cri-resource-manager
cri-resource-manager copied to clipboard
intel
Kubernetes Container Runtime Interface proxy service with hardware resource aware workload placement policies
This patch series adds support for switching policies without shutting down `cri-resmgr`. To allow this to happen, the patch set also reworks/refactors 1) the low-level configuration handling infra, and 2)...
Exists with proper exit code to (0-ok) and prints error message. Can be used as a check before applying new configuration. Just an idea - probably in our extension we...
Tested with ``` reinstall_cri_resmgr=1 k8scri="containerd&cri-resmgr" distro=fedora \ containerd_src=../containerd vm=fedora-nri-containerd ./run.sh interactive ``` Works well. By default e2e run.sh creates DaemonSet if NRI support is enabled. If NRI needs to be...
This draft PR adds a new command line option `--use-nri-plugin`. When passed to `cri-resmgr` this option disables CRI request proxying and enables experimental code for resource management as an NRI...
- Tested on debian, fedora and ubuntu. - Example: distro=fedora on_vm_online=' vm-install-minikube; vm-create-minikube-cluster; vm-command "minikube kubectl -- get pods -A"; exit' test/e2e/run.sh interactive
**Describe the bug** Can not get the adjustment status: _**kubectl get -n kube-system adjustments.criresmgr.intel.com -ojson | jq '.items[].status' null**_ **Expected behavior** For document, the result should be: kubectl get -n...
Ignore before #865 is merged. This draft build on top of PR #865. Add explicit "unlimited" to both MaxCPUs and MaxBalloons. Needs testing "unlimited" CPU balloon inflation and MaxBalloons behavior.
Much of the stuff is now automated.
