I´m thinking about different ways how to handle Server Name and IP-Address for PI Servers. The current reason is beginning of planning for my next upgrade (hardware, OS and PI applications) of my PI Servers. The background is to minimise the Impact on data collection and data retrieval while an upgrade happens. In my case there is no collective so I do not take this into account.
Due to my knowledge there are two common ways to do an upgrade:
1) Setting up new Servers with new Server Names and IP Adress. In the first step the same versions of the PI applications like on the legacy Servesr will be installed. After data move the applcaitons will be updated. After this the Interface nodes Need to be reconfigured to send the data to the new PI Data Archive Server. Clients also need to be reconfigured to retrieve the data from the new Servers (from Data Archive as well as from AF-Server).
2) Shutting down the old Servers and installing the new Servers with the same Server Names and same IP-Adresses. In the first step the same versions of PI applications like on the legacy Servers will be installed. After data move the applications will be updated. In this case the clients and Interface nodes do not need to be reconfigured.
So far nothing new. there are some points to be considered for these ways:
1) The buffer queues on Interface nodes are dedicated for an PI Data Archive Server. If you change the PI Server for buffering a new buffer queue is beeing generated. The data insinde the old buffer queue will not be delivered to the new PI Server. This means that you have to carefully plan the configuration switch to avoid data loss. If you have redundand Interfaces it is more complicated.
2) While installing and the update process there is no access to the data. In Addition you have to make sure that the new Server will have the same PI Server ID as the old Server had. It could also be an problem with Network routing or other IT related topics when there are different servers with the same name and IP Adress (e.g. it is not possible to have the same name and IP-Adress in the Network).
The last times we did an upgrade we used way 1. Im currently trying to evaluate a new way. Therfor I hope for hints and comments from you:
My idea is to create an DNS alias for my PI Servers. The alias would be pointing to the current Server name. In case of an upgrade I would prepare the new Server. When it is finished I would change the DNS Alias. So no configuration chage should be nessesary on clients and Interface nodes.
I need more information about the PI Server IDs in this Scenario. There is some additional study of the administration guides nessesary.
Are there any hints or comments for this Scenario? Anything that I did not take into acount? Any experiances with this apporach? How do you handle this when you do an upgrade?