1.36.19 Decoding is not keeping up

[lunaticds]

Hi,
ZM newbie here.
I performed a new build on Ubuntu 22.04 in a VM yesterday - 16vCPU/24GB RAM. The host is a HPE Proliant DL380 Gen 9 running ESXi 6.7. I've got around a dozen cameras - mostly 4-6MP Hikvisions that I want to record at full resolution. At the moment they're running at 20fps, and most are configured for H265.
Once I had a couple of cameras in and configured, I noted the logs started reporting "Decoding is not keeping up. We are xxx seconds behind capture".

Having worked through the various camera settings, I reduced them down to the Record function with no analysis or decoding enabled, no Save JPEGs and Camera Passthrough. This managed to get me to a point where I was able to record multiple cameras successfully overnight.

Turning decoding back on so I can enable Live View seems to bring the error back. With it off on all cameras, the 3 test cameras I left recording overnight seem to be recording fine and the error logs aren't overrun with warnings.

For the sake of trying to understand - I did a second install in another VM, and added a single camera running as Monitor only in order to narrow down where this issue might be coming from. I've left it at 3072x2048 with Save JPEGs and Video Writer both disabled. My understanding is that in this mode it should be doing nothing more than rendering the feed from the camera.
Overnight the "Decoding is not keeping up" error has returned. Looking at the video feed, the camera's timestamp is now 4 hours behind.
This VM is currently tying up nearly 2 vCPUs and around 11.6GB RAM merely decoding a single camera with a load average around 2.00. The other VM doing the recording of the 3 cameras at full resolution/20fps in comparison has a load average of 0.55 and is using less then 2GB.

Am I missing something? Any help appreciated.
Thank you

[lunaticds]

In the end I...
- Set the cameras to record at high res only
- Set up new monitors using substreams with modect at low res to use for live view.
I did try linking and running the high res cameras set to nodect with the low res cameras, but they weren't triggering. I can at least this way find events in low res and pull the high res video if needed.
It's been otherwise stable overnight so it seems to be a reasonable way around it. Resource utilisation is a lot better to the point I should be able to cut the RAM back to 16GB and still not worry about OOM issues.

[asunturd]

Thanks for sharing your experiences

Im planning to do some 1080p and my tests have show sluggish recordings so far(centos8, 8GB, 1 camera) but Im in process of setting up new server and doing a clean install on ubunto 22.04 as you have.

[aj3]

I just setup a new Debian10 server and ZoneMinder 1.36.20

I'm getting this error in the Logs... and I have about 50% of the cameras recording.

Any idea if I can make a change and have them record?

The same cameras were previously recording in ZoneMinder 1.34

[apber]

I started seeing this after the wireless traffic increased significantly on vlan where my two cameras are. Solved it by separating the traffic through less busy switch ports.