Strange behavior of Docker monitoring

[tlamik]

Hi,

I have a couple of Dockers "servers", mostly I have Docker running in LXC. They was builded by Proxmox cloning possibility. But, when I add some of that "clones" to Pulse. I can see only one of them, for some short period of time it changed to another one and immediately switch back. If I add three of them, then all three dockers are switching in monitoring. Every LXC has it's own IP address, different MAC address, name, etc.

[rcourtman]

Seeing one Docker host at a time that keeps swapping is almost always the duplicate Docker engine ID problem described in issue #590. Cloned LXCs keep the same /etc/machine-id, Docker derives the engine ID from that file, and every heartbeat from those agents overwrites the previous one.

Each container needs its own machine ID first, then its own Docker engine ID. Run these steps inside every clone:

1. Regenerate the machine ID with systemd-machine-id-setup.
2. Remove the old Docker engine ID and restart Docker: rm /etc/docker/key.json && systemctl restart docker.
3. Confirm every container now reports a different ID: docker info --format '{{.ID}}'.
4. Make sure each agent uses a unique API token so uploads stay separated, especially on older Pulse builds.

Once the IDs differ the hosts stop clobbering each other, which is exactly how users in #590 resolved the same symptom. If the IDs already differ and the agents still rotate, share those values so the next checks can focus elsewhere.

[tlamik]

Thanks a lot, I have had unique machine-id, but the same Docker engine ID.