[mythtv] Problems with using LVM for large storage

[Keith C]

> Consider having three disks mounted as /store1, /store2 and / 
> store3.  Now you can take down any one of them and the other two  
> are still accessible.
> With Raid0 though you lose the whole array.  There is sometimes  
> another mode called JBOD which can sometimes work where a failed  
> disk just looses you some (parts) of the files, but I have never  
> tried this...
>
> I can see some definite merits in the idea of having multiple store  
> locations, even if the record location is fixed and you have to  
> shuffle them around later.
>
> I just bought a new 400Gb drive to replace the 300Gb on in my  
> current machine.  There is 1Tb on another fileserver in RAID5.  One  
> thing which becomes apparent with this amount of storage is that  
> it's quite hard to shuffle things around later and upgrade the odd  
> disk here and there...  It's very easy to sit there and say make it  
> Raid 1/5, but the cost, heat, power noise and size requirements are  
> considerable - not to mention the hassle when you need to upgrade
>

I used to have a master and slave backend, each with their own local  
big disks for storage.  Each backend was set to use its own disk for  
buffers and recordings (/video1 and /video2).  And then I used NFS  
automounting to have both shares available to both backends.  It  
seemed to work fine, although MythWeb was a little confused by it (if  
I tried to click a thumbnail to download a video, otherwise it was  
mostly ok).

However, it was not stable.  One backend or the other would lockup  
randomly, with no logs or segfaults to help me debug it.  So I moved  
all the tuners and disks to one backend and it never locks up now.   
This was just my experience though, it might be completely possible  
to have this work without locking up.

One of the main drawbacks I see to an LVM is the inability to upgrade  
a disk.  If I want to pull a 120GB drive out of the LVM and drop in a  
300 GB, it might actually be possible, but a massive in the arse to do.

I too would be interested in a better solution.  The main reason I  
tried the master/slave before was to have my HD recordings local on  
the frontend (so as to not bog down the network), but not have all 3  
disks (540 GB total storage) whirring away in my living room.  If I  
could instead say : Use video1 for new recordings, and instead of  
deleting old video, move them to video2 (which could be an NFS  
mount).  If video2 gets full, delete the oldest recording from there.

Overly complicated?  Maybe.  Hard to configure using Setup?   
Probably.  But a neat idea none the less.

Keith C