translate raid controller "units" to mounted disks

[wolvie]

Hi,

I have been handed a fileserver that was run by a string of grad students (ie no sysadmin) that has a number of disks that are apparently not mounted. It is an off-brand system with 24 disks, 3ware 9550SX-12MI raid controller, running centos 5.2 final.
I can see which physical drives are in which "unit" on the raid controller, but I can't find a way to translate that into the /dev/sdxx assignments. The vgscan shows that there appears to be a missing volume group with some drives assigned to it, so I want to remove that vg, the physical drives in it and replace them with some new larger size drives. But there are also 3 or 4 single drive units that I cannot pin down. Unit 0 on controller 0 is a single drive that seems to be /dev/sda and /boot, but if I am wrong about that and pull the wrong single drive obviously problems could ensue.

Unfortunately, I am told the system has not been shut down or restarted in years because the last time the grad students did so, several services stopped working properly and they had no idea how to fix them. The grad students seemed to have a habit of finding "interesting" but non-standard ways to implement things and not remember to apply changes after testing them, which I think is why things stop working after reboots. Since everything is hot swappable, I am hoping to do this without a restart. (I am also eager for suggestions on how to "image" this system to allow recovery in case /boot dies at some point. I have been using Acronis for imaging windows servers with some success.)

Anyway, my question is what do I do to determine exactly which units in the raid controller translate into which disks (/dev/sdxx) in the system?
BTW it does seem to be using device mapper. Also the disks in question are all the same size and model, so I can't use that to figure it out.

Thanks for any help you may offer and I apologize if I am not concise - it is my first post here. (I did search for an answer for quite a while before posting the question, but I apologize if I should have found it.)

[TrevorH]

CentOS 5.2 :-( You know that's only 3.5 years out of date with security patches etc

Start by looking in the output from `dmesg` and see if there are mentions there of the 3wre module loading as that normally tells you what disks it has found.

I have to say that from your description of the state of the system, I do wonder if it's worth setting up a parallel install and migrating the data and services from this machine to it so that you have a known stable baseline to start from. What you have now sounds like it might be tricky to tell what's been done and how.

[wolvie]

Thanks for the reply.

I would love to build a new system to transfer services, unfortunately, they have no budget for another system and my time is limited to 50hrs/year by the powers that be, so I am trying to do what they ask. (I have told them it is a terrible idea to run such an old version, I can't guarantee things will keep working without the proper work to maintain the system, etc. )

I thought I had seen some discussion at some point, years ago, of commands that would translate units into /dev/ ids, but vgscan, pvscan, lvscan do not seem to do so.

I will parse the output from dmesg and see if it helps.

Thanks again.