The algorithm to determine the time range changes quite a bit from one version to another. I'll try to shed some light on the behavior.
Which version are you running?
Both 2017R2 and 2018 (different projects, differences between environments).
3 of 3 people found this helpful
Thanks for the information. I provided a brief description below on how the time range is determined in general in each version. There are a few particular situations for which the logic differs. If you witness a different behavior than my description, let us know with some details (last analysis output time, execution time, amount of time the service was down, if the analysis was running prior to the service restart, version...) and we will provide you with an explanation.
You can find the information in the release notes under WI 181275: https://techsupport.osisoft.com/Viewer/File/b7d62798-8e0a-48b8-b49c-c0fb4238b3e0
For convenience, here is the info:
Previously, the auto-backfilling feature of Analysis Service determined the backfilling time ranges based on the latest analysis output time stamps before the service shut down. While it was mostly sufficiently accurate, it caused some undesirable behaviors. For example, newly created analyses (during the time the service was not running) were also backfilled after the service was started. In addition, service fail-over would have resulted in some inconsistent results. These problems are now mitigated by backfilling analyses based on the known latest evaluation time of each individual analysis.
- The last execution time is used as the start time of the autobackfilling time range.
- Analyses that were not running before the service was shutdown are not backfilled during autobackfilling.
- Analyses that were running but never evaluated will be backfilled with a start time corresponding to the minimum execution time of all analyses that were running and previously executed.
The description in the work item contains the information.
- For Expression and Roll Up analyses, the latest output timestamp of the analysis is used as the start time of autobackfilling (I think this explains the behavior you described).
- For Event Frame analyses, the latest event frame end time for a given analysis is used as the start time of autobackfilling.
Hope this helps,
Seems like a perfect answer! Exactly what i was looking for.
I still need to dig through the oddities i see in practice. But that will take some more time, and this is not too high on my prio list at the moment.