diff options
Diffstat (limited to 'health/guides/kubelet/kubelet_10s_pleg_relist_latency_quantile_05.md')
-rw-r--r-- | health/guides/kubelet/kubelet_10s_pleg_relist_latency_quantile_05.md | 35 |
1 files changed, 35 insertions, 0 deletions
diff --git a/health/guides/kubelet/kubelet_10s_pleg_relist_latency_quantile_05.md b/health/guides/kubelet/kubelet_10s_pleg_relist_latency_quantile_05.md new file mode 100644 index 000000000..595fae8a5 --- /dev/null +++ b/health/guides/kubelet/kubelet_10s_pleg_relist_latency_quantile_05.md @@ -0,0 +1,35 @@ +### Troubleshoot the alert + +1. Check Kubelet logs + To diagnose issues with the PLEG relist process, look at the Kubelet logs. The following command can be used to fetch the logs from the affected node: + + ``` + kubectl logs -n kube-system <node_name> + ``` + + Look for any error messages related to PLEG or container runtime. + +2. Check container runtime status + Monitor the health status and performance of the container runtime (e.g. Docker, containerd) by running the appropriate commands like `docker ps`, `docker info` or `ctr version` and `ctr info`. Check container runtime logs for any issues as well. + +3. Inspect node resources + Verify if the node is overloaded or under excessive pressure by checking the CPU, memory, disk, and network resources. Use tools like `top`, `vmstat`, `df`, and `iostat`. You can also use the Kubernetes `kubectl top node` command to view resource utilization on your nodes. + +4. Limit maximum Pods per node + To avoid overloading nodes in your cluster, consider limiting the maximum number of Pods that can run on a single node. You can follow these steps to update the max Pods value: + + - Edit the Kubelet configuration file (usually located at `/etc/kubernetes/kubelet.conf` or `/var/lib/kubelet/config.yaml`) on the affected node. + - Change the value of the `maxPods` parameter to a more appropriate number. The default value is 110. + - Restart the Kubelet service with `systemctl restart kubelet` or `service kubelet restart`. + - Check the Kubelet logs to ensure the new value is effective. + +5. Check Pod eviction thresholds + Review the Pod eviction thresholds defined in the Kubelet configuration, which might cause Pods to be evicted due to resource pressure. Adjust the threshold values if needed. + +6. Investigate Pods causing high relisting latency + Analyze the Pods running on the affected node and identify any Pods that might be causing high PLEG relist latency. These could be Pods with a large number of containers or high resource usage. Consider optimizing or removing these Pods if they are not essential to your workload. + +### Useful resources + +1. [Kubelet CLI in Kubernetes official docs](https://kubernetes.io/docs/reference/command-line-tools-reference/kubelet/) +2. [PLEG mechanism explained in Redhat's blogspot](https://developers.redhat.com/blog/2019/11/13/pod-lifecycle-event-generator-understanding-the-pleg-is-not-healthy-issue-in-kubernetes/) |