LVM Striping On Top Of Raid5 Striping - Is this Possible??
 

I have little experience with RAID and no experience with LVM so I'm going to introduce my situation here on very generic terms. From my recent research I have learned what striping means and that both RAID and LVM are capable of striping.

To begin, I started with two identical RH9 servers that I can only access remotely. Identical in *every* way so please take that literally (excluding network address). One is the active server and the 2nd is a backup server. They use RAID 5 and all the partitions are ext3(/ext2) except for the LVM data partition (and one other small LVM partition that has no obvious purpose to me).

My goal is to convert these servers to Cento 5.x so the backup server is my target for the first conversion to Centos. Afterwards (when I have a seemingly reliable strategy worked out) I will deploy this backup server as the new live server and then duplicate the conversion process on the old live server to create an identical backup server. Maybe not the best plan of attack, but the one we have chosen and already commited to so far.

In the process of installing Centos from CDs, we selected the custom installation. The idea here was, simply reformat the existing partitions as provided by Centos during the installation in hopes that we would not disrupt the existing LVM partition(s). That didn't work out as planned. Centos complained it could not reformat the LVMs so we selected options to delete and recreate them. That all seemed to work out fine according to fdisk and df results -- the resulting partitioning scheme remained comparable. Then I used 'lvmdiskscan' and got quite a different picture. Here my main concern is that we stared out with 'lvmdiskscan' showing "1 disk" and ended up after the Centos install showing "4 disks". If I interpret that correctly, it sounds like LVM was not striping in RH9 and after the Centos install, striping was added to LVM. That, on top of the inherent striping performed by RAID 5, means I have LVM striping on top of RAID striping. Is that possible, or is this (hopefully) a result of a more sophisticated 'lvmdiskscan' command simply informing me that this partition is being striped?

BEFORE (as seen via RH9 'lvmdiskscan'):
[root@axelserver root]# lvmdiskscan
lvmdiskscan -- reading all disks / partitions (this may take a while...)
lvmdiskscan -- /dev/sda1 [ 39.19 MB] Primary [0xDE]
lvmdiskscan -- /dev/sda2 [ 2.50 GB] Primary LVM partition [0x8E]
lvmdiskscan -- /dev/sda3 [ 101.97 MB] Primary LINUX native partition [0x83]
lvmdiskscan -- /dev/sda4 [ 98.96 GB] Primary Windows98 extended partition [0x0F]
lvmdiskscan -- /dev/sda5 [ 9.76 GB] Extended LINUX native partition [0x83]
lvmdiskscan -- /dev/sda6 [ 1.95 GB] Extended LINUX swap partition [0x82]
lvmdiskscan -- /dev/sda7 [ 1 GB] Extended LINUX native partition [0x83]
lvmdiskscan -- /dev/sda8 [ 1 GB] Extended LINUX native partition [0x83]
lvmdiskscan -- /dev/sda9 [ 9.76 GB] Extended LINUX native partition [0x83]
lvmdiskscan -- /dev/sda10 [ 75.49 GB] Extended LVM partition [0x8E]
lvmdiskscan -- 1 disk
lvmdiskscan -- 0 whole disks
lvmdiskscan -- 0 loop devices
lvmdiskscan -- 0 multiple devices
lvmdiskscan -- 0 network block devices
lvmdiskscan -- 10 partitions
lvmdiskscan -- 2 LVM physical volume partitions

AFTER (as seen via Centos 'lvmdiskscan'):
[root@localhost ~]# lvmdiskscan
/dev/ramdisk [ 16.00 MB]
/dev/vg01/lv_data [ 75.38 GB]
/dev/ram [ 16.00 MB]
/dev/sda1 [ 39.19 MB]
/dev/ram2 [ 16.00 MB]
/dev/sda2 [ 2.50 GB] LVM physical volume
/dev/ram3 [ 16.00 MB]
/dev/sda3 [ 101.98 MB]
/dev/ram4 [ 16.00 MB]
/dev/ram5 [ 16.00 MB]
/dev/sda5 [ 9.76 GB]
/dev/ram6 [ 16.00 MB]
/dev/sda6 [ 1.95 GB]
/dev/ram7 [ 16.00 MB]
/dev/root [ 1.00 GB]
/dev/ram8 [ 16.00 MB]
/dev/sda8 [ 1.00 GB]
/dev/ram9 [ 16.00 MB]
/dev/sda9 [ 9.76 GB]
/dev/ram10 [ 16.00 MB]
/dev/sda10 [ 75.49 GB] LVM physical volume
/dev/ram11 [ 16.00 MB]
/dev/ram12 [ 16.00 MB]
/dev/ram13 [ 16.00 MB]
/dev/ram14 [ 16.00 MB]
/dev/ram15 [ 16.00 MB]
4 disks
20 partitions
0 LVM physical volume whole disks
2 LVM physical volumes

Main concern here is: why does it now show 4 disks?

Thanks, if you can answer!!! (or offer better plan of attack)

I can't help you much, but here are some suggestions:

If it's possible to test it, try to see what the output would be after the raid5 setup and before the centos install - use the CentOS rescue/install disk to do the lvmdiskscan. Also, check to see if there are any version differences between lvmdiskscan on RHEL9 and CentOS 5.x. I don't know how you specifically did the CentOS install, but I can't imagine that the install procedure would go ahead and do a lvm strip setup without you specifically telling it to do so. My feeling is that your setup is fine, but you most likely should find an answer to your situation to make sure it won't be a problem source in the future.

It might help others help you if you give more details on how you did the CentOS install - specifically how you setup the LVM - were any command line commands executed or was the entire install done through the CentOS install procedure program.

If you don't get an answer here, the CentOS community is another place to look for help.

Hi btncix... The Centos 5.4 installation was performed exclusively and directly from the CDs. I do not have firsthand access to these servers so I am relying on a third party to do the installation. Neither one of us know anything about line command installation. I can tell you the servers use raid 5 configured inside the BIOS -- settings which we did not change so they presumably are identical to those that existed during the original RH9 installation.

I like your idea to use the CentOS rescue/install disk to do the lvmdiskscan on the sister server, but problem there is that is the live server. The third party I am working with represents the most paranoid employer on the planet so I don't think they will allow us to touch that live server until we have a proven replacement ready to drop in its place. I am hoping I can get my third party tech guy to run the install again and scrutinize that LVM partitioning stage for any possible tweaks in the setup.

Haven't played with CentOS in a while, and I dislike LVM, so treat this as you may.
Possibly that's LVM1 on RH9, and LVM2 on CentOS - likely to be all sorts of changes if true. Try "lvmdiskscan -l" so it only reports the pvs.