Incus container memory metrics have spikes that are not visible in metrics from the inside of the container

Incus
1

Hello everyone!
Basically, when I check data from incus_memory_MemAvailable_bytes of my container, then I’m getting the following spikes:

image
image1606×1016 71.1 KB

However, when I check the available memory in this container instance with node_memory_MemAvailable_bytes, I’m getting:

image
image1908×949 65.6 KB

No spikes visible.

I’m using Grafana Alloy, to export node metrics. These spikes are very troublesome because they trigger my alarms. Does anyony have an idea what’s going on here?

My incus version is 6.0.4 on Ubuntu 24.04.2 .
This instance is a container.

2

Hi!

In the graph I can see some very short dips. Most likely this are related to the tool since the metric is memory. It’s not likely that a 6.5GB memory usage would drop to around 200MB momentarily so that it rises up to 6.5GB on the next moment.

The interval is 1 minute. A lot of things can happen within at 1 minute.

Either the tool, when it runs in the container, misreports the memory use, or the container restarts or the process is swapped out (is that possible?).

3

Yes, dips. I forgot the correct word ^^". This happens in all of my containers, and they have pretty much nothing in common.

image
image786×323 16.6 KB

It doesn’t happen on VMs though.

These spikes are not reported by the node_memory_MemAvailable_bytes metric at all.