Seeing a stream of warning messages when setting up debug console with kubernetes.
Closed this issue · 6 comments
On setting up debug console with kubernetes pod, I am seeing a stream of warning message on the debug console:
time="2019-06-21T23:35:21.762247108Z" level=warning msg="cpu is not in guest cpu list, using guest cpus" cpu=1 cpuset=0-3 guest-cpus=0 name=kata-agent pid=80 sandbox=fe8e6e345bc9a572e4f16a6a734f763a82d9a00d6b7ced2f6bd0d334d3efba1f source=agent
time="2019-06-21T23:35:31.884159759Z" level=warning msg="cpu is not in guest cpu list, using guest cpus" cpu=1 cpuset=0-3 guest-cpus=0 name=kata-agent pid=80 sandbox=fe8e6e345bc9a572e4f16a6a734f763a82d9a00d6b7ced2f6bd0d334d3efba1f source=agent
The pod did not have any resource constraints.
We need to look into why these warnings are generated. If these are benign messages, then we should turn these to debug messages, as currently the debug console is quite unusable with these messages being printed to the console continuously.
@amshinde are you using static cpu policy?, I think those warnings are to just not apply cpuset if is not possible
@jcvenegas Yes, This is the one used for the clear-linux stack as well:
https://github.com/clearlinux/cloud-native-setup/blob/master/clr-k8s-examples/kubeadm.yaml#L7
I had to manually patch the agent and install it the rootfs to get around this.
@amshinde interesting, this warnings actually where to prevent the container execution failed (instead of just fail). What are the additional changes you needed to add?
@jcvenegas I just commented out this line:
https://github.com/kata-containers/agent/blob/master/cgroup.go#L55
So that I could type things on the debug console again without being interrupted by the continuous warning messages.
oh I see what you mean, It is just a very annoying message. Yeah make sense, probably we can change the log from warn to debug. This happend all the time because k8s keeps doing a kata-runtieme update per each container.