--- Begin Message ---My system:
- To: lvm10@packages.debian.org
- Subject: lvm setup problems
- From: Bob <bgeer@linkmail.net>
- Date: Fri, 20 Jun 2003 13:47:58 -0400
- Cc: lvm-common@packages.debian.org
- Delivered-to: lvm10@packages.debian.org
- Delivery-date: Fri, 20 Jun 2003 18:50:14 +0100
- Envelope-to: patrick@localhost
- User-agent: Mozilla/5.0 (Windows; U; Win 9x 4.90; en-US; rv:1.0.2) Gecko/20021120 Netscape/7.01 (NSCD7.01)
CPU: 200MHz Intel Pentium MLB: AOpen AP53 IDE Drive: 540M Maxtor SCSI Adapter: Compaq wide ultra scsi (68 pin) (uses ncr53c8xx.o) SCSI Drives: 2 Compaq (Seagate) 4.5G 1 Fujitsu 4.5G 1 Seagate 18G
All of the above hardware is 100% functional and works with this and other software. I'm exploring storage management for myself and for a friend who runs an internet server farm.
My reference data:
Man Pages On-line howto from Sistina various files included in the package
Problems:
1. The howto tells me to alter the command lines described if I'm using "devfs" but doesn't tell me (a novice) how I'd know whether or not I have it. In /usr/share/doc/lvm-common/readme.gz I am told how to check by inference: it says that a test is made in /etc/init.d/lvm. However, there is no test in the script I have.
2. The stock installation tries to install a module "lvm" in /etc/modules. Exploring my system there is no "lvm.o" anywhere in the /lib tree. Consequently this gives me a modprobe error on every startup. I solved this problem by changing the "lvm" line to "lvm-mod" and everything seems to work okay. There is a related error in the /etc/init.d/lvm script, but it doesn't seem to have any harmful effect. Its modprobe output is redirected to /dev/null. I commented it out anyway.
3. This is a strange and hard to discover problem. I played around with it for one whole day. Only my faith in the Linux community and Sistina kept me after it. The simple explanation is:
--------------------------------------------------------------
fdisk /dev/sda1 . . pvcreate /dev/sda1 ..repeat the above for /dev/sd[b-d]1...
-OR-
dd if=/dev/zero of=/dev/sda1 bs=1K count=1 pvcreate /dev/sd[a-d]
-THEN-
vgcreate -n vg0 /dev/sd[a-d]1 lvcreate -L 9 -n lv0 vg0 lvcreate -L 20 -n lv1 vg0
-the following is NOT included in the howto-
vgscan
-----------------------------------------------------------------
All up to the last command ("vgscan") works perfectly and produces usable devices. However, if you don't do a "vgscan" before you reboot you can "vgscan" to your heart's content and all you'll get is error message after error message. The volume group cannot be used because it can't be found. If you try to re-create it, THEN it can be found and you can't do that either. Even "lvmchange -R" doesn't remove all traces of the volume group(s) created. I used "find / -name vg0" (and variants) to hunt them down an kill them in places such as /dev. Also common sense made me look in /etc/lvm and the like.
BTW the difference between "pvcreate /dev/sda" and "pvcreate /dev/sda1" is not well explained (at least not where it matters.) I think I used the "pvcreate /dev/sda1" version. I can't remember for sure because I'm old and my memory isn't as good as it used to be. Guess I'll try both ways later.
Not withstanding all of the above and more, the end result was a fine storage system and great results. We'll probably use lvm on my friend's Red Hat (oops) internet systems too. I hope their lvm layout isn't too different from Debian's.
I also hope that this is of some help to you and the Sistina howto people. I hate to seem like I'm griping or whining because I'm not. This is a genuine attempt to improve an already great product. Assuming you got this far, thanks for your time and effort in reading this account.
Sincerely,
Bob Geer
--- End Message ---
