1 of 1 people found this helpful
Victor it is not possible to replicate just one database. PI AF Collectives use transnational replication to replicate the PIFD database. A workaround, would be to install a second AF Server and use a different SQL Server instance, and only host one database.
There is also the PI System Connector, but it really is intended for a different purpose.
The name of the database also suggests that you are running calculations in AF. Please note that the secondary member of an AF Collective is read only due to the way SQL Transactional Replication works. As Gregor Beck mentioned, OSIsoft does not recommend AF Collectives as an AF HA option.
Out of curiosity, what is your use case for AF HA? I think you may be better served by a different HA setup.
Thank you Sebastien,
I work in an ISO and we are using PI Analysis for real time calculations. Since those calculations are critical, we managed a workaround to keep it running initerruptly.
I created a Powershell script that follow the steps below:
1. I export an xml file to a second AF server daily;
2. In this second AF server I have a database with the same name of the first "CalculosTR", I UNDO CHECK-OUT in this second database;
3. I import the xml file created from the first server to the second one, but I do not check-in, because if it was done, there would be two different analysis writing in the same tags.
4. In case of problems with server1, it is possible to check-in the changes in server2 and the analysis will keep running.
In case of problems I have to check-in manually the second server;
I am copying data from the first server to the second once a day.
I am trying to find a more suitable solution for this tine but very important database.
PS: I am sorry for the time taken to reply, I was in a trip for business and had limitations to access internet.
There is no out of the box supported way of having AF HA for a single AF database. Our recommended AF HA architecture is an NLB front end for the PI AF Application service, Always On SQL for the back end and Analysis only supports Windows Fail over Clustering.
You could replace your manual check in operation with an automated programmatic check in using the AF SDK: https://techsupport.osisoft.com/Documentation/PI-AF-SDK/html/1a02af4c-1bec-4804-a9ef-3c7300f5e2fc.htm
That being said, I'd recommend using a supported HA mechanism. If you have your Analysis and PI AF Applications Service on the same machines you could use Windows Fail over clustering as an HA solution for the front end.
That's ok, at least I tried...
Thank you so much for clarifying, Sebastien.
I will strive to expand our hw destined to PI System and mount a decent HA mechanism.