On Fri, 2010-04-02 at 14:45 +0200, Volker Armin Hemmann wrote: > On Freitag 02 April 2010, meino.cra...@gmx.de wrote: > > Neil Bothwick <n...@digimed.co.uk> [10-04-02 14:08]: > > > On Fri, 2 Apr 2010 13:04:53 +0200, meino.cra...@gmx.de wrote: > > > > only to be sure to have understood everything correctly: > > > > Suggestion is to create for example one root partition and a swap > > > > partion. And I will create on big "rest of the disk"-partition. > > > > The last one will be subdivided with LVM into portions as needed. > > > > > > Yes. > > > > > > > Since the last big partition is big due to physical reasons (not for > > > > logical one): What will happen, if -- for example -- one portion will > > > > be not unmounted cleanly and while booting/checking fails to recover? > > > > Are all others damaged/lost? > > > > > > No, because the failure you describe is at the filesystem level. Even the > > > volume containing that filesystem will retain integrity, only the > > > filesystem itself will be corrupted. As you have left free space on the > > > volume group, you can just create a new volume, format it and copy over > > > everything you can recover from the broken filesystem before deleting it. > > > > Hi Neil, > > > > yes, sounds good, very good. > > Last question: How heavy is the performance impact of such a setup ? > > seriously lvm sounds nice. But it isn't. It easily breaks. > > You want a save setup? Go raid5 or raid6. As a bonus - you can get more space > if you need it by just adding another disk. And you are not depending on some > complex stuff to get it working. >
My experience is lvm itself is quite robust and very low impact on performance. More reliable than linux software raid at least (well the raid 0 that I was using: ) - never had a problem I could trace to lvm. The only thing thats affected lvm for me were hardware errors (disk died). My experience was with raid 0, while the higher raid redundancy will shift the reliability figures back the other way. Its really down to space and management or losing space to redundancy. Yes its an extra layer on top of the raw hardware (but so is raid really) so its the flexibility thats important. BillK
