Over the past few weeks, ZeroTier has been at 100% CPU usage on a Raspberry Pi 3 (linux/armv7l).
It seems like the usage spikes started once the install was upgraded to 1.10.1. Here’s a graph showing the usage changes (full disclosure, the install is done using Home Assistant). Upgrade occurred ~14 days ago.
The two graphs show network traffic and cpu usage of the container over 3 months (May until now). I attempted uninstalling and clearing all data and then reinstalling and still had issues. I haven’t changed anything with my ZeroTier networks in a long while.
I tried that out and didn’t get anywhere. I dumped the network traffic and there seemed to be some issues with packets being sent over ipv6. I disabled ipv6 on the host and that seems to have normalized things. Not sure why this seems to have just started being an issue after the update (I’ve been dualstack for a few years). I’ll have to do some more debugging this weekend. Sorry for the noise!
Having same symptoms with 100% CPU in the following environment: 3x Proxmox kvm host in cluster,
zerotier on all 3 for external access. As soon as I’m trying to upgrade from 1.8.7 to any higher version on a node, zerotier CPU% on this node is increasing. After the 3rd node has e.g. 1.10.1 - all 3 nodes show 95…100% CPU for zerotier process. Same on another cluster with 3x Proxmox (i.e. debian 11 bullseye).
ZT is not configured for ipv6, but I cannot try to disable v6 on the host, since it is needed.
So I’m stuck at 1.8.7 and set zerotier upgrades on hold using dpkg --set-selections.
