samba via nfs = bad

Paul Oldham paul at the-hug.org
Sun Oct 26 00:01:45 CEST 2008


On Sat, October 25, 2008 14:47, Drew Fitzsimmons said:

> At work we have got some linux machines which run a proprietary
> software product. The vendor supplied the whole system on fedora core
> 2 boxes and the set up runs something like this:
> One of the boxes acts a s a server and contains the data which is used
> by all the other systems.
> All the other 6 boxes connect to a single nfs share on that machine.
>
> Recently there was a disk failure on the server box and all the data
> went away...
>
> I haven't had much to do with the systems but it turns out that the
> server runs a single 10k rpm disk, no RAID no redundancy at all. This
> is clearly a very bad idea and now that one failure has happened it
> has been decided that the system should be a little more robust
> (always the way) [...]

Why not just slap a second disk drive of the same spec as the original in
the server box (assuming there's a spare slot) and then configure the
drives as a software RAID 1 pair? Minimal hardware cost and simple enough
to do.

-- 
Paul




More information about the CLUG mailing list