Resource Requests

May 15, 2024 · 3 min read · k8s kubernetes ·

(this post is part of the material I cover in my devops course)

When you specify a Pod, you can optionally specify how much of each resource a container needs.

Requests and limits

Generally, you have these tools to control the resources that a pod will use:
- Resource Requests
- Resource Limits (which we'll demonstrate in another post)
Resource Requests are evaluated before a pod is scheduled, and after that use weighting to make sure the pod gets its requesged CPU. Memoryy is evaluated mainly before scheduling.
if there is an available CPU on a node, then pods share it using weights according their assigned requests.
The CPU limit defines a hard ceiling on how much CPU time that the container can use.
Same is done for memory limit.
See more details here

Too many CPUs requested

Here's an example of a pod definition, with a request that cannot be granted
The pod requests 60 CPU units:

 1apiVersion: v1
 2kind: Pod
 3metadata:
 4  name: big-cpu-pod
 5spec:
 6  containers:
 7  - name: busybox
 8    image: busybox:latest
 9    command:
10      - sleep
11      - "3600"
12    resources:
13      requests:
14        memory: "50M"
15        cpu: "60"

This is what we see when we apply the definition:

 1$> 
 2$> kubectl apply -f big-cpu-pod.yaml 
 3pod/big-cpu-pod created
 4$> kubectl get pods
 5NAME          READY   STATUS    RESTARTS   AGE
 6big-cpu-pod   0/1     Pending   0          6s
 7$> kubectl describe pods big-cpu-pod
 8Name:             big-cpu-pod
 9Namespace:        default
10Priority:         0
11Service Account:  default
12Node:             <none>
13Labels:           <none>
14Annotations:      <none>
15Status:           Pending
16IP:               
17IPs:              <none>
18Containers:
19  busybox:
20    Image:      busybox:latest
21    Port:       <none>
22    Host Port:  <none>
23    Command:
24      sleep
25      3600
26    Requests:
27      cpu:        60
28      memory:     50M
29    Environment:  <none>
30    Mounts:
31      /var/run/secrets/kubernetes.io/serviceaccount from kube-api-access-rp2fd (ro)
32Conditions:
33  Type           Status
34  PodScheduled   False 
35Volumes:
36  kube-api-access-rp2fd:
37    Type:                    Projected (a volume that contains injected data from multiple sources)
38    TokenExpirationSeconds:  3607
39    ConfigMapName:           kube-root-ca.crt
40    ConfigMapOptional:       <nil>
41    DownwardAPI:             true
42QoS Class:                   Burstable
43Node-Selectors:              <none>
44Tolerations:                 node.kubernetes.io/not-ready:NoExecute op=Exists for 300s
45                             node.kubernetes.io/unreachable:NoExecute op=Exists for 300s
46Events:
47  Type     Reason            Age   From               Message
48  ----     ------            ----  ----               -------
49  Warning  FailedScheduling  21s   default-scheduler  0/5 nodes are available: 5 Insufficient cpu. preemption: 0/5 nodes are available: 5 No preemption victims found for incoming pod.
50
51$>

Note the Pending state, the 60 cpu requests and the Events section at the end/

Too much Memory requested

Here's a simmilar experiment with memory.
The pod definition:

 1apiVersion: v1
 2kind: Pod
 3metadata:
 4  name: big-mem-pod
 5spec:
 6  containers:
 7  - name: busybox
 8    image: busybox:latest
 9    command:
10      - sleep
11      - "3600"
12    resources:
13      requests:
14        memory: "64Gi"
15        cpu: "250m"

Note that the request is for 64 Gigibytes of memory.
Now running it:

 1$> 
 2$> kubectl apply -f big-mem-pod.yaml 
 3pod/big-mem-pod created
 4$> kubectl get pods
 5NAME          READY   STATUS    RESTARTS   AGE
 6big-mem-pod   0/1     Pending   0          4s
 7$> kubectl describe pods big-mem-pod
 8Name:             big-mem-pod
 9Namespace:        default
10Priority:         0
11Service Account:  default
12Node:             <none>
13Labels:           <none>
14Annotations:      <none>
15Status:           Pending
16IP:               
17IPs:              <none>
18Containers:
19  busybox:
20    Image:      busybox:latest
21    Port:       <none>
22    Host Port:  <none>
23    Command:
24      sleep
25      3600
26    Requests:
27      cpu:        250m
28      memory:     64Gi
29    Environment:  <none>
30    Mounts:
31      /var/run/secrets/kubernetes.io/serviceaccount from kube-api-access-kr9jk (ro)
32Conditions:
33  Type           Status
34  PodScheduled   False 
35Volumes:
36  kube-api-access-kr9jk:
37    Type:                    Projected (a volume that contains injected data from multiple sources)
38    TokenExpirationSeconds:  3607
39    ConfigMapName:           kube-root-ca.crt
40    ConfigMapOptional:       <nil>
41    DownwardAPI:             true
42QoS Class:                   Burstable
43Node-Selectors:              <none>
44Tolerations:                 node.kubernetes.io/not-ready:NoExecute op=Exists for 300s
45                             node.kubernetes.io/unreachable:NoExecute op=Exists for 300s
46Events:
47  Type     Reason            Age   From               Message
48  ----     ------            ----  ----               -------
49  Warning  FailedScheduling  25s   default-scheduler  0/5 nodes are available: 5 Insufficient memory. preemption: 0/5 nodes are available: 5 No preemption victims found for incoming pod.
50$> 
51$>

Pod is pending, with simmilar notes.

We'll cover more in further posts.