We use Baan IVc3 on Informix with AIX Version 4.3.3. IBM has come out with a disaster recovery system called HACMP (High Availability Cluster Multiprocessing) for AIX which automatically detects system or network failures and eliminates a single point-of-failure by managing failover to a recovery processor with a minimal loss of end-user time using Cluster technology.

In essence, the entire application can be transferred over to another server from the initial server on which it was mounted with minimal outage period.

My question is, can Baan make or be made to use this Clustering feature? Remember that Baan makes use of the .brandfile which includes server name as one of its entries hence making it difficult to “Validate” a new server during the transfer process of HACMP….can we apply for a “network license” from Baan and give the other (recovery) server as the secondary server to make this procedure work?

Any help in this regard will be highly appreciated.

Thanks.

My company (OST) has customers running HA implementations of Baan with Oracle or Informix on all of the major UNIX vendor's HA implementations. HACMP (or the HACMP/ES enhanced version) can be used for Baan failover. There are of course a number of issues that must be addressed at the hardware, database and Baan application level especially if you have third-party applications or EDI in use within the environment.

The licensing is done with a combination of the network and HA license. Since the .brand file is on the root file system, this does not failover so each system running the license daemon is independently licensed.

Feel free to email me for specific information. I will be happy to direct you to Baan HA reference sites for you to talk to.

-- Jim

As Jim said, running Baan on an HA cluster is done by several Baan customers. But you should get yourself a consultant, who knows, what he is doing.

My experience is all with HP systems, but I guess the following is valid for IBM as well:

1. You need to take care with the validation of your Baan system. Basically you need to validate both servers, so that the licd can run on both. (You will get two sequence numbers from Baan for this)

2. The annoying thing of licd is, that it wil half the licences, if it notices that a second licd is running. To avoid this you have to make sure, that you get the "networking" and "high availibility" option in the valiadation process.

3. Sometimes Baan says, they generated a key with these two options, but they didn't do so correctly. To check type "licmon6.1 -B" after validation. You should see a "long" option key, let's say more then 8 digits. Normally then it is ok.

4. Try to configure your Cluster as "defensive" as possible. Do not switch a package from one node to the other, just because a process does not seem to respond etc. In our experience it is best to restrict the monitoring to pure HW and network. The Baan Background processes shmtimer6.1/licd6.1 hardly ever fail at all and you shouldn't use pdaemon in the first place in my opinion.

Regards

Markus