Lan server disaster recovery

I would like to build up a lan server disaster risk free, i'm considering
the idea to buy 2 identic pc and setup one with 2 hd in raid 1 so when an hd
or any other hardware component blow up i can move the working hd to the
second pc and have the server up in a few minutes. This also could leave me
from the problem of do daily backups.
I only have 2 dubts:

1 - When i put the full hd (with linux configurated and all the progs and
files of the server) as master in the second pc with a blank hd as slave,
all the contens will be auto copied to the 2nd hd recovering the raid
situation on the 1st pc or not?

2 - Is this a good solution or are there others less expensive/complicated?

Thanks.

Rasta99 <(E-Mail Removed)> wrote:
> the idea to buy 2 identic pc and setup one with 2 hd in raid 1 so when
> an hd or any other hardware component blow up

If you use raid 5 you can keep running with one disk down, if you use
hot-swappable hardware (SCSI) you can even swap the disk while the
machine is running. But _nothing_ will save you if your machine burns
down or is stolen (it happened to me).

> This also could leave me from the problem of do daily backups.

No. It won't. If the controller goes yippy-dippy and screw both the
disk you need a backup. Also, see the previous 'burns-down' scenario.

> 2 - Is this a good solution or are there others less expensive/complicated?

Buy _good_ hardware, have spare parts handy when is necessary, implement
a good backup strategy, test your backups. Cheap - Reliable. Pick one.

Davide

--
| Genetics explains why you look like your father, and if you don't, why
| you should.
|
|

Davide Bianchi <(E-Mail Removed)> said:
>Rasta99 <(E-Mail Removed)> wrote:
>> the idea to buy 2 identic pc and setup one with 2 hd in raid 1 so when
>> an hd or any other hardware component blow up
....
>> This also could leave me from the problem of do daily backups.
>
>No. It won't. If the controller goes yippy-dippy and screw both the
>disk you need a backup. Also, see the previous 'burns-down' scenario.

Also, no amount of mirroring will save from administrative errors
("What was that file I just removed?", or "Now, where did I run
that 'rm -rf'?"). Of course, if there's another way to rebuild
the server from scratch (in some desired timeframe), then backup will
not be needed.
--
Wolf a.k.a. Juha Laiho Espoo, Finland
(GC 3.0) GIT d- s+: a C++ ULSH++++$ P++@ L+++ E- W+$@ N++ !K w !O !M V
PS(+) PE Y+ PGP(+) t- 5 !X R !tv b+ !DI D G e+ h---- r+++ y++++
"...cancel my subscription to the resurrection!" (Jim Morrison)

Rasta99 wrote:

> This also could leave me from the problem of do daily backups.

Maybe get a PC with the discs, RAID if you like, and use rsync to automate
the backups?

Backups don't just cover against fire/theft/physical damage. Have you
never accidentally deleted a file? I've run rm -rf * in / before, thinking
I was in another folder. Probably everyone does something like this once.

What if a file is corrupted, then the backup overwrites the original? You
might not notice for weeks.

--
Matt


-----= Posted via Newsfeeds.Com, Uncensored Usenet News =-----
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
-----== Over 100,000 Newsgroups - 19 Different Servers! =-----

On Thu, 15 Jul 2004 16:22:03 +0000, Juha Laiho wrote:
> Davide Bianchi <(E-Mail Removed)> said:
>>Rasta99 <(E-Mail Removed)> wrote:
>>> the idea to buy 2 identic pc and setup one with 2 hd in raid 1 so when
>>> an hd or any other hardware component blow up
> ...
>>> This also could leave me from the problem of do daily backups.
>>
>>No. It won't. If the controller goes yippy-dippy and screw both the
>>disk you need a backup. Also, see the previous 'burns-down' scenario.
>
> Also, no amount of mirroring will save from administrative errors
> ("What was that file I just removed?", or "Now, where did I run
> that 'rm -rf'?"). Of course, if there's another way to rebuild
> the server from scratch (in some desired timeframe), then backup will
> not be needed.

Yes, that's true. I suspect that most of the use of backup/restore is
actually to take care of this kind of "human error"?

OTOH, leaving the human error aside for the moment, one might argue that
using RAID could justify a change in the frequency of backups. Perhaps
instead of doing incrementals daily, you could do them weekly? Or maybe
keep daily incrementals (might be small anyway) but stretch out the
schedule of full backups? The rationale here would be based on the fact
(? observation?) that most H/W failures causing loss of data are actually
failed disks. RAID mitigates against these kinds of data loss due to disk
failures. Hence, the probability of losing data due to failed disk is much
less. Therefore, if the previous backup schedule was considered adequate
(for probability of loss of data), then the new schedule could be
"stretched" (for the same probability of loss of data, or rather risk of
loss of data, or probable amount of effort to restore data, etc.). I find
I'm implicitly adopting this kind of policy, after having implemented
RAID5 on my code repository. I'm intending to do daily incrementals (not
doing them that often now), but I'll stretch out full backup schedule.

