rootvg spanned on multiple disks
Hi All,
After installation of linux RHEL-5 on VM through TPM (tivoli provisioning manager),the rootvg is spanned accross two disks (1st and 7th disks),which is intended to use only one disk i.e 1st disk.
Any ideas on why it is behaving so?
Kind Regards,
Ramakrishna p
Responses
Guru
6863 points
I don't know how TPM specifically works. However, when using kickstart (in general) it will refer to a kickstart configuration file (typically located on a web server, or NFS share). If you do not specify which disks to use, the results can be unexpected, and usually are.
You can check /root/anaconda-ks.cfg to see how your kickstart/installation had configured the drive and report that back and we should be able to identify exactly how/why it used multiple devices.
Here is an example of my (fedora) installation:
#clearpart --linux --drives=sda
#part pv.008002 --onpart=sda2 --noformat
#volgroup vg_morpheus --pesize=4096 --useexisting --noformat pv.008002
#logvol swap --name=lv_swap --vgname=vg_morpheus --useexisting
#logvol / --fstype=ext4 --name=lv_root --vgname=vg_morpheus --useexisting
#logvol /export/isos --fstype=ext4 --name=lv_isos --vgname=vg_morpheus --useexisting --noformat
#logvol /home --fstype=ext4 --name=lv_home --vgname=vg_morpheus --useexisting --noformat
#part /boot --fstype=ext4 --onpart=sda1
Hello Ramakrishna,
One of the solution to this problem is to craft a kickstart that does not statically reference a specific device node (such as /dev/sda), but rather inspects each device and builds a dynamic list of devices to use based on certain criteria.
Both RHEL 5 and RHEL 6 support the ignoredisk option, which allows you to define a list of devices that anaconda will restrict itself to (--only-use) or a list of devices to ignore (--drives).
Combined with the use of a script in the %pre section, one can loop through all available devices and only use (or ignore) those that meet certain criteria.
Please refer the following article for detail configuration
https://access.redhat.com/knowledge/solutions/58561 Order of storage devices during installation causing difficulties with anaconda kickstart in RHEL
Hope this helps.
Regards,
Nitin Yewale
Guru
6863 points
hello Ramakrishna - without seeing your kickstart file, we will not be able to deduce why it behaved that way. We don't need the entire file, just the portion that addresses the storage. Once we see what is in your kickstart (or anaconda) file, we can then try to get to the root cause, which is likely how TPM is setting up your host for installation.
My guess it there is a "drive layout" portion of the TPM interface that you can configure for new machines being provisioned and deployed.
Ah, yes: the joys of automated provisioning. You'll also run into this kind of fun when P2Ving (or even V2Ving) a system that has more than one partition. A lot of tools don't really grok that there's not necessarily a 1:1 relationship between mountpoints and underlying (virtual) disks. Depending on how many "decisions" you leave up to your provisioning (or virtualization-conversion) tools, you can come up with some *really* sub-optimal storage layouts.
Another fun thing your provisioning tool probably did was created partitions that are poorly-aligned to the underlying storage subsystems (more problematic with RHEL 5 than RHEL 6). Not a big deal in a small environment, but when you have tens or hundreds of poorly-aligned virtual disks on a given physical data-store, the tiny inefficiencies can mount up to noticeable negative impacts within your environment.
At any rate, when you want to optimize the results of automated provisioning, you need to go into a fair amount of detail in your provisioning rules/scripts to ensure the results you get are the results you wanted. Automated provisioning is yet another one of those things that acts as a great illustration of the principal of "conservation of complexity".
Welcome! Check out the Getting Started with Red Hat page for quick tours and guides for common tasks.