This content has been marked as final. Show 6 replies
Either Jumpstart or Flash Archives should fit the bill and you can find the documentation on docs.oracle.com. For Solaris 11 when you get ready to make the jump look into Automated Installer (Ai).
first I would convert both drives to either ufs or zfs.
There are 2 common methods in use to quickly recover a server after a drive failure
The first method is to use "live upgrade" to clone the os (/var and /root). If one drive fails, or if important os data is erased, then the server can be booted to the other disk. You could probably use solaris volume manager to mirror the date.
The second method is to use Solaris volume manager to mirror the os and and the data. In my opinion this is a bit more complicated to implement, and the recovery is a bit more complicated. In addition if the drive fails bad data can be written to the remaining good disk. Mirroring does not protect you from data being accidently erased. On the other hand the alternate root drive is always up to date. With live upgrade you either need to update the alternate disk manually or use a script to automatically upgrade the script.
Flar looks good and maybe more appropriate for my use case than Jumpstart. Just need a full image to keep around as a baseline that I can recover from as needed.
(No need for continual updating per Paisley's drive sync suggestion.)
But, will flar correctly back up both the UFS and ZFS drives into one big archive?
If not, is there something else I need to do?
If you must use both UFS and ZFS and you are tight on disk space, then you will need bare metal
recovery for your UFS root file system and snapshots of your ZFS data pool.
If you only have one spare disk to hold both UFS and ZFS data, then I think it will be a bit more
difficult. You could create a flash archive of your UFS root file system and store it on your spare disk
in a ZFS storage pool. Its just a cpio archive, but I don't think you could also use JumpStart to install
this flar. I think the JumpStart server and client file system need to match.
Another idea, and this is untested by me, is that you can install the system manually by pointing
the installer at the UFS root flar on a ZFS file system shared over NFS. This needs to be tested,
You could create a ZFS storage pool on the whole spare disk and create a separate file system for
the UFS flash archive (and confirm that you could install this flar) and depending on how you
create your ZFS snapshots, the ZFS file systems would get created automatically when the file
systems are received. Or, you could store the snapshots as files in a ZFS file system.
If the "store a UFS flar on a ZFS file system" idea doesn't work out, you could create partitions on the
spare disk for a UFS file system for storing the ufsdump files and a ZFS pool for storing your ZFS data
My advice is to ask Santa for new disks or a new system for Xmas and set up your system(s) so that
management and recovery of your data is much easier:
ZFS root pool (mirror of 2 disks)
ZFS data pool (mirror of 2 disks)
ZFS backup pool (mirror of 2 disks)
A little more info is needed to give you the best solution.
How quickly do you need to restore the system?
What is your backup method? (tape drive, disk)
Will all of your data fit on one disk?
Will you be able to convert ufs to zfs?
I think there is an article on how to use Live Upgrade to move from ufs to zfs.
There are 2 methods in use to quickly recover a server after a drive failure.
1. live upgrade:- using Lucreate you can have two (BE) Boot Environment on same harware. If one drive fail you can boot with second BE.
2. Flash Archive:- using flarcreate you can have a flar image of running OS on SAN/NAS or FTP location. it is easy to recover OS using flash installation.