[08:07:11] hello hello [08:07:35] oh yeah, we worked on that to release canary versions of the agent [08:08:45] some versions, especially when rolling out new eBPF features, have proved to induce pretty high CPU usage swings, and that was a way to release a canary version of this Daemonset to production [08:09:11] as some performance degradation behaviours would only appear in our production system, and not staging [08:10:11] I was not involved with the development of that feature, but I've been involved with the release of some canary versions in which I was introducing new checks/features, meaning I needed to keep an eye one performance of the canary release [09:36:35] Hey everyone. Due to the way we (ML) run pods, we need a larger kubelet partition (/var/lib/kubelet). In the past, we just used LVM commands to bump the partition, but with more big models like LLMs coming in the future, I'd prefer to have a larger partition by default/right after (re)image. The ticket for this is T365971 and I've filed a WIP change (1036195). This would basically copy [09:36:37] the k8s overlay recipe and tweak the kubelet partition size, thne wire it up via preseed.yaml. Thoughts? [10:42:38] klausman: we have traditionally done something similar for openstack hypervisors and the nova instance storage (before ceph) [10:49:48] So at leasst I am not completely off the mark :) [10:51:14] yeah [10:51:28] we used to have dedicated recipes for the installer [10:53:52] I think this is one: modules/install_server/files/autoinstall/partman/custom/labvirt-ssd.cfg [11:34:06] Any volunteers for the code review? :)