Synchronizing 2 unix servers

[thewho?]

Hello,
HP-UX 11.0, Trusted systems, N4000
I'm researching several options (NIS+ and LDAP). I don't think I'll go thru NIS+ because of the overhead that it will create, specially for only 2 servers. LDAP looks more feasible but...
Is there a simpler way? Here's the scenario, one of the boxes is my production box, in the event of a failure of this box I want to be able to switch to the development box (by changing its IP address). I already took care of the synchronization of the databases, what I'm trying to figure out is how to keep the passwd, host, etc in the development box up to date.
Any ideas will be deeply appreciated.
I thought of using incremental fbackups and send them thru the network, Is this feasible for a trusted system?
Thanks in Advance,
Luis

[linuxfan]

Hi Luis,

As far as syncing up the files are concerned,
you could probably write a script that ftps a few files over to the other machine and run this script through cron or you could use something like rdist. I am not sure if rdist works between trusted systems.

The source code for rdist can be obtained from
http://www.magnicomp.com/rdist
or from the HP's porting center
http://hpux.cs.utah.edu/hppd/hpux/Networking/Admin/rdist-6.1.5

-HTH
Ramesh

[Sridhar Bhaskarla]

LUIS,

Did you think of MC/Service Guard?.

-Sri

[Deshpande Prashant]

Hi
MC Service Guard will help you do that. This package is an extra cost though.

Prashant.

[Rainer von Bongartz]

No, no
Service Guard will not solve all problems. For sure it is the way you should go to build a high available database system, but it will not free you from the mentioned tasks of keeping /etc/passwd, /etc/groups /etc/hosts and others consistent on multiple hosts.
But you should not think of copying thgos files to different machines and hope to have a replicated evironement. You will fail.
As mentioned keep looking for a central repository for theese files and I really would recommend LDAP. You can install Netscape Directory Server and HP ldap-ux services (both available for free) and build the repository around this.

It is not so complex as you might think but it will save you a lot of problems for the future

Regards
Rainer


on

[Timothy Czarnik]

Luis,

You may want to look at the rdist command as well. You can build a list of files that rdist will automatically verify are the same from one server to the other. If they are not, it will update the out of date file. This not only works for files, but for directories as well (/home, /database, etc...).

Another more drastic alternative is to use a make_recovery tape of the production server on the standby server. ALL your hardware would have to be using the same SCSI ID's on each server, so you may even have to swing cables in the event of a production failure, but its doable.

My $.02

-Tim

[Timothy Czarnik]

Luis,

Your 7 points gave me my first hat! Thanks!

-Tim

[thewho?]

Congratulations Timothy!!!

[Brian Markus]

We have a something setup that sounds like what your looking for. It was a bit expensive, but you could scale it down for your environment. I have an XP256 Disk Array. I have my Disk system shared out to both systems. So it's one phical volume to each logical volume on the systesm. (same disks two systems) I use MC/Service guard to switch back and forth. You can pick and choose which mount points are 100% identical per node in the cluster.

Hope this helps

Brian