[SATLUG] Re: now help with raid setup on linuxmce box

John Pappas j at jvpappas.net
Tue Sep 23 09:46:56 CDT 2008


On Tue, Sep 23, 2008 at 07:14, Todd W. Bucy <toddwbucy at grandecom.net> wrote:

> On Mon, 2008-09-22 at 17:58 -0500, John Pappas wrote:
> > On Mon, Sep 22, 2008 at 13:11, Todd W. Bucy <toddwbucy at grandecom.net>
> wrote:
> > >
> > > In this particular box I will have 2 320 gig and 2 500 gig drives.
>  with my
> > > raid I want it all redundancy, as much
> > > space as possible and speed.
> >
> > How does this sound:  Partition up the 500GB disks like so:
> > SDx1 = 128MB RAID (Linux fd = raid autodetect) Bootable
> > SDx2 = 6GB RAID (linux fd)
> > SDx3 = ~174GB RAID (linux fd)
> > SDx4 = 320GB RAID (linux fd)
> >
> <<SNIP>
> >
>
> I thought of something like this but was warry.  I seem to remember a
> discussion on this topic a while back that trongly discouraged setting
> up multiple raids on the same drives.  Are there any particular worries
> that I need be aware of?


If this were a server hosting more than a household, "best practice" would
suggest to split RAID types on different raidsets, and there would be $$ to
support it.  In this case, the I/O loads are not going to be such that it
will matter.  With another drive (a 500 probably) you could do a RAID 5, get
1TB and have 33% overhead, and use the 320's for the OS mirror


> I know that if I have a failure of one of the 500 GB drives getting it
> replaced will be real complicated.  Are there
> strategies for addressing this now so as to make my life easier then?


In that case, you simply replace, repartition the replacement (Use the good
one to match partitioning) and run a couple commands (maybe not at the same
time, as the multiple rebuilds are I/O heavy).  You may have to manually
fail the partitions, but in the case of a real failure, that will be usually
taken care of by the system, so all of them will be 1/2 or 3/4, so you will
need to run the following to add the replacements and initiate rebuild:
`mdadm /dev/md0 -a /dev/sdx1`
`mdadm /dev/md1 -a /dev/sdx2`
`mdadm /dev/md2 -a /dev/sdx3`
`mdadm /dev/md3 -a /dev/sdx4`
Run `cat /proc/mdstat` to monitor progress.

Not too hard, right?

HTH,
jp


More information about the SATLUG mailing list