OneWorld Mirrored Installation

List-

Need some insight if anyone has had experience in this kind of situation. A client of mine has a fully live OneWorld installation up and running in location 1 with an enterprise server, deployment server, and four terminal servers/NFuse running in a farm. They have just purchased a company several states away, and are looking to implement some serious disaster recovery procedures involving the new location. They are thinking that they want to mirror the entire installation (all servers and database) at the new location, so that the operation could fail over quickly in the worst case scenario. This would basically involve almost constant replication between the installations. My first thought is that they could introduce a server cluster into the installation, thus mirroring all code and data on the servers as soon as it is written, but I haven't thought this through all the way. So, my questions are:

1) Has anyone attempted this kind of scenario?
2) Can server clusters (Win2K Advanced Server) be implemented over a WAN? Can the code and data be replicated this way?
3) Does the introduction of a cluster affect performance in any way?
4) Is there a better solution?

As always, thanks in advance for your thoughts.

OneWorld Xe SP 17.1, Update 4
Win2K Advanced Server/SQL 2000 Enterprise Edition
Metaframe 1.8, NFuse 1.61

Hi there :

Unfortunately, clustering on WAN links isn't a good idea.
Clusters require lots of bandwidth 24 hours a day, and most hardware
providers strongly recommend
you to have dedicated fiber LAN cards for heartbeat (the sync process among
its two nodes).
Thus, I conclude that WAN links won't be large enough for you.
For SQL databases, I would recommend you to use a log shipping strategy,
you've got a primary
DB server and you backup its databases. Then, you restore this tape on a
distant location.
After that, you periodically backup transaction logs and send them to the
alternate place,
let's say twice a day over the WAN link.
Of course, its technical viability depends on the amount of data you expect
and what your
WAN link is like!
Code (basically packages and ESUs) may be replicated automatically or
manually. Given that
code modifications may add up to several Gb (such as a full package), I
would sync them manually.
Clustering requires some CPU cycles, but I think that the overall result is
quite positive.
I've been testing B7332 clusters on NT+SQL7 and they run well, but don't
forget that its
reliability is limited by the intrinsical nature of Windows OS.

Sebastian Sajaroff

Hey jde_cnc,

We are currently implementing a solution very similar to this. We are literally right in the middle of this now so I will explain our intended architecture.

Production site has the following 1 x Enterprise\DB Server, 1 x Batch Server, 1 x Deployment Server, 3 x Citrix Servers.

DR site has 1 x Enterprise\DB Server, 1 x Tiered Deployment Server, 1 x Citrix Server.

Basic theory is that transaction log shipping is used ½ hourly for the JDE7333, JDE_PD7333 and JDE_PRODUCTION databases.

A tiered deployment server (at DR site) is added to the production deployment server. This effectively just acts as a file server, when a package is deployed at the production site the .cab files are deployed to the tiered deployment server. From here it is deployed to the DR site.

The Citrix server at the DR site is in a farm of it’s own, with access controlled by a Windows global group that nobody is normally a member of.

When (hopefully if) the DR solution ever needs to be invoked the following will take place.

1) A script will be run on the DR Enterprise server to change it’s name to the name of the production enterprise server.
2) Selected users will be added to the Windows Global group that controls access to the DR Citrix published application
3) Clients will need to re-discover the published app, this time it will find the DR site.
4) In our case we will need to change a couple of OCM mappings ie Our default UBE mapping that currently points to our batch server which will not exist in a DR environment. This will also be scripted.

Once we decide to go to the DR solution we are estimating we can have it running in 10 minutes or less.

While our DR site is connected to our WAN via normal methods we are going to be using dedicated ATM links to ship the transaction logs and deploy to the tiered dep server so as not to impact on the WAN

NB: Obviously there will be no development while in DR mode.

Anything else I can do to help just let me know. If anybody else has any comments or gotcha’s that we have not thought of please let me know.


