Originally posted by: Louis_Hicks

All,

I am currently involved in a project to transfer a number of workloads from a struggling AIX 5.3 machine to a new machine with the same OS. This is required in order to accommodate expansion, and to allow the deployment of new HA & DR software.

Due to the importance of the project we are looking at outsourcing for the skills required to achieve this goal. However, I will be required to carry out such activities in future, and therefore am trying to train up on this type of exercise. I am effectively a one man team on this, as the company I work for is just moving into pSeries. I would really appreciate any help you can offer even if it is to direct me to a document on how to approach the tasks.

There are the obvious things which I have thought about; to first create a system back-up, as well as a database back-up, and to analyse the configuration of the current system.

This where my lack of experience fails me, I am learning quickly, but have never been required to migrate a workload.

My initial thoughts where that I could create a mksysb, restore it to the new system, vary off the old volume group and then bring it up on the new system by transfering the disks. However, I am told that the current solution is on local storage.

Again, any help would be appreciated

Thanks
#AIX-Forum

Originally posted by: alethad

I would suggest looking for the document on cloning a system from a mksysb. But applications are generally NOT installed in the root VG nor would you want to install them in the root VG.

So you may have to re-install the application software OR if the software is contained completely in a single filesystem(with few or no extraneous pieces in /usr and you're not using raw LV's for the database) it would be alot easier to build the filesystem structure on the new system and just rcp the files over.

We use HACMP in our shop. Just 2 node clusters so far. Here's a link to get you started on that but I would HIGHLY recommend at least getting into the install and planning class if you can. The class is really worth it. If not the main thing you really need to be annal about is not letting anyone make changes to the system without your knowledge. I have the problem with vendors coming in and upgrading, adding disks or making changes to filesystems/VG's and then I don't get told. Therefore the changes aren't included in HACMP config and when the system fails over to the other node the application doesn't function properly. Really frustrating for an admin if you know what I mean. The class is really worth it.

http://www.redbooks.ibm.com/redbooks/pdfs/sg246769.pdf

Is your new system local disks or do you have it on a SAN? We used to be on local disks but we have IBM SAN here now.

Good luck.
#AIX-Forum

Originally posted by: Louis_Hicks

Thanks for the information
I am gathering now that it may be best to simply rebuild the configuration on the new hardware, including apps and the DB, then restore the DB file to the new DB instance.

Once the data is verified manually change the IP addresses and do some testing. I will hopefully get some more input from other admins too. I know what you mean about people changing things and not filtering it to people. All too well.....

Thanks again
#AIX-Forum

Originally posted by: alethad

I'm sure there's others that may have better suggestions.
But yeah, that's probably what you'll wind up doing. I have found it the easiest and less painful. But then all my apps were all contained in filesystems that were easy to re-create and port over. No raw LV's.

Well then go ahead and breeze thru the HACMP cookbook and get a feel for that at least. You're going to need to plan out the topology and the start & stop scripts for the apps really well. Hopefully you'll get the systems before the users so you can test your failover. Really painful if you don't. :)
#AIX-Forum

Originally posted by: Louis_Hicks

I'm not actually using HACMP for this solution. I am going to be using a Vision Solutions HA & DR offering that we are currently testing and developing. It looks like a good system and I have been working with it for about four months.

I have decided to speak with the DBA's at the customer site and ask how they would carry out a failover using their current DR. Then I will know how problematic restoring to DB etc may be once the system is built. Hopefully it will be a matter of then restoring the data. We will have the old system running in the event it all goes pear shaped, so I'm not too worried yet.

Thanks

Louis
#AIX-Forum