[mythtv-users] Question re: available SATA ports and linux software RAID

[Simon Hobson]

f-myth-users at media.mit.edu wrote:
>Could I -please- ask that people who keep citing Newegg URLs actually
>include THE REAL MODEL NUMBER of the product as well?
>
>Depending on Newegg URLs to be stable forever is a bad plan, and some
>models outlast them.  They also completely foil people doing web
>searches for the pros and cons of any of these devices---they'll
>never find your URLs.  And finally, they make even random readers,
>some of whom may already have these devices, have to stop and look
>up every single URL to see what you're talking about.  That pretty
>much maximizes the chances that someone will just say "screw it"
>instead of actually commenting on what luck they've had with the
>device.

Well said ! I see the same problem elsewhere where people link to an 
item on eBay :-/



Bobby Gill wrote:

>So now 3 questions arise:
>
>RAID is for availability, rather than backup, so unless you like the idea of
>doing unscheduled OS/MythTV reinstalls/restores, then it might be a good
>idea to put your OS on more than one drive too.
>
>
>1. ^^^^ I don't understand what you mean by unscheduled OS/MythTV 
>reinstalls/restores.

You can have as much RAID protection as you like, but if it doesn't 
protect your OS then the system will die if the drive containing the 
OS dies.


>3. It's looking like RAID6 is a better choice. For the cost I do not 
>mind losing some space for better protection, if that's how one can 
>deem it. I just used this RAID calculator: 
><http://www.icc-usa.com/raid-calculator.asp>http://www.icc-usa.com/raid-calculator.asp 
>and it shows that if I have 8x2TB for example in RAID6, I'd have 
>12GB usable. This is acceptable to me. My question is about having 
>extra drives to swap for failed ones: a) Must the drives all be 100% 
>exactly the same and b) How many would be advisable to have on hand? 
>(sticking with the 8x2 example).

You shouldn't need a RAID calculator - though of course if you aren't 
familiar with the technology then it's probably easier than 
remembering how to do it.
Striping - you just add up the drives (no redundancy)
Mirroring - you lose half your space (everything is written to two drives)
RAID 5 - you lose one drives worth of space out of your set
RAID 6 - you lose two drives worth of space.

As to sizes, this is an easy one to get caught by - and some hardware 
RAID controllers really don't help by only allowing you to use the 
whole disk. If you replace a disk, then the new disk must be NO 
SMALLER AT ALL than the smallest disk in the set (or allow you to 
create a partition which is no smaller than the smallest partition 
used in the set).
I have been caught by this in the past - drive fails in server, call 
up support company and they send out a new drive, but it won't work. 
Spend several days explaining to several different people that "yes 
it's a 9G drive, but it's smaller than the 9G drive that's failed, 
the replacement needs to be no smaller than <x> blocks". IIRC it took 
them about a week to source a replacement that wasn't slightly 
smaller - which was a "slightly worrying" time as there were errors 
starting to appear on several other drives and I think we ended up 
replacing about 4 before we'd done. I think we started doing mid-day 
backups of the data during that time as this server basically 
supported everything the company did (accounts, sales order 
processing, stock control, goods in, manufacturing processing, etc, 
etc, etc).

Whether to keep a spare is up to you - there isn't a stock answer. If 
you go with RAID6 then I don't think it would be worthwhile. With 
RAID5 it's a case of how long do you think it would take to get a new 
drive ? If you monitor drive errors and are able to see if one is 
starting to go, then you still probably don't need a spare.
On the other hand, if you buy 8 identical drives, in one go, from one 
supplier - then you stand a good chance of them coming from the same 
batch and that vastly increases the chances of multiple drive 
failures correlated in time. Think about it - you have a set of 
drives, from the same manufacturing batch, and then you run them for 
the same power on hours, at the same duty cycle - that has to 
increase the chances of a common defect getting more than one drive. 
 From that point of view, you would be better getting the drives one 
at a time, either from difference suppliers, or from one that has a 
good turnover - so as to maximise the chances of them coming from 
different batches. If you're paranoid, you'd mix manufacturers as 
well.
-- 
Simon Hobson

Visit http://www.magpiesnestpublishing.co.uk/ for books by acclaimed
author Gladys Hobson. Novels - poetry - short stories - ideal as
Christmas stocking fillers. Some available as e-books.