K
Q

Question

kubectl describe node does not report proper pod usage

6/27/2018

I have a k8s v1.10.2, 3x3, cluster which I was trying to stress. I put together the command:

kubectl run stress --replicas=1 --image=lorel/docker-stress-ng -- --cpu 8 --io 8 --vm 4 --vm-bytes 1024m  --fork 4 --timeout 5m --metrics-brief

and when I look at the node usage via:

kubectl describe node addons-worker-01

the node reports no usage by the pod:

Non-terminated Pods:         (6 in total)
  Namespace                  Name                                  CPU Requests  CPU Limits  Memory Requests  Memory Limits
  ---------                  ----                                  ------------  ----------  ---------------  -------------
  default                    stress-765b45bdd5-qwqbj               0 (0%)        0 (0%)      0 (0%)           0 (0%)

yet when I look at top, the node shows the usage I would expect. Is this expected? Am I missing something?

-- Sancho

kubernetes

1 Answer

6/28/2018

kubctl describe is used to show your current node configuration.

For example:

Namespace       Name                CPU Requests    CPU Limits Memory Requests  Memory Limits
 ---------      ----                ------------    ---------- ---------------  ------------- 
 default        stress                  100m (1%)       0 (0%)          0 (0%)          0 (0%)
 kube-system    fluentd                 100m (1%)       0 (0%)      200Mi (0%)      300Mi (1%) 
 kube-system    kube-dns                260m (3%)       0 (0%)      110Mi (0%)      170Mi (0%) 
 kube-system    kube-proxy-gke-cluster  100m (1%)       0 (0%)          0 (0%)          0 (0%) 
 kube-system    kubernetes-dashboard    100m (1%)    100m (1%)      100Mi (0%)      300Mi (1%)

Allocated resources: 
(Total limits may be over 100 percent, i.e., overcommitted.) 
CPU Requests CPU Limits Memory Requests Memory Limits 
------------ ---------- --------------- ------------- 
660m (8%) 100m (1%) 410Mi (1%) 770Mi (2%)

It means that there are 5 pods on the node and, for example, fluentd requires 100m of CPU to start with no limit set and 200M of memory to start which is limited to 300M.

You can set limits inside a config yaml file for the pod like the following:

spec:
  containers:
  - name: text
    image: nginx
   resources:
      limits:
        memory: 512Mi
      requests:
        memory: 128Mi

More about setting limits of memory and CPU you can read in Kubernetes manage-resources docs.

For monitoring I would recommend Prometheus or Google Cloud Monitoring.

You can also use kubctl top nodes which will show current load on the nodes:

NAME                                       CPU(cores)   CPU%      MEMORY(bytes)   MEMORY%
gke-cluster-1-default-pool-1                7969m        100%      4708Mi          17%
gke-cluster-1-default-pool-2                56m          0%        491Mi           1%
gke-cluster-1-default-pool-3                60m          0%        568Mi           2%

-- Crou

Source: StackOverflow

KQ

kubectl describe node does not report proper pod usage

Similar Questions

1 Answer

K
Q