LSI SAS HBAs (mps) utility

[Borja Marcos]

On Mar 11, 2014, at 5:59 PM, Alan Somers wrote:

>> And now, same hardware, different firmware, I don't see the "ses0" expander, odd.
> 
> Surprising.  What does "camcontrol devlist" show you?

It shows this:
root at pruebassd:~ # camcontrol devlist
<ATA OCZ-VERTEX4 1.5>              at scbus0 target 0 lun 0 (pass0,da0)
<ATA OCZ-VERTEX4 1.5>              at scbus0 target 5 lun 0 (da1,pass1)
<ATA OCZ-VERTEX4 1.5>              at scbus0 target 7 lun 0 (pass2,da2)
<ATA OCZ-VERTEX4 1.5>              at scbus0 target 9 lun 0 (pass3,da3)
<SEAGATE ST9146853SS YS08>         at scbus0 target 12 lun 0 (pass4,da4)
<SEAGATE ST9146853SS YS08>         at scbus0 target 13 lun 0 (pass5,da5)
<ATA OCZ-VERTEX4 1.5>              at scbus0 target 15 lun 0 (pass6,da6)
<ATA OCZ-VERTEX4 1.5>              at scbus0 target 16 lun 0 (pass7,da7)
<Dell Internal Dual SD 1.:>        at scbus2 target 0 lun 0 (da8,pass8)

The ses0 device is gone. Full story follows below the "I can't understand your English". ;)

>> Anyway, I pulled one of the disks (physical slot 4) and it wasn't the disk identified by sas2ircu as connected to "1:4" but
>> a different one.
> 
> What do you mean by "physical slot 4"?  The HBA doesn't know anything
> about physical positions.  Even the SES expander probably isn't
> numbering the slots in the order that you expect.

Yes, that´s what I mean. I was hoping to be able to determine the slot for a given disk, but it's clear that I am wrong.

>> 
>> That I mean. Chaos... ;)
> 
> I'm having trouble understanding your English.  What does this sentence mean?

Sorry, I wasn't explicit at all. Full story follows.

I'm building a storage server and I'll probably build several ones. It's based on ZFS of course, and I'm using SSDs for now.

Due to the insistence of several manufacturers such as Dell or IBM, I had several "RAID" cards, which, of course, we don't want
in such a configuration. Turns out some of them (the ones sold by Dell as H200 and the cards sold by IBM as M1015) can be
turned into simple HBAs by replacing their firmware by what LSI Logic calls "IT Mode" firmware. 

I did it using the latest firmware version I found on the LSI Logic website. The cards work, the performance is very good, and I
don't need to battle "JBOD" pseudo disks and such. The SSDs receive the TRIM commands perfectly, etc. 

So far so good, but of course I need to make sure that an operator can identify a failed disk without confusion. And that's where the fun
is. Even being substantially cleaner with the "IT mode" firmware, these cards are stilll a puzzle to use. 

What I am looking for is a way to identify a physical disk. I guess I will end up printing paper labels with the serial numbers, but it would
be incredibly clumsy and error prone. Looking at the downloads section on the LSI Logic website I found this tool, which seems to be a rough equivalent of mfiutil(8).

I didn't see this program mentioned on this list (or I don't remember) and, as I have seen mentioned that there is no equivalent to mfiutil(8) for the "mps" cards, well, I thought that it could be useful to someone, even though it doesn't serve the purpose I hoped to achieve: identifying the physical slot for a disk. That's why I pointed out that the "enclosure:slot" information is useless.

I hope it's more clear now.;) 

Sorry if I was so vague. I've been playing with this for two weeks, trying to find a solid configuration (and, for me, the definition of "solid" includes "the operator can identify a disk without confusion") and just now I notice that, after flashing the HBA, the enclosures have disappeared in two machines :/

Cheers,





Borja.