The server has an HBA now…
The server we bought last week came with a Dell H700, which is apparently a reasonably competent RAID controller, with a battery backup and half a gigabyte of cache and all that good stuff, but that’s not reallysuitable for our purposes. I’m using ZFS for our main storage, which apparently works better when given raw access to the disk, something this controller can’t provide. How much of that is true, and how much of that is the cargo-cult mentality that surrounds the community that tends to fuck around with old servers? I’m not qualified to say.
After much research, I determined that the safest bet would be getting an LSI SAS controller of the sort that could be flashed into IT mode (Initiator-Target, which basically just means controller and drive, both dumb as they come). The LSI SAS-9211-8i looked to be just the thing, and I found two a few days later for about $70 each shipped. Snoozing on them because of my indecisiveness, they sold rather quickly. I very nearly plopped the cash for a $150 card, brand new from LSI, pre-flashed to IT mode, when I happened upon a IBM ServeRaid M1015, which is an LSI-9220-8i, which is effectively the same card only sold to be badge-engineered.
As it turns out with a bit of finagling, these can be cross-flashed with the same firmware from a 9211-8i and give me everything I want. And it’s thirty bucks! Definitely worth the gamble, so I bought it… and it arrived tonight.
I hooked it up and verified it worked, then set about flashing it. It seems that because of the vendor specific firmware, modern versions of LSI’s flash tools don’t want to know it. There’s another tool, MegaRec, that folks use to write the SBR (“Serial Boot ROM”, apparently), and someone’s knocked together an empty SBR image that basically zeroes it. This fools LSI’s tool and lets you flash it. Awesome!
Except MegaRec didn’t work on my server. This had me concerned, because there’s no shortage of folks saying “do not reboot after this point, you’ll brick your card!”, so I didn’t know if it was safe to do anything! I left it go for about ten minutes, hanging at (or similar, typed up from memory, screenshots, and searching):
MegaRAID HWR Controller Recovery tool. Version 01.01-004 February 05, 2010 Copyright (c) 2006-2008 LSI Corp. Supports 1078 controller and its Successors.
As above, I let it sit for 10+ minutes and nothing, so I rebooted, and the card came up fine, and I determined it was fairly safe at this point, maybe the card hadn’t started being erased? That appears to be the case.
Trying the sas2flsh utility, and I get the familiar “No LSI controllers found” message that the rebranded cards are apparently prone to. I tried to find older versions of the utility that allegedly don’t have this limitation, to no avail. Tried the UEFI equivalent, no luck there either… I couldn’t get an EFI shell on the R510, and my desktop came up with the same message. But what about MegaRec on my desktop?
Success! I deleted the SBR and wiped the equivalent area of the flash. Awesome! Tried sas2flsh… No PAL. Fuck, that means you’ve got to do it in EFI mode. But if I reboot, the internet says I may brick my card? Not much can be done now, I’ve either fucked thirty bucks or I haven’t, so I rebooted into the EFI shell and did the sas2flash.efi commands, and it worked!
I put the card back in the 510, and installed Ubuntu without issue on the internal 2.5” disk that’s inside the machine. But there are some complications with that process also, that I skipped over above.
The internet likes to tell you that in the “storage” slot of the PowerEdge machines, you must use a Dell branded card with Dell firmware, or else it won’t boot. This is either wrong, outdated, or doesn’t apply to the 510, because this one booted fine with both the IBM branded RAID card, and the LSI branded SAS firmware. The problem? It’s physically too long for the fucking slot. So I was trying to flash it in one of the rear slots, and when I finally got it flashed and went to put it back in… my SFF-8087 cables are too short, no matter how they’re oriented.
So what to do? Order longer cables naturally, but in the mean time I’m impatient. I pulled the riser out, and shucked it from the metal hanger it’s attached to, slotted it back in, then sat the HBA in the storage slot. This let it clear everything, albeit unsupported, but with the orientation of both the fan power cable and the SFF-8087 cables, it’s held in there fairly well. It’ll only be for a few weeks, and I’ve done much worse with my home servers.
Long story short: I’m not sure how much to believe when someone says you must do something a certain way or you risk losing your data or hardware.