Memory requests and limits in Kubernetes - Step-by-Step Execution

Process Flow - Memory requests and limits

Pod Spec Created

↓

Set Memory Request

↓

Set Memory Limit

↓

Scheduler Checks Requests

↓

Pod Scheduled on Node

↓

Container Runs

↓

Memory Usage Monitored

↓

If Usage > Limit

↓

If Usage < Request

The pod spec defines memory requests and limits. The scheduler uses requests to place pods. The runtime enforces limits during execution.

Execution Sample

Kubernetes

apiVersion: v1
kind: Pod
metadata:
  name: memory-pod
spec:
  containers:
  - name: app
    resources:
      requests:
        memory: "200Mi"
      limits:
        memory: "500Mi"

This pod requests 200Mi memory and limits usage to 500Mi.

Process Table

Step	Action	Memory Request	Memory Limit	Scheduler Decision	Container State
1	Pod spec created	200Mi	500Mi	Pending scheduling	Not running
2	Scheduler checks node capacity	200Mi	500Mi	Node with >=200Mi free selected	Pending
3	Pod scheduled on node	200Mi	500Mi	Scheduled	Starting container
4	Container starts running	200Mi	500Mi	Scheduled	Running, memory usage 150Mi
5	Memory usage rises to 450Mi	200Mi	500Mi	Scheduled	Running, memory usage 450Mi
6	Memory usage exceeds 500Mi	200Mi	500Mi	Scheduled	Container killed by OOM
7	Container restarted	200Mi	500Mi	Scheduled	Running, memory usage 180Mi

💡 Container killed when memory usage exceeded limit 500Mi

Status Tracker

Variable	Start	After Step 4	After Step 5	After Step 6	After Step 7
Memory Usage	0Mi	150Mi	450Mi	Killed (OOM)	180Mi
Container State	Not running	Running	Running	Killed	Running

Key Moments - 3 Insights

Why does the container get killed even though it requested only 200Mi but used more?

The container requested 200Mi for scheduling but can use up to 500Mi. If usage exceeds 500Mi (limit), the container is killed as shown in step 6.

Can the scheduler place the pod on a node with only 150Mi free memory?

No, because the pod requests 200Mi memory. Scheduler must find a node with at least 200Mi free (step 2).

What happens if the container uses less memory than requested?

The container runs normally. Requests are minimum guaranteed memory for scheduling, not limits on usage (step 4).

Visual Quiz - 3 Questions

Look at the execution table, what is the container state at step 5 when memory usage is 450Mi?

A Pending

B Running

C Killed

D Not running

💡 Hint

Check the 'Container State' column at step 5 in the execution_table.

At which step does the container get killed due to exceeding memory limit?

A Step 4

B Step 5

C Step 6

D Step 7

💡 Hint

Look for 'Container killed by OOM' in the 'Container State' column.

If the memory request was increased to 600Mi, what would happen during scheduling?

A Scheduler would find a node with at least 600Mi free memory

B Scheduler would ignore the request and schedule anyway

C Container would be killed immediately

D Memory limit would automatically increase

💡 Hint

Memory requests affect scheduler decisions as shown in step 2.

Concept Snapshot

Memory requests reserve minimum memory for scheduling.
Memory limits cap max memory usage.
Scheduler uses requests to place pods.
Container killed if usage exceeds limit.
Requests ≤ Limits always.
Requests do not limit usage, limits do.

Full Transcript

In Kubernetes, memory requests and limits control how much memory a container reserves and can use. The pod spec sets these values. The scheduler uses the request to find a node with enough free memory. The container can use memory up to the limit. If usage goes beyond the limit, the container is killed by the system. This trace shows a pod requesting 200Mi and limiting 500Mi memory. The scheduler places it on a node with enough memory. The container runs normally until usage exceeds 500Mi, then it is killed and restarted. Requests ensure scheduling success, limits protect node stability.