Blue Iris alerts disappearing

[n2qew]

I've run into the issue of the system deleting all the alerts. You get a database error in the logs - mentions corruption detected, and proceeds to wipe the alerts. This is a 3 week old install that was running on 5.6.6. I just updated to 5.6.8.2 after it's second round of wiping the alerts last night. hoping that maybe it's fixed in the "various fixes". Another user mentions similar Constantly seeing no alerts in database - even after restoring via 'repair/regenerate'

So... since it appears that the alerts are only DB stored, and cannot be recovered from the clips in the event that the DB gets hosed, I'm looking for a way to recover for when this happens. I'm thinking snapshotting the DB directory periodically, and then restoring it in the event that it decides to kill what's in the active DB. I'll just lose whatever alerts happened between the time of the last snapshot and when I restore it. Until... whatever is actually causing it gets fixed. Or.. Is there some other way to get it to regenerate the alerts from history?

Win10 Pro 22H2 under VMware
8 gig ram
8 cores Xeon Silver 4208 @ 2.1Ghz - normally under 20% CPU

Local disk is 250 gig for 20 gig new, 140 gig Stored Rolling to Aux1 2TB / 30days on a 90 TB file share.

 

[wittaj]

Have you tried stopping BI, deleting the entire DB folder, and restarting BI?

The way the alert images are created and stored, they are lost on delete or rebuild of the DB.

Keep in mind you don't lose the video, just the alert thumbnails. Yeah it is a pain, but it isn't a complete loss of the events.

What you are trying to do about taking snapshots of the directory and restoring periodically is a recipe for disaster.

If the deleting of the DB folder doesn't fix it, then rollback to the last stable version and hold tight there until BI either adds something you want or it has been noted that this issue is fixed.

 

[fenderman]

I've run into the issue of the system deleting all the alerts. You get a database error in the logs - mentions corruption detected, and proceeds to wipe the alerts. This is a 3 week old install that was running on 5.6.6. I just updated to 5.6.8.2 after it's second round of wiping the alerts last night. hoping that maybe it's fixed in the "various fixes". Another user mentions similar Constantly seeing no alerts in database - even after restoring via 'repair/regenerate'

So... since it appears that the alerts are only DB stored, and cannot be recovered from the clips in the event that the DB gets hosed, I'm looking for a way to recover for when this happens. I'm thinking snapshotting the DB directory periodically, and then restoring it in the event that it decides to kill what's in the active DB. I'll just lose whatever alerts happened between the time of the last snapshot and when I restore it. Until... whatever is actually causing it gets fixed. Or.. Is there some other way to get it to regenerate the alerts from history?

Win10 Pro 22H2 under VMware
8 gig ram
8 cores Xeon Silver 4208 @ 2.1Ghz - normally under 20% CPU

Local disk is 250 gig for 20 gig new, 140 gig Stored Rolling to Aux1 2TB / 30days on a 90 TB file share.

There were issues with the database in 5.6.6...

 

[n2qew]

Have you tried stopping BI, deleting the entire DB folder, and restarting BI?

I have. Of course that brings back the clips, but the alerts are still missing.

The way the alert images are created and stored, they are lost on delete or rebuild of the DB.
Keep in mind you don't lose the video, just the alert thumbnails. Yeah it is a pain, but it isn't a complete loss of the events.

That's the problem I'm trying to avoid. We've got an old Lorex DVR that I'm trying to use this to kill off. I'm getting video for each camera via RTSP from the Lorex, so both are recording now. Not all the cameras have MD, and this allows searching for times of motion in areas for all cameras. I run low res continuous - 500kbps or so - for all, and run full res record on motion. Works pretty good except when it's deleted the alerts. I have had two others in service since 2021 at my mom's and daughter's, so I'm not new to BI. MD not nearly as critical there, and they've behaved well, though they haven't been updated since 2021. This one is at work, and MD / alerts are more important.

What you are trying to do about taking snapshots of the directory and restoring periodically is a recipe for disaster.

