[K12OSN] Data Storage - Redundancy and Backup Suggestions

[Jim Kronebusch]

On Thu, 3 May 2007 17:38:57 -0700 (PDT), Nick Fenger wrote
> Hello fellow K12ltsp'ers,
> 
> Currently, all of my student's files (and some of mine) are spinning on an old 
> 80GB drive in the smb-ldap PDC with no backup and no redundancy. I can barely 
> sleep at night.
> 
> We are starting a online student portfolio project next year that will add 
> lots of data that needs to be safely stored, backed up, etc...
> 
> Any suggestions for size, type of drives, should I stick them in the PDC and 
> just use software RAID or do I need some kind of Hardware solution? Should I 
> buy some kind of external drive array? How will I know if something goes 
> wrong? and what is the best way to backup all of this data?
> 
> We are a single site independent charter school with a maximum enrollment of 
> 320 students K-12.
> 
> Thanks in advance for your suggestions,

I am guessing that speed at this point isn't an issue if you are getting by with a
single 80GB IDE.  But if you are adding much more data in the future you may start to
see performance hits with a single IDE where you don't right now.  My suggestion would
be to add in a PCI RAID card to handle some redundancy and fail over.  I would think
that if you were happy with a single IDE in the past a SATA card would still meat your
budget needs and be fast enough.  If you have the money I would recommend SCSI, but that
would add some extra stress on the budget.  If you just want redundancy get a card that
can mirror (RAID 1), if you want to be safe on the speed side get a card that can handle
striped mirrors (RAID 0+1 or RAID 10, mirrored stripes or striped mirrors respectively).
 I am not sure how many SATA RAID cards handle RAID 0+1 or RAID 10 so that may put you
into the SCSI realm.  You could go with a RAID 5 but my experience is that the
redundancy is great, but the speed isn't quite there.  I recommend hardware RAID
whenever possible, I am not much of a fan of software RAID.  If your PDC is what
currently contains your /home directory, then I would put the drives in the PDC.  I
recommend putting your OS on a separate drive and running your /home alone on the RAID.
 You could easily move your data by adding the RAID Array and mounting it temporarily as
/newhome then rsync /home to /newhome, then remount the array as /home.  Very easy way
to add the new drives but not have to mess with your server at all.  Also then you can
enable disk quotas on the /home drive.

But first, BACKUP YOUR DATA!  You should not sleep at night until you get a backup!  I
hate to shout, but I see users lose their data every day.  Get a backup.  If you have to
move over a network share, plug in a temporary USB drive, clone with some sort of magic
device, whatever, get a backup.  After you have the backup, get a good nights sleep and
then mess with adding new drives.

As far as a permanent solution, either add in a large IDE drive, or get yourself a large
external USB drive and schedule some rsync's or whatever you desire.  I usually
recommend maintaining a few backup sets.  If you have the money get a drive large enough
to hold nightly incremental backups and a weekly full backup.  If you can, store a
separate nightly backup for each night of the week.  Backups aren't always for hardware
failure.  Mine get the most use via user mistakes.  On our main school server I store a
full month of backups by day.  Then when a student or teacher realizes 7 days from now
that they accidentally deleted some project that took them 1 month to create, I can go
back 7 days and get it for them.  If you only have one backup set, you lose that
ability.  The more backup sets, the farther back you can go.

Hope that helps.
Jim



-- 
This message has been scanned for viruses and
dangerous content by the Cotter Technology 
Department, and is believed to be clean.