Brian Martin gave an (as always) excellent talk about LVM Thursday night. As I get older and more cynical, I look at software that seems very nice and wonder "what happens if this goes wrong? How do I debug it?"
My concerns about LVM are about rapidly dealing with mistakes and hardware failure in stressful situations (ever had a laptop drive go flaky 4 hours before a presentation? And successfully copied the presentation to a live spare with minutes to spare?). Brian gave the example of moving data from a failed drive to a working drive in the same volume group. But if all you have for checking data integrity is fsck at the file system level, and a file system is spread across many drives in a volume group, how do you tell which block errors correspond to which drive - quickly? For those of us with file systems smaller than the largest drives available, there is something to be said for buying a drive that is 3x larger than the current data set, and buying a new and larger drive when we fill the current drive to 80% or more. For those of us using laptops, adding a second drive usually isn't practical. So for many situations, LVM doesn't sound practical. OTOH, if one needs a 20TB file system on a server class machine, and assuming there are simple ways to identify starting-to-fail drives, then LVM sounds great. I run Redhat-derived distros, but bypass the default LVM setup of the disks and use bare metal partitions instead. It sounds like I should keep doing that until I cannot fit my system on one drive. Keith -- Keith Lofstrom [email protected] Voice (503)-520-1993 KLIC --- Keith Lofstrom Integrated Circuits --- "Your Ideas in Silicon" Design Contracting in Bipolar and CMOS - Analog, Digital, and Scan ICs _______________________________________________ PLUG mailing list [email protected] http://lists.pdxlinux.org/mailman/listinfo/plug
