On Wednesday, Mar 30, 2005, at 19:10 US/Central, Mark wrote:
1. What happens if a LV spans over mutliple hard disks and one of the disks fails while the server is running? Can the server still keep on running on the other disks as long as no data from the broken disk is needed? Or does everything just crash once one of the disks dies? As an extreme case, lets assume the broken disk does not contain any files or directories that are used at all.
Short answer: you're totally hosed. Long answer: you're somewhere between partly hosed and totally hosed. But even if you're partly hosed, by the time you fix things and get back up to speed you will have wished you had setup a RAID.
2. How big is the performance loss? Obviously, an additional layer in the file system hierarchy does mean additional mapping and handling, meaning additional processor usage, meaning performance loss. In most cases this probably does not matter, but what if I use the server for example as a mysql database server?
You're running a server. You want data redundancy. You want performance. Sounds like you want at a minimum RAID-5 and possibly RAID-10. LVM gives you none of the above. The only thing LVM gives you is flexibility in how you manage your disk space, the ease with which you can create/reduce/enlarge/remove virtual partitions, i.e. logical volumes.
But the best way to answer your questions is to actually setup a test system and experiment with it. You don't even need a system with multiple drives. Just partition a single disk with a few (at least 3) extra partitions of say 200 MB each. Then create the LVM and later remove one of the partitions. What do you think will happen? What actually does happen? Now repeat, testing for performance.
And please, let the list know how things go.
Regards, - Robert http://www.cwelug.org/downloads Help others get OpenSource software. Distribute FLOSS for Windows, Linux, *BSD, and MacOS X with BitTorrent
