lists at jnielsen.net
Thu Mar 14 09:45:39 MDT 2013
On Mar 13, 2013, at 6:21 PM, Daniel Fussell <dfussell at byu.edu> wrote:
> On 03/13/2013 05:06 PM, Corey Edwards wrote:
>> On 03/13/2013 04:53 PM, John Nielsen wrote:
>>> Using LVM would give me what I'm looking for, but I really would just
>>> use it for the naming so it seems kind of silly: - each drive would
>>> be its own volume group - each volume group would have exactly one
>>> logical volume
>> I would still recommend LVM. The overhead is essentially nil, it's quite
>> standard across distributions, and it's robust. You'll also have
>> flexibility if your needs change down the road.
> And by flexibility, he means crazy awesomeness like moving the volume to
> a different drive while it's being used.
> Just make sure you have a decent replacement plan for _when_ a drive
> dies. LVM will only make that situation worse, or at least more
> complicated, if you can't handle and recover from the failure.
Thanks for the responses! LVM it is.
The data on these disks will be replicated at the application layer, so a disk failure is not an emergency. When the disk is replaced, it just needs to have volume group and logical volume names that match what is in fstab, then get a new filesystem. Once it's mounted the application can repopulate its contents.
More information about the PLUG