After installation, the PI Analysis Service can run under a local account even if the PI Asset Framework is installed on a different computer, since NTLM authentication can be used between both computers. However, the PI Server install kit does not recognize this and instead forces a domain account to be used for PI Analysis Service if the PI Asset Framework is installed on a different computer. See the attached screenshot.

If the computer running the install kit is part of a domain, then each time the user wants to upgrade the PI Analysis Service, they must use a domain account temporarily and then switch back to the local account after installation, which is a pain. If the computer running the install kit is not part of a domain, then the PI Analysis Service cannot be installed or upgraded on that computer.

Please allow local accounts to be used during the installation of the PI Analysis Service if the account can be NTLM authenticated against the PI Asset Framework server.

For more information on NTLM authentication and PI, see this article:
https://customers.osisoft.com/s/knowledgearticle?knowledgeArticleUrl=KB00354

Currently, the procedure to recover buffer queue files consists of parsing the queue file into a temporary archive and then merging this archive with the registered archive files that overlap with the impacted period.
This final merge requires to take the registered archive offline, which risks dataloss if new data arrives for the period of this archive. Alternatively, the archive subsystem can be stopped to prevent this dataloss, but this will impact users and calculations that need archive data, which in turn leads to incorrect data or inconsistencies.
Additionally, in the case of collectives, this needs to be repeated for all collective members, risking differences in data between the members while the merge process is done.

The only available option to do this without risking dataloss or inconsistencies is setting up a temporary server with a backup and merge the queue/archive to this temporary server, which then allows you to extract the data with standard tools. This is not the easiest or most practical solution.

There should (at least) be an option to extract events with their point ID from buffer queue files or offline archive files to a CSV format, so that the events can be manually processed with the SDK without the need to unregister archives or stop the archive subsystem.

Even better would be the possibility to merge a buffer queue or offline archive file with a registered archive file (as an extension to the current online reprocessing).
Understandably, this is much more complex, so a minimum viable solution is a way to extract the events to a CSV format.

It would sure be nice if OSI had a way to better manage users access, because we know there are a lot of people who have the client tools but move to a different job or end up not needing them, so we want to pull their license/programs away and give it to someone else opposed to having to buy more. It would be really nice if as an option in the Mappings & Trusts beside the Trust name there was a counter that kept track how many times that trust was used, and have it reset to 0 every year Jan 1st.

We have an AF structure that uses templates for a few device types. Our templates are setup to resolve PI tags for the device if we have them available. If we don't, it says PI Tag Not Found - which is fine. Once the device is added in the field (out of the office), that AF element will update and work.

When in PI Vision, the user gets thousands of errors saying the PI tag is not found. I would like an option in PSE to set the element attribute to ignore the error if the tag is not found, or simple return something I specify such as, not available in the field at this time.

The purpose of templates is to implement something that works for most. I do not want to disable fields not working one at a time.

I use analysis to toggle a boolean or integer from 0 to 1. This is used as source to close/open event frames (step continious mode). this is working fine in real-time. But when I do a recalculation for a part of time in the past (not until *) on the analysis toggling the boolean/integer, the result can be that the last event (on thePI point) in that time frame is equal to the next event. Causing wrong results when the recalculation of the event frame generation analysis is done. If we could tell to close/open an every event we would not run into that issue

As a user I would like to have versioning control of all ad-hoc element attributes and its analysis as well as element attributes that are part of an element template.

Currently we have to manually export XML files of everything in our Database for every single change that is made. Then we have to keep track of the changes in a word document. This is cumbersome to do for every change.

It would be nice to have a Git like ability where multiple individuals can work in a PSE database and commit to a master version of the database once their changes are completed.

Currently, once we click "check-in" everything is a go and there is no going back in case there are problems unless we manually backup.

Our code has become extensive and grows everyday. This is a real need and anyone adopting extensive use of AF into their organization will quickly realize this is a must.

We have needed this for well over a year.