--
Juhan Leemet
Logicognosis, Inc.

Thanks to all, you're right i still need to do backups.
Anyone could help about what i asked in point 1?

@Maro Bianchi
I used wrong words, i didn't mean blow up seriously, i mean every hardware
fail, hope no one hate me so much to place a bomb in my pc...
A saperlo scrivevo in italiano che almeno lo parlo meglio

Rasta99 <(E-Mail Removed)> wrote:
> Anyone could help about what i asked in point 1?

It depends by your RAID configuration, in hardware raid the disk will
be resilvered and initialized immediately, in software raid, the problem
is that you need a disk from which to boot, then you can add the other
disk to the raid and have it reinitialized.

> A saperlo scrivevo in italiano che almeno lo parlo meglio

There are many it.comp.os.linux groups then.

Davide


--
| Diplomacy is the art of saying "nice doggy" until you can find a rock.
|
|
|

I will go for the software raid, do you mean is enought a simple boot disk
(like a floppy) or a disk containing the whole OS?

> It depends by your RAID configuration, in hardware raid the disk will
> be resilvered and initialized immediately, in software raid, the problem
> is that you need a disk from which to boot, then you can add the other
> disk to the raid and have it reinitialized.

On Fri, 16 Jul 2004 08:31:29 +0000, Davide Bianchi wrote:
> Rasta99 <(E-Mail Removed)> wrote:
>> Anyone could help about what i asked in point 1?
>
> It depends by your RAID configuration, in hardware raid the disk will
> be resilvered and initialized immediately, in software raid, the problem
> is that you need a disk from which to boot, then you can add the other
> disk to the raid and have it reinitialized.

I haven't setup "root mirroring" on my Linux machines yet, but I have done
it on a couple of Sun Solaris machines. I believe the Linux approach is
very similar. You cannot boot off S/W RAID5 because you have a
"chicken/egg" kind of problem: to read the S/W RAID5 you need a working
kernel with RAID drivers, but you haven't loaded them yet. With mirroring
it can work: you boot off 1/2 of the mirror (or if that fails you boot off
the other 1/2) and then when the kernel and drivers are loaded, the system
starts working with the full mirror (RAID1).

You were right, but you should reread your man pages to make sure of the
commands. Typically you define the metadatabase (which records the RAID
definitions), then you define the mirroring and allocate the 1/2 half of
the mirror. I think you might have to reboot at this point, so that you
are using the md metadevices (even though you only have 1/2 mirror setup).
Then when you are running with the RAID (1/2 working), when you attach the
2nd half of the RAID1 mirror, the RAID software should replicate the 1st
half onto the 2nd half, and then keep them in sync after that.

If one or the other 1/2 mirror fails, the system can still run (and boot),
but you should hurry up to replace that disk, else you're vulnerable. If
you lose the other 1/2 before you have replicated it, obviously, you lose.

Of course, one could go to more elaborate setups: RAID1 on boot disks, and
then RAID5 for data, as an example. You can also combine striping and
mirroring (RAID0+1 or RAID1+0, but I get confused as to distinction).
BTW, remember to mirror your swap too, if you want to keep running (and
not panic). Read the man pages & docs. It's not that hard to setup.

--
Juhan Leemet
Logicognosis, Inc.

"Rasta99" <(E-Mail Removed)> writes:

> I would like to build up a lan server disaster risk free, i'm considering
> the idea to buy 2 identic pc and setup one with 2 hd in raid 1 so when an hd
> or any other hardware component blow up i can move the working hd to the
> second pc and have the server up in a few minutes. This also could leave me
> from the problem of do daily backups.
> I only have 2 dubts:
>
> 1 - When i put the full hd (with linux configurated and all the progs and
> files of the server) as master in the second pc with a blank hd as slave,
> all the contens will be auto copied to the 2nd hd recovering the raid
> situation on the 1st pc or not?
>
> 2 - Is this a good solution or are there others less expensive/complicated?

I would recomend two computers, configured so that you have the data on a
separate disk/partition than the OS (in fact if I have the space, I always make
2 root partitions on separate disks, which allows you to experiment with
alternate distributions/releases while keeping your primary system so you can
go back to it). No user data is to be stored on the OS partitions, so that you
can blow it away, and install a new OS if needed/desired. If you want maximum
reliability, put the user disks in an appropriate RAID configuration. The
second machine is similarly configured, but you can save some $$$ by not making
the disks RAID. You use rsync or a similar tool to backup the data to your
hot-server and then do backups from that system. I use this system for my
photo files, and find it useful if I deleted a file on the main server, to
recover it via rsync. Since I control when the copy to the hot backup is done,
I do it when things are in a consistant state.

--
Michael Meissner
email: (E-Mail Removed)
http://www.the-meissners.org