If the deleting of the DB folder doesn't fix it, then rollback to the last stable version and hold tight there until BI either adds something you want or it has been noted that this issue is fixed.

Can you elaborate and what problems this could cause? I'm not suggesting that I will be restoring the DB periodically. The idea I had was to grab a snapshot of the DB files maybe daily, and hold them for a bit... Maybe a few days. If I find that the system has a hiccup related to the DB and wipes the alerts, I'd go back a day or so, and put that set of DB files in the DB folder, and run a compact - repair. That should leave me with only a day or so of missing alerts rather than absolutely none. I'd think that since I can just wipe the DB folder, and restart the service and it recreates the clip data based on where they actually are, this shouldn' cause any additional problems. Am I unaware of or missing something?

You mentioned "noted that this issue has been fixed". Is there an actual change log or list of known outstanding issues you could point me in the direction of?

 

[wittaj]

Keep in mind as pointed out that this particular 5.6.6 version update of BI has been problematic.

Hopefully your update to the latest resolves it for you and others. If not, rollback to a stable version prior to 5.6.6.x.

Keep in mind most here run older versions of BI and only update when BI adds something that we really want or need. Is there anything in 5.6.6.X or above that you really need?

Deleting rebuilding the database has always lost the alerts. It is a pain but usually it is a rarity that one has to do that. But nothing is lost - all the video is still there.

But in doing this periodic restoring the DB, it is sure to cause DB errors at some point. Maybe you just get in the log file error messages and nothing really happens except missing alert thumbnails, or maybe it really messes it up and something worse happens.

Why would you want to lose a few days of alert thumbnails in favor of longer alert images?

While having lots of days/months of alert thumbnails sounds great, the reality of it is unless it was something catastrophic (which you would have known about sooner anyway), most are not going to start scrubbing video for something that may have happened a few weeks ago.

By spending time to dial in the alerts and a frequent peek at what is going on, you would have noticed something around your property within days. I literally every morning in under 30 seconds can scrub what happened the night before and see if anything happened I need to look at further

If a neighbor comes up to me and says "sometime around 2 weeks ago someone backed into my car, can you see if you caught it?" You will find that even with the best scrubbing this is a monumental task. Unless they can narrow down the day/time window, most of us are not going to scour it.

Now I recognize you said this is for work, so the use case may be different and maybe the thumbnails are more important, so if so, once you are on a BI version that is stable FOR YOU, do not update it until they add something you want or need. Otherwise you run the risk of issues like this or worse that sets you back.

As I said, many here run older versions. Some are still on BI4 as it fits their needs. I am still using Deepstack so I am on the version prior to when they switched to CodeProject. Nothing has been added since in BI that I need or want, so why mess with a system that is working for my needs?

 

[Village Guy]

My system is running 5.6.7.3 and I'm also experiencing issues with Alerts disappearing. If that is not bad enough I'm unable to regenerate them.

I have been in contact with Ken in support, alas he is putting it down to pilot error. Ken has advised that alerts are kept in the database and are independent of the Alert folder which in my case is a 5GB RamDrive that gets cleared on re-boot. I have been adjusting the size of the RamDrive this evening and restarting BI and have just lost all the Alerts in the calendar for the last week!

This Alerts issue started around the time Ken was making database changes. While I appreciate that the alerts folder is effectively a momentary storage that in theory is emptied after clip examination. My present theory is that unless the the Alerts are pointed to a traditional drive, Alerts can get lost after a re-boot. Before anyone says it does not make sense, I agree!

I plan to test the theory by making a change tomorrow. I should add before the database change my system has been running faultlessly for several years.

 

[digger11]

I'm trying to work through database corruption issues with Ken. For me 5.6.3.7 runs with no issues, but BI has logged frequent db corruption messages on every version that I've tried from 5.6.6.1 - 5.6.8.1.
Ken said that he rolled back some of the database changes to their pre 5.6.6.1 state for 5.6.8.2, and so far it seems to be less prone to db corruptions than the other releases from 5.6.6.1 and beyond.

