Note_Tech

All technological notes.

Project maintained by simonