samba - Mar 2005 - A probably silly thought but...

Not a silly thought.    

I've instead of spending x hunderd thousand on brandname ibm servers and
sans, how you could create a RAID array of PCs running IDE hard drives.
If a PC dies just plug in some more and rebuild.   If you want to add
more space just add more PCs.  :)      that would be cool.

Obviously you wouldn't use it still for critical data like databases
etc, but our user directory would be a good candidate.     

-----Original Message-----
From: samba-bounces+mwestern=sola.com.au@lists.samba.org
[mailto:samba-bounces+mwestern=sola.com.au@lists.samba.org] On Behalf Of
S Clark
Sent: Thursday, 3 March 2005 8:09 AM
To: samba@lists.samba.org
Subject: [Samba] A probably silly thought but...

Here's a completely off-the-wall thought.

I have no idea if it would work, but would it be possible for you to
combine the two RAID5 systems into a "software" RAID1 system?  Perhaps
via iSCSI?

Kind of an overcomplicated method for doing what you need, but it WOULD
keep both sets of RAID5 arrays in sync with each other in realtime, and
would allow things to keep running without a pause even if one of the
RAID5 systems failed.

If it's even possible...

(Told you it was a silly thought...)


On Wednesday 02 March 2005 09:19 am, Greg Freemyer
wrote:
> On Wed, 02 Mar 2005 10:33:34 +0100, Michael Gasch wrote:
> > hi,
> >
> > well, i was also wondering how to build up a very redundant solution
> > for my samba installations
> >
> > at the moment i'm using rsync twice a day to sync about 2TB amount
> > of data between two hardware raids (both raid5 with 2 hot spare)
> >
> > advantage: if filesystem is corrupt on one raid, the other raid is 
> > normaly not affected
> >
> > disadvantage: because analyzing data to sync by rsync takes time 
> > it's senseless to sync every our so you have no realtime backup 
> > (only 12h
> > before)
> >
> > how do you avoid this filesystem issue with drbd? doing rsync every 
> > night seperatly? i don't know of statistics about filesystem
damages
> >
> > cheerz
>
> DRBD would not help this problem.  As you say the filesystem 
> corruption would immediately be duplicated to the alternate server.
>
> OTOH a good journelled filesystem combined with dual-power supplies 
> and dual ups's should have a very high relaibility rate.  EXT3 seems 
> to get mentioned as the most reliable linux filesystem, so go with 
> that if reliability is your top concern.
>
> Greg
> --
> Greg Freemyer
--
To unsubscribe from this list go to the following URL and read the
instructions:  https://lists.samba.org/mailman/listinfo/samba

On Friday 04 March 2005 00:30, Matthew Western, IT Support, Lonsdale
wrote:
> Not a silly thought.    
> 
> I've instead of spending x hunderd thousand on brandname ibm servers
and
> sans, how you could create a RAID array of PCs running IDE hard drives.
> If a PC dies just plug in some more and rebuild.   If you want to add
> more space just add more PCs.  :)      that would be cool.
Network block device + dm.
 
> Obviously you wouldn't use it still for critical data like databases
> etc, but our user directory would be a good candidate.     
> 
> -----Original Message-----
> [snip]
> > > at the moment i'm using rsync twice a day to sync about 2TB
amount
> > > of data between two hardware raids (both raid5 with 2 hot spare)
> > >
> > > advantage: if filesystem is corrupt on one raid, the other raid
is
> > > normaly not affected
This is the recipe for high reliability:
* RAID5 protects you from hardware failure
* backup protects you from human error/kernel bugs

One of your RAIDs is effectively a backup.
You don't really need to store it on the RAID tho,
unless you want to be protected from *simultaneous* hw failure
on both machines. Very unlikely. You can keep backup on plain ol' disk.
--
vda