Is it possible to configure the AF analysis failover in PI AF 2017 R2?
Yes as of AF Analytics 2016 R2 it is possible to utilize Windows Failover Cluster technology for AF Analysis failover.
Thanks for your reply.
But we are using PI AF 2017 R2 version is it possible to configure failover and do we have any known issues?
Let me clarify why we want to configure failover, We are going to upgrade OS for Analysis service hosted server and AF hosted server during Upgrade there will be Outage to avoid this we want to configure failover( yes data will be auto back filled once servers are up)
Do you want to permanently configure Asset Analytics in an HA configuration or are you simply looking for a way to avoid any downtime for this OS upgrade on your analysis and AF Server ?
Something to note, while it is possible to avoid READ downtime for AF by setting up a parallel system, it is impossible to avoid any WRITE downtime as the database schema must be upgraded. Consequently, even if you setup a fail over system for Asset Analytics, due to the AF upgrade there will be some unavoidable downtime.
Please submit your feedback for multiple instances of AF analysis below
Multiple instances for Asset Analytics – User Feedback for OSIsoft Products
In addition to Sebastien Raposo , To enable high availability for PI Analysis Service, you must install the service and enable Buffering Manager in each node of a failover cluster. You can install PI Analysis Service in the same failover cluster as PI Notifications Service, or you can install the services in a different failover cluster. I request to refer the link PI Server for more details.
The comment so far have probably given you some idea that this is not trivial to achieve. While technically possible, once i hear the terms 'redundancy' or 'failover' or ..., i immediately start asking about the goals to achieve, the risks that are to be mitigated, and the overall system stacks.
So be sure to ask the right questions, analyze the perceived risks, and make holistic decisions!
These are excellent points Roger Palmen thanks!
This is especially true with Asset Analytics given that it is capable of auto backfilling if the service is down for less than 72 hours (default auto backfilling time range) and it's always possible to manually backfill/recalculate the analyses. In some use cases this is sufficient instead of setting up a failover cluster.
Thanks for your feedback.
I've seen too many times, that a technical solution is chosen for a perceived business requirement (yes, of course we cannot loose data!), but when the costs are involved things might be different (ehm, well, if the price doubles and we need to wait another 2 months, then let's first make this work properly...)
Not a problem we have a server which we can use temporarily.
We have decided to Upgrade OS on weekends (Business impact will be very less)
Retrieving data ...