Topic: replicate, mirror, clone

[.nate]

What is the "best" way to replicate, mirror, clone, whatever, an e-smith server and gateway box that your whole company depends on every second of every working day?
 
Ultimately I need to be able to replicate the whole box exactly. I have SME implemented in a small company as a gateway, PDC, e-mail server, web server, etc., etc., etc…. If the box goes off-line, or even hiccups, I have people running into my office like a swarm of angry bees looking for blood!
 
In need of easy redundancy.
 
Thanks,
 
.nate

[Dave Liquorice]

> What is the "best" way to replicate, mirror, clone, whatever, an
> e-smith server and gateway box that your whole company
> depends on every second of every working day?

So it doesn't matter if the box dies outside of the working day taking all your data with it? Strange...  B-)

How much money have you got?

A UPS would be a good start to remove the reliance on the power company. Followed closely by some form of RAID, probably a multi drive array with hot swap of failed drives and automatic rebuilding of the array when a drive is changed, not forgetting a stock of (tested) drives sitting on the shelf.

A second machine sitting on the network that accepts regular backups from the main machine, also with hot swap auto rebuilding RAID. Both machines fitted with tape backup devices that are used regulary in a proper rotation and stored *of* site. Maybe even backup to a machine across the internet to get the data off site.

Have more than one machine, each one preforming a seperate task. One for the web, one for mail, one for samba etc etc. All machines identical as far as the hardware is concerned so you can "borrow" bits if you get a serious failure and you haven't got spares on the shelf.

Don't forget to have spare switches on the shelf and cabling, connectors and required tools. Trace the route a  packet would take from a workstation to server/internet and if you haven't got a complete spare (don't foget the PSUs) for anything in that path get at least one to act as a spare.

My thoughts, there is a High Reliabilty FAQ or document somewhere but I have not read that.

Cheers
Dave.

[Kelvin]

Nate,

http://www.star-support.com/downloads/mitel/contrib/Linux-HA/SME%20High%20Availability%20How-To.html

Kelvin

[nate]

“Strange... B-)” ?  ...as in blood brother?
 
Anyway, I am concerned with the time it would take me to bring a new machine online, restore data, etc.  If this happened off-hours I could get things going again by the time people came back to work.  However, if it were to crash at 9:35 am on a busy Tuesday morning, my life would be hell until the network was back up.  
 
I would not loose any data as I have redundant cron job backups to DLT, and a backup server w/SCSI RAID5 hot-swap,  ....and Yes, I have extra drives on the shelf.
 
"UPS - remove the reliance on the power company"  - Well yes, you could say we have that part covered.  We are a systems integrator of solar electric power systems and we have the server room powered by three independent power sources, which provide clean, perfect sign-wave, ultra redundant power!
 
As far as splitting up the services up among multiple boxes - the whole point of SME is a simple, all in one, low cost, powerful solution for a small company.  Each individual service does not get hammered very hard resource wise, certainly not enough to warrant multiple boxes, etc..
 
Anyway, I either need an active mirror or a machine on the shelf that I can grab, hook up the KVM, Net, and Power, …hit the on switch and I’m done.
 
.nate

[Brian Hand]

I had a 2 scsi mirrored system with a third ide drive installed in the system which is mounted and rsync'd hourly or daily, etc. I took it out as mitel recon it is an unsupported configuration, but it ran fine for 6 months. I had problems (still ongoing) since upgrading to update 6 of 5.5.

I now have an old (second) pc (pentium 166) with a small 1gb drive running e-smith and the IDE drive setup in it for the same purpose. whenever I have a problem, I do a clean install on the main server and rsync back - about 10 mins and all running ok.

A bit basic, but it works.

[dave]

Nate,

Really, the biggest question that comes to mind is how often does your data change?  This will tell you how often you need to backup.  

Rsync would be an ideal tool, you can use cron to build hourly backups of really key data through the day.  Another cron job each evening would get everything.

The quantity of data will be another factor.  The more data, the longer the backup will take and the greater impact on server/network performance.  

Probably the simplest recovery would be to create an image of the SME system volume and store it on the network.  Norton Ghost should do the trick there.  In the event of a major failure, replace the server (if you're using ghost, the replacement should be identical to the original).  Restore the image file and you're back up as an internet gateway in the amount of time it takes to plug in the backup server and Ghost to copy the file down.  Use rsync to restore the balance of the user files once the initial image is restored.  You should also perform full backups to tape as frequently as is reasonable, store a full set of tapes offsite too.

Dave

[.nate]

Thank you all so much for the great information.
I really love the Linux community.
I will post a schema of my final solution when it is all up and running.
 
.nate

[John Sequeira]

I'm not sure if this is what you're looking for,  but g4u is a free disk imager that totally rocks:  
http://www.feyrer.de/g4u/

It's a simple front end to the even simpler dd and netcat commands.

Here's some other links I came up with.

Dolly
http://www.cs.inf.ethz.ch/CoPs/patagonia/

cloneit
http://www.ferzkopp.net/Software/CloneIt/CloneIt.html

and also partimage (link?)

Cloning on ASK Slashdot

http://ask.slashdot.org/article.pl/06/26/1912256&tid=

http://slashdot.org/article.pl/06/27/1934226

Cloning dd, netcat
http://www.rajeevnet.com/hacks_hints/os_clone/os_cloning.html