[mythtv-users] Faster CPU and more RAM in SD frontend or backend?

[Kevin Kuphal]

On Tue, Jun 3, 2008 at 5:59 AM, Adolfo R. Brandes <arbrandes at gmail.com> wrote:
>
> Kevin Kuphal wrote:
>> Don't bother [using RAID-5].  Myth can use each drive independently
>> (...) and I find the possibility of losing everything if the RAID
>> array fails is far worse...
>
>        I respectfully disagree.  These days, software RAID on Linux is
> simultaneously robust and flexible (which is more than you can say for
> hardware RAID).  I've been running 2 arrays (RAID-1 and RAID-5) for
> months, and they have survived the following:
>
> 1) Lightning strike (RAID-5)
>
>        An electrical surge "stunned" one of the hds in the array,
> essentially killing it.  I was able to take it out, fix it, and put it back
> in the array without losing any data, AND with the backend running in
> the meantime (very little downtime).

Agreed.  It certainly does cover you in case of hardware failure.  But
again, it's just TV and anything that's worth keeping can also be
backed up onto DVD-DL or other media.  This is the area where RAID-5
has an advantage over plain storage groups but for my use, it is an
advantage not worth any potential problem.  I stress potential because
while RAID in software on Linux is stable (heck, I'm using it on my
iSCSI storage mount to my Vista machine for all my photo processing),
I use it when I need hardware redundancy and for me, TV isn't that
place.

>
> 2) RAID expansion (RAID-5)
>
>        Just last month I was able to stick in a new hd, expand the
> RAID-5 array, extend the LV, AND extend the ext3 filesystem on top, all
> with the backend online (except for the 10 minutes required to plug the
> new hd in).  No failures whatsoever.  Talk about flexibility!

By the same token, I've also found difficulty going the other
direction.  Shrinking storage is darn near impossible from my
experience while storage groups just allows Myth to use whatever
storage is available regardless of shrink/grow.  RAID-5 isn't unique
in allowing for expansion and there is more chance for failure in
growing a RAID-5, LV, and ext3 filesystem than adding a drive,
mounting it, and adding it to a storage group.

>
> 3) Flaky S3 (RAID-1)
>
>        My frontend runs on software RAID-1, and I like to suspend it
> instead of powering off.  However, kernel support for the Asus
> M2NPV-VM's ACPI is flaky at best, so occasionally the machine will hang
> on waking (the ugly kind of hang: no response at all from the system).
> I then need to hard-reset the box, which forces the driver to rebuild
> the array on the next boot.  Nevertheless, even after dozens of resets,
> I have yet to lose one tiny file on it.

Again, not necessarily specific to RAID but the filesystem on it.  But
I understand your point.  Just offering another perspective that for
many, RAID-5 can be an added set of hassles that can be addressed with
storage groups as long as you don't require hardware redundancy and
can live with the additional complexities it adds.

Kevin