Thanks for you reply. I didn't try the "reprocess step", instead I just deleted some datapoints and recreated the test data and all was well... I wanted to see what would happen if a new snapshot was added after the rogue point but didn't get the time. Also, sorry I missed that "Known Issue" when doing the search before I posted a question.
4 of 4 people found this helpful
As Thyag noted, you were probably running into the known issue he referenced. The known issue is encountered when the snapshot event for a point is "out of sequence" with the archived data, as you can see with the events you provided in the description. There are several scenarios that could cause this behavior, but the most common is probably a crash of the system or PI Snapshot Subsystem (as seems to be the case here). The out-of-sequence event occurs because by default PI Snapshot Subsystem only flushes snapshot events to disk every 15 minutes. So if the system or process crashes, the next time PI Snapshot Subsystem starts up, the snapshot event will be whatever was last flushed to disk, which may potentially be up to 15 minutes prior to the crash.
Note that the actual problem event here is the out-of-sequence snapshot event, which means that reprocessing the archive does not resolve this issue. In order to resolve it, you would have to somehow make sure that this snapshot event is no longer out of sequence. This means there are the following options (in order of preference):
- Delete the snapshot event - this will pull the most recent archived event to replace the snapshot event
- Write an event to the point with a timestamp more recent than the snapshot event timestamp - this will potentially archive the snapshot event and may cause a duplicate event
- Delete the archived events in backwards order until the snapshot event is no longer out of sequence - probably not as preferred as the options above, since it could potentially be a large number of events, and it may not be preferable to delete these events
I have gone ahead and added these options to the workaround section of the known issue page.
Great answers, Gavin Chen. I just wanted to chime in that we experienced this with one of our clients where we did a system recovery after they lost their PI server. We restored the system using a PI backup. The problem "goes away" after a while once a new snapshot event comes in with a timestamp greater than the last archived event. It's transient but it is annoying to be sure.