago-controller-card-vmware-esxi/

https://www.broadcom.com/products/storage/host-bus-adapters/sas-9207-8i#downloads

Bob

srodenburg
Expert
Expert
Jump to solution

You are not going to like my answer...

I use the exact same cards. All nodes have direct card-to-cable-to-single slot connections (no SAS-expander in between). Port 0 on each cable is always connected to slot 0 in the chassis. Port 1 goes to slot 1 and so on and so on.

So you would think that if you look in vCenter and see the vmhba0:C0:T3:L0 mapping that VMware creates from it, you would be able to deduct which disk is in which slot. Right?  Wrong !

This mapping works "most of the time" but it get's mixed up every now and then. You cannot rely on it.

I have spend a LOT of time on this topic and I never got a reliable, consistent mapping of chassis-slots to VMware's notation (C0:T2:L0 etc.) system.

The only way to find out which disk is where, is to install the SMIS and LSI/Avago MSM (as described above) and that is the ONLY place you should look. There, you can use the NAA.xxxx names that you see vCenter and look for that number (without the NAA part) in the MSM and it will tell you which controller-port it is connected to (0 to 7 on a LSI 9207). If you then mapped the controller ports on the cables consistently to backplane slots (like I did), you can reliably find the disk. Those NAA numbers in vCenter are actually the unique SAS addresses of a disks (much like a MAC address of a NIC).

Mind you, i'm not talking about backplanes with expanders where a "bundled cable" is used to connect both end. I'm talking about a 4-port SAS connector going to a fan-out cable where each "port" (they are called P0, P1 etc. in the Supermicro world) goes directly into a single-disk backplane SAS connector.

Things like "blinking" are mostly just sending IO's to a disk in a time-pattern (twice a second, or whatever the vendor choose) but that really does not help much if the disk is busy. When the disk is dead, well good luck sending IO's to that disk...

The only reason why on many machines, finding a disk through a visual indicator in a GUI works, is because they use a SAS-Expander which is hooked up via a "SCSI Enclosure Services (SES)" cable and specially written firmware for the whole combination. So messing around like this on Cisco UCS or a HPE ProLiant is not needed. But if you go the Supermicro route, or any other self-built system, then you have to "roll your own" in regards to disk identification.

This is the price you pay for using "DIY".

Reply
0 Kudos
wreedMH
Hot Shot
Hot Shot
Jump to solution

Yeah that is what I have figured out, of well, no big deal. How is the performance on the 9207 cards? I have been building a at home VSAN lab and I just think I cannot get them to perform. I dont know what it is. I have LSI 9300-8i cards on the way now to try.

See my other thread.  Poor All Flash vSAN Performance

Reply
0 Kudos
wreedMH
Hot Shot
Hot Shot
Jump to solution

srodenburg,

I got the SMIS provider installed and MSM connected to the host, but the NAA and SAS address numbers do not match at all Smiley Sad

Reply
0 Kudos
srodenburg
Expert
Expert
Jump to solution

Hi,

The performance of the LSI-9207-8i is excellent. I never have issues (but must use Firmware 19, NOT Firmware 20 despite FW20 being on the HCL now).

This card is the exact same card as the HP H220, a certain Dell card that i cannot remember the name of and a certain Fujitsu card. All OEM'ed 9207's.

But those manufacturers write their own Firmware for them so they work properly with their expander-backplanes etc.

Reply
0 Kudos
wreedMH
Hot Shot
Hot Shot
Jump to solution

srodenburg,

I got the SMIS provider installed and MSM connected to the host, but the NAA and SAS address numbers do not match at all

Reply
0 Kudos
srodenburg
Expert
Expert
Jump to solution

"but the NAA and SAS address numbers do not match at all"

Yes they do. The last character always changes but the entire ID after the "naa." up until the last character does match. In MSM, lookup the SAS address and subtract the "0x" and the last character. Then look at a ESXi Server -> Configure -> Storage Devices and the device ID and substract the "naa." and the last character.  You will see that the ID's match.

Reply
0 Kudos
wreedMH
Hot Shot
Hot Shot
Jump to solution

What am I missing? It must be because I am using a Dell Poweredge with a backplane. I see in your post above you use a fan out SAS cable. Grr my issue continues. This is on a R620, I am going to try a R720 in a minute, doubt its any different.

Capture.PNGCapture2.PNG

Reply
0 Kudos
DyJohnnY
Enthusiast
Enthusiast
Jump to solution

Why is it that the last number in the naa ID do not match?

IonutN
Reply
0 Kudos