Today I discovered that my BI server has logged the exact same bugcheck and bluescreened a total of 5 times since mid-December.
It seems somewhat suspicious that the first one of these occurred 1 day after I upgraded BI to 5.6.6.1, and none of them occurred during the week I rolled back to 5.6.3.7.

The computer has rebooted from a bugcheck.  The bugcheck was: 0x0000010e (0x0000000000000017, 0xffffffffc000000d, 0x0000000000000000, 0x0000000000000000). A dump was saved in: C:\WINDOWS\MEMORY.DMP. Report Id: a428d0f0-a6bd-445f-bdbc-9adf6e5d1178.

 

[Mike A.]

Yeah, lots of us have been having the same sorts or errors in recent versions so it's not pilot error. I've lost track of what happened exactly when at this point but after deleting all video and letting it rebuild from scratch I've been OK on 5.6.6.1 for a while now. Previously without deleting older files it was back and forth with rebuilds and then errors. Makes me think that there's something retained in the pre-change files that affects the database after.

 

[digger11]

Overnight more db corruption, so for me at least, 5.6.8.2 doesn't fix the corruption issue.
Ken asked me to run BI NOT as a service to see whether that has any effect on the problem.
I'll post an update if/when additional corruption is logged.

 

[Village Guy]

Apart from the strange behavior of losing Alert events 5.6.7.3 works just fine. Perhaps its worth starting a thread on DB corruption.

 

[digger11]

According to Ken he made some significant changes to 5.6.8.4 to address database corruption issues.

By the time I had the opportunity to upgrade BI on Jan 25, 5.6.9.2 was available. After installing it, since it appeared that BI had to rebuild the database from scratch and based on recommendations I had seen here on the forum, I did shut down BI and deleted the database directory to insure that BI would have to rebuild it from scratch.

Since then I've run 5.6.9.2, 5.6.9.3, 5.6.9.6, and now 5.6.9.7 without seeing any more database corruption.

Now if Ken puts out a version that changes the ALPR behavior so that the highest confidence plate result is returned rather than the plate result from the image with the highest DayPlate/NightPlate confidence, I'll be a happy camper.

 

[Gyula614]

Hi,

I have 5.7.7.18 and after periodic DB compacting the last few hours of alerts disappear.
Compacting is set to 2 am each day. Alerts from the previous night to the compacting event go missing.
I was just replayed the alerts now when it started compacting. Then when I came back after it finished, the clips I was watching few minutes before went missing.
Does anybody have similar issues and know a solution to it? Thanks.

Gyula

 

[AlwaysSomething]

I have 5.7.6.7 and last night I lost 2 days of alerts and flagged clips. To clarify, I mean on the left hand side in the Clips list using the filter. I did have alerts and flagged clips from prior to 2 days ago (i.e. a weeks ago) so that was strange. What's more strange is that I just watched the alerts and flagged some less than 12 hours ago so why go back to the point it did (2 days ago). FYI - I have it set to compact the DB every night as well but mine is 8PM.

I was actually missing all clips for the last 2 days but I did a Rebuild/Repair on the DB (without deleting the folder) and it added back the full clips but not the alerts or flagged clips. My understanding is that is normal though since the DB is where the alert/flag information is contained. It is just reading the files in the folder to get the full videos back into the DB.

However, my issue may not be BI because I have been having issues with Windows/PC. Sometimes it won't wake up the monitor and I have to force shutdown the PC (hold the power button down). Been chasing that for a while and may have to just rebuild Windows at this point. I won't hijack this thread with that.

Keeping on topic with this thread...
I wish there was a way to take a select group of videos (the full clips) and reprocess them to at least regenerate the alerts clips. That would at least reduce what I have watch to get back what I need to flag/export. In my case it would reduce 2 days of videos I have to watch to just about an hour or so of viewing.

Even if it was one video at a time that I could reprocess it would help and save a lot of time. Possible enhancement to make a right click -> reprocess triggers/alerts? It would also help when adding a new camera and trying to get the motion and AI settings dialed in. There may be a way to do this but I haven't found it and if there is my apologizes in advance since I'm still learning (discovering new amazing things every day).