Migration Strategy
New linode.com member here, and I have a few questions about the best practice and pros/cons about various migration strategies.
I will be using a custom CentOS 5 install with LVM, completely replacing the linode.com version. The way I see it, I have a couple of options:
1. Perform the install locally, then after changing the drive block devices, etc, upload it and dd replace what's there using Finnix.
2. Set up my drives using the web interface, then using Finnix simply copy the files into the new respective partitions, while excluding things like /proc, etc.
With option #1, will it essentially be running LVM on top of LVM (since I understand that's what linode.com uses under the hood)? Then again, perhaps that will not be the case since the "physical" drives are changing. Will that add extra I/O overhead? Finally, would I need to use the initrd from my own image and use the web interface to specify that?
Option #2 seems pretty straight-forward and easy. Any gotchas pros/cons with that?
Thanks in advance.
9 Replies
UML
@anderiv:
I've always used the "canned" distros that linode provides, so I can't comment on your migration strategy. I can say, however, that linode is currently using
under the hood, not LVM. UML
They are two different things. UML is the virtualization layer, while LVM (logical volume manager) is a way to manage disk space.
@mstarks01:
They are two different things. UML is the virtualization layer, while LVM (logical volume manager) is a way to manage disk space.
Hah - I mistook LVM for KVM. Nevermind me…
When you copy across, you have to 'Linode-ise' your distro - the instructions are in the wiki.
@pclissold:
I use LVM on my Linode without problems. I use a regular Linode partition for the root partition (no initrd) and I have a minimal Debian distro with the LVM tools installed - instead of using Finnix. I used the Debian distro to create the LVM volumes. I built my custom system on the Linode, rather than copying it across - but the principle is the same.
When you copy across, you have to 'Linode-ise' your distro - the instructions are in the wiki.
Do you mean that you did the actual custom Debian install on the linode? If so, what did you boot from and how did you get the install going? Please elaborate. Thanks.
After that, you need to create UML disk devices, edit fstab, make sure that you're starting a getty process on tty0 (for the Lish console) and add tty0 to /etc/securetty.
Use LPM to set your new root partition as the boot device.
@pclissold:
I then installed my system from scratch. You could copy your system to the new root + mounted LVM logical volumes.
OK, here's where you lost me. I understand how you created the partitions and so on. But to install the system did you simply copy over files that you had created by doing an install somewhere else? Or did you (perhaps chroot) run an installer?
@mstarks01:
But to install the system did you simply copy over files that you had created by doing an install somewhere else? Or did you (perhaps chroot) run an installer?
Both - I've downloaded a Gentoo tarball, untarred and the chrooted to do the install, and I've copied across a Linux-From-Scratch system built elsewhere.
Just to warn you in case it cropped up at all.