Vettester
Known around here
- Feb 5, 2017
- 1,106
- 1,113
Off topic, but did you know that you can significantly reduce your RAM usage by decreasing the "max connects" setting?
Version 6.0.1
- Thread starter Blue Iris Update Bot
- Start date
Wow, no I did not. I went to 20 and it saved 1GB mem!
Vettester
Known around here
- Feb 5, 2017
- 1,106
- 1,113
And the logs show that the alert trigger is happening, but nothing is found:
For some reason it won't let me add my BI AI screen, but I have Built-In selected as "provider", and yolov8m.onnx selected as my primary model. Nothing else is checked.
Last edited:
Vettester
Known around here
- Feb 5, 2017
- 1,106
- 1,113
From your example I noticed extremely high analysis times (over 1 1/2 minutes) which is extremely slow. Are these wifi cameras?
CrazyAsYou
Getting comfortable
is there any way to tell how many "max connects" a system needs? i.e 1 per camera or maybe x2 per camera? I have 8 cameras, so would 8 "max connects" be enough?
As this is a setting for the BI web server, I would suspect that it refers to the max web server connections, presumably from UI3 clients, I have no idea how the BI Android/Apple apps connect, this is just a guess on my part!
biggen
Known around here
- May 6, 2018
- 2,874
- 3,295
For your system to be up for about the 4hrs. shown in your log, you have WAY too many errors going on that need to be addressed. Is it affecting your AI, maybe, but I would still like to see your main BI/settings/AI screen.
Vettester
Known around here
- Feb 5, 2017
- 1,106
- 1,113
This is an AI setting and it controls the maximum number of concurrent requests (or simultaneous connections) that BI sends to an AI processing server. It was created for integrated AI servers like DeepStack and CPAI. It helps manage system resources, preventing memory spikes, backlogs, or overload on slower hardware when processing images for AI.
I've been playing around with this setting for a few days now and it doesn't seem to be needed if you're using the AI built into BI6. I currently have mine set to 1 and haven't seen any issues. I haven't verified this with Ken yet so if you change this setting your results may vary.
Last edited:
The help files refers to this box as follows, perhaps it is out of date?
EDIT - I see now that there is also a max connects setting on the AI page, I was looking at the one on the advanced page of the web server!
The Connection limit box should be left at its default unless you are battling a system
resource issue. One “real” user may be responsible for transiently adding multiple rows to
the Status/Connections list due to connectivity issues for example—and that user may be
unable to re-connect until other connections time out.
Last edited:
I have a couple cameras that are disconnected right now... That's where the errors are coming from. Will be replacing them in the next week or two.
Vettester
Known around here
- Feb 5, 2017
- 1,106
- 1,113
You’ve got some network issues somewhere. If you ping one of your cameras from a command prompt what kind of response times are you seeing?
Network is solid; running UDM Pro with two gigabit Unifi POE switches. Cameras and PC isolated on their own VLAN.
My guess is that these times reflect the time it takes to fire up the AI, as I had just rebooted the BI machine and this was the first alert.
FWIW, it seems to be working now, with no additional modifications. Not sure what was going on, or why it seems to work now, but anyways... Thanks for the assistance! Much appreciated.
I had it set to 32 sessions and had some issues with AI overwhelming my system. Ken asked me to set it to 8 (default for built-in AI). It's been running much better since....
Vettester
Known around here
- Feb 5, 2017
- 1,106
- 1,113
I have 15 cameras all using the built-in AI and have been running mine at 1 for a few days now and it has been working better than ever.
From [6.0.1.9] changelog:
"...If errors are “caught” when using the built-in AI, they are now logged. As a result an error, the analyzer instance may be reinitialized as well, adding some delay. Some users are finding the need to lower the number of simultaneous analyses from the previous maximum value of 32 down to something more reasonable, perhaps 8 or 4, or even 2. Each instance can use significant RAM and graphics memory, which not all systems can accommodate..."