Guidance on Migrating activity for version 14.4

[Atifah Abdul Latif]

Hi All,

Based on techdocs, it is recommended to gradually remove old nodes (CentOS 6) after added the new nodes (CentOS 8). If customer want to use the same IP from the old nodes to the new one, is that possible? or is it not a best practice considering the connection of application-to-application.

Thanks,

Atifah

[Rinat Matityahu]

Hi Atifah

With gradual migration, for a certain amount of time you have both Centos6 and Centos8 machines running together in the same cluster. Since they cannot have the same IP while running together and keeping in mind it is not possible to change IP address of any appliance once it has been sent once, not certain how you can achieve reuse of IP addresses.

Regards

Rinat

[Adrian Bullock]

We reused our IP addresses when upgrading from CentOS6 to CentOS8 vapps using gradual migration.

In our 2 node Development environment this meant instead of having one application node and one provisioning node, we changed to having all services deployed on both nodes.

We then chose one CentOS 6 node (old node 1) and powered it down and spun up a new CentOS 8 node with the same IP address (new node 1). This new node was added using the dashboard on the existing old node 2 (CentOS 6) and thus got updated with all configuration information.

This now gives a mixed env with one CentOS 6 node and one CentOS 8 node - same IP addresses in use as when starting.

We then powered down old node 2 (CentOS 6) and spun up new node 2 (CentOS 8) with the same IP old node 2 had, and added it to the configuration using new node 1 dashboard.

This now gives a fully CentOS 8 config.

One thing we did discover when upgrading Test and Prod (we had 4 node clusters of vapps here) was that when CentOS6 and CentOS8 vapps ran in the same cluster they generated separate cluster configs that did not talk to each other.  This was back in 14.4.0 so may have been fixed by now, but it meant that when we did the Prod upgrade we did not run mixed CentOS6 and CentOS8 there, rather upgraded one leg to CentOS8 and after adding it to the cluster using the CentOS 6 node we stopped the CentOS 6 nodes and ran Prod with a single CentOS 8 leg for a week (until we upgraded the remaining nodes to CentOS8 and added them to the config).

But this relies on having multiple vapps in an environment and upgrading them one at a time.

Adrian

[Rinat Matityahu]

Good feedback Adrian, thanks for that!

Clever way to workaround the IP limit.

[William Cheang]

Hi Adrian,

I just want to confirm following "This new node was added using the dashboard on the existing old node 2 (CentOS 6) and thus got updated with all configuration information."

• Since the old node1 is already added into the cluster(with old node2), when u "switch"/powerup to CentOS8 node into this cluster, after that do u perform redeployment using Dashboard-Setup ? 

regards,

William

[Adrian Bullock]

Yes William that is correct.  After adding the new node via the dashboard the services need to be deployed (time to get a cup of coffee!).

Time for me to update that set of instructions for upgrading from 14.4 CentOS8 to 14.5 CentOS9 :)

[William Cheang]

Hi Adrian,

Thanks for the tips, I guess we can use this method to upgrade 14.5(CentOS8) to 14.5(Centos9) without using the "Gradual Migration" method.

[William Cheang]

My lab IGA cluster env

node1-14.4.1 Centos8 (1st installed node)

node2-14.4.1 Centos8 (2nd installed node)

 

Then i perform in-place upgrade for node1,so node1 become 14.5

node1-14.5 Centos8

node2-14.4.1 Centos9

 

Then i shutdown node2

I deployed a new VM using 14.5 OVA file, configured with same IP as node2, 

node1-14.5 Centos8

node2-14.5 Centos9

-> Then goto Setup and click Deploy (this will redeploy all the services again)

-> Deployment was successful without any errors.

-> Restarted node2, all services is up.

 

But when upgrade node1 with the same method, then the deployment hang there...

it seems like this method only works for second node.