"Hard Disk Not Found" Error When Trying to Register Server

[Jamie Rozek]

Hello and good morning, Foggers. I’m having a new problem, which is probably just an ID10T error, so if you could point out the obvious, that would be great.

I’m trying to register a DL380 G3 Server. (Yes, I know. We have reasons for this…and we’re working on it. For the time being, it is what it is.)

This G3 server is running a Compaq Smart Array 5i+ in a RAID5 configuration. (Yes, I also know, gotta keep things as they are, unfortunately. Some things might break if they’re changed.)

The server PXE boots into the FOG GUI menu. I choose “Full Host Registration.” Then, all hell breaks loose.

I receive the following:

mptbase-ioc7 Error: WAIT IOC_READY State (0X20000000) timeout(15)!

That repeats a few times, then arrives at would be the registration screen. Instead, I get:

“Hard Disk Not Found.”

I went back and ran some compatibility checks.

Partition Information

Model: Compaq Smart Array (cpqarray)
Disk /dev/cciss/c0d0: 291GB
Sector size (logical/physical): 512B/512B
Partition Table: unknown
Disk Flags:

FOG Compatibility Test

Error: /dev/cciss/c0d0: unrecognized disk label

This computer appears to be compatible with FOG!

Results

Network…[Pass]
Disk…[Pass]

Any ideas? Basically, we’re trying FOG in a very serious disaster recovery scenario. (This is for a very critical server. If it goes down, we need it back up ASAP.)
I’m encouraged by the fact that it says it’s compatible, but I know I’ve had this issue with these servers before. It’s just, now it’s important. Thank you!

[Jamie Rozek]

While I’m awaiting an answer, I’m going to dig around and see if this is a kernel issue. Maybe the Smart Array driver isn’t included in the most recent release?

[ArchFan]

I think you’re on the right track with trying a newer kernel
[url]https://mastacontrola.com/downloads.html[/url]

[Jamie Rozek]

I tried the latest kernel. Same problem.

[Tom Elliott]

The part that weirds me out is the /dev/ccis/cc0d0 or somethign

FOG Only checks for /dev/sd{[a-z]} and /dev/hd{[a-z]}

Even then I think it stops trying if it hasn’t found anything through h

[Jamie Rozek]

Tom, cciss is a block driver for older HP Smart Array controllers like this machine. The C0D0 naming scheme indicates controller 0, disk 0, whole device.

Here is more information I found on it: [url]http://man7.org/linux/man-pages/man4/cciss.4.html[/url]

Is it possible that this controller is so old that fog won’t support it? Is it possible for me to build a kernel that includes the CCIS driver?

[Jamie Rozek]

This concerns me, does this mean that I may not be able to use this particular hardware for disaster recovery?

[Tom Elliott]

I’m fairly sure the drivers are already there.

However, the device it’s looking to get is not setup as a native device. You could try booting into debug and run fdisk -l. Does your device show up? If it does, what is the device listed as? You can then place that in the “kernel device” box for that host on the gui and all should work.

[ArchFan]

According to Tom’s 64-bit kernel config:
[CODE]
CONFIG_SCSI_HPSA=y
[/CODE]

According to some Googling, the cciss driver was deprecated and replaced with the hpsa driver, which is present in Tom’s kernel. It also appears, however that the 5i controller is compatible with the cciss driver only though, and I don’t see it in the .config. I don’t see any mention of a 5i+, only a 5i.
Source: [url]http://cciss.sourceforge.net/[/url]

[Tom Elliott]

While ArchFan is correct, the 5 series smart array is there:
BLK_CPQ_CISS_DA=y

[Jamie Rozek]

Alright, thanks for the reply guys. I seem to be getting a bit further.
I manually created the host in the GUI, and set it to run in debug mode.
When running FDISK -l I received this:

[root@192 /]# fdisk -l

Disk /dev/cciss/c0d0 271.3 GiB, 291331192328 bytes, 569006235 sectors
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes

So, I took that, went to the host in the GUI, and placed /dev/cciss/c0d0 in the “Host Primary Disk” box.

I then setup a debug download job, and went to it. Here is the output for that:

Below is a variable dump from FOG

osid=4
osname=Windows (Other)
mbrfile=
type=down
storage=192.168.6.8:/fog/images/
img=CRUSADERBASE
hd=/dev/cciss/c0d0
part=/dev/cciss/c0d01

I then ran the “fog” command:

Checking Mounted File System…Done

#Starting Image Push
#Using Image: CRUSADERBASE

#Looking for Hard Disks…Done
#Using Hard Disk: /dev/cciss/c0d0
#Erasing current MBR/GPT Tables…Done
#Restoring MBR…Done
#No extended partitions…Done
#Checking Hard Disks…Done
#Processing Partition: /dev/cciss/c0d0p1 (p1)
#Partition File Missing: /images/CRUSADERBASE/d1pp1.img*
Done
#Changing hostname…Done

#Updating Computer Database status

#Database Updated!

#Task is completed, computer will now restart.

[Jamie Rozek]

I’m also going to throw in there (while I’m trying to configure a kernel) that the DL380 G3 is strictly an x86 unit.

[ArchFan]

[quote=“Jamie Rozek, post: 38807, member: 24394”]I’m also going to throw in there (while I’m trying to configure a kernel) that the DL380 G3 is strictly an x86 unit.[/quote]

From how I understood what Tom said and looking at even the x86 kernel, the drivers you need are there, so it doesn’t sound like a kernel issue. I would think it’s related to the fact that “FOG Only checks for /dev/sd{[a-z]} and /dev/hd{[a-z]}” per Tom. That’s my take anyway, as a non-developer. I’m sure Tom will explain and correct me

[Jamie Rozek]

Right. I was just building a kernel for the heck of it while specifying the HP/Compaq drivers directly. Like I said previously, I got it to the point where I was doing a debug download. Now it is throwing a “Partition File Missing:” note.

[Jamie Rozek]

Yeah. The kernel didn’t work.

[Tom Elliott]

I’m taking a look to see if there’s a better way for me to get a list of hdd’s and partition data.