Many new errors after upgrading to 1.36.11

[rilny]

Hi there,

I just upgraded from 1.36.5 to 1.36.11 and updated the database. Good news: the "Error writing trailer Error number 67 occurred" messages are now gone (these were previously logged at the end of every capture). And the CPU usage per monitor is reduced by about 1/3, so that's good too! But now I've got a whole bunch of new errors.

Settings first: 4 identical Foscam cameras. Each one is 720P, max bitrate 1M (VBR), 5fps, key frame interval 20. The monitors are all configured the same:

Function: Mocord (with analysis enabled)
SaveJPEGs: Analysis images only
VideoWriter: Camera Passthrough
EncoderParameters: movflags=frag_custom+dash+delay_moov
ImageBufferCount: 10
MaxImageBufferCount: 200

I have 16G of RAM on this box (10G free, with zm running) so I'm not worried about memory usage. Also 8 cores and it's not very heavily loaded. Each monitor is using blobs. Here are the errors I'm getting now, quite frequently:

1. Max blob count reached. Unable to allocate new blobs so terminating. Zone settings may be too sensitive.
2. You have set the max video packets in the queue to 200. The queue is full. Either Analysis is not keeping up or your camera's keyframe interval is larger than this setting.
3. Analysis daemon for 2 cam-xxx needs restarting, time since last analysis 61 seconds
4. 'zmc -m 2' has not stopped at 21/12/17 19:37:58 after 10 seconds. Sending KILL to pid 2386110

Before the upgrade, I would occasionally get the "Max blob count" error, but not very often. Since upgrading, it's much more frequent. The other errors are brand new and never happened before. All of these errors happen across all monitors.

Any ideas?

[iconnor]

The first is obvious, your zone is too sensitive.

The second is due to the keyframe interval being set too large in your camera or the analysis process waiting on decode or locked writing to disk.
#3 is actually just a warning I think no action is taken anymore. However it indicates that analysis is spending too much time doing something (perhaps searching the queue for a keyframe) or maybe locked waiting on the database. use mysqltuner. Also reduce keyframe interval.
#4 is bad, it means everything is locked waiting for something to happen, which shouldn't happen. Makes me think of database problems.

Also, we have .12 now.

[rilny]

1. What does "too sensitive" mean? Here are the settings I use for all zones (I think most of these are the defaults):

Units: Pixels
Min/Max Pixel Threshold: 60 / blank
Filter Width/Height: 3 / 3
Min/Max Alarmed Area: 1000 / blank
Min/Max Filtered Area: 3 / blank
Min/Max Blob Area: 3 / blank
Min/Max Blobs: 1 / blank

Which of these should be adjusted to make it less sensitive?

2. I'm currently at 5fps, keyframe interval = 20. Should I change the keyframe interval to 10 or do you have a recommendation?

I'll check for any db performance issues. Otherwise the biggest question is why these issues are new in 1.36.11 and didn't happen in 1.36.5. (Also I get my packages from rpmfusion so I'll update to 1.36.12 as soon as it's available there.)

[Andyrh]

I set my keyframes to as low as it will go on my cameras which works out to be 1 keyframe per second. Yours is 1 every 4 seconds. Once I did that the error went away. For me that was a 1.34 error, so not new to 1.36.

[iconnor]

Min max areas should at least over 100. I generally have them 1000 to 3000.