Paul

Paul :

Be aware that when you change the machine name of your DR SQL machine,
you'll have to reboot the machine and SQL Server won't be able to
start its services unless you do some Registry tricking (or rerun
SQL CDROM Setup). You may also face some problems with DNS and WINS.
I would prefer to leave your DB Servers with different names and use
two OneWorld environments : one of them pointing to the main DB servers
(PD7333) and another one pointing to the secondary machine (DR7333).
When a disaster occurs, all would be instructed to log into DR7333,
and that environment would be pointing to your auxiliary server.

Sebastian

hmmmmm. so your datasources are set for the name of the Production Enterprise Server?



Matthew Scott
XE, SP 17.1, AS/400, Win2000 Logic Servers, Win2000 Term. Servers, Win2000 Java Servers, Central Objects in Oracle.

Re: RE: OneWorld Mirrored Installation

Sebastian-
Just to clarify---I understand your points about clustering over a WAN link, as I had these same feelings about it. But could you elaborate on your thoughts about the code replication? If you are not using clustering for this purpose, then how would the replication take place?

Thanks in advance.

OneWorld Xe SP 17.1, Update 4
Win2K Advanced Server/SQL 2000 Enterprise Edition
Metaframe 1.8, NFuse 1.61

Re: RE: OneWorld Mirrored Installation

Sebastian/Paul,

If you are running SQL Server 2000 and you rename the server it will pick up the new name upon reboot without all the stuffing around you used to have to do in earlier versions of SQL. Apparently.
I have never tested it...

OneWorld Xe Sp16_018, W2k, W2K TSE, SQL 2K on Compaq Proliant.

Re: RE: OneWorld Mirrored Installation

I am restarting the discussion for this thread as a need a little more information on mirrored installation.

As it was suggested by Sebastian Sajaroff that to create a seperate environment called DR7333 for the disaster recovery solution. But
How can I build a seperate enterprise server at DR site. do i need to
add a secondary enterprise server using add on servers plan. If so then in the plan which i create for add on server i cannot load the shared datasources for the new enterprise server.

My requirement is that when my primary enterprise server goes dowm i have to connect to the auxillary standby server at DR site and continue my business.

will you please suggest me how to add a new enterprise server in detail and how to setup a new environment called DR pointing to this server.

Sebastian Sajaroff, will you please me a detailed document so that i can go ahead ad setup the same.

Thanks in advance,
SP

Re: RE: OneWorld Mirrored Installation

I am listening in as well. I called JDE to get
standard practice for a High Availability site and
they said there was not one, that it handled case by
case. "Can I have services give you a call?"

I suppose it is good opportunity to bill. We have the
same delima.
--- sangameshwar_pa <[email protected]> wrote:
http://www.jdelist.com/cgi-bin/wwwthreads/showflat.pl?Cat=&Board=OW&Number=32134


=====
Dan Eppich
The Anschutz Corporation
B7332 Coexistant, V4R5 w/central objects
INS Card deployment server
SP16.1
Optio DCS 6.3.2
Fat and Citrix ICA Web Client

__________________________________________________

Hello jdecnc!
I will tell you what we got!
I work in an industry far away from the city, and we have implemented the following disater-recovery (backup)

The first reason we did not implemented cluster servers, was the virus problems, it already happens to us , so we eleiminated this problem in the followin way"

1. A entrire network was created with a Primary Domain Controller was installed with the same name as we have it in the production world, besides all the required servers are installed in this network, such aseployment server, enterprise servers, web servers and 4 data servers for the production environment (we have distributed de system and it works ok. Financial - Distribution - HR - Manufacturing) each module is located in a diferent server.
This servers were all installed exactly in the same hardware in which the productions servers are installed.
This backup network is completely disconnected from the real world.
What we do...
We clone the disks of the production servers to a bakup server (not to the other servers) with the Norton Ghost Utility, then we check the backup server for any virus. This server has to network cards when the virus checking has finished the network card for the backup environment is turnned off troguh a bat file and viceverse.
If you have any quaestions you can write me at [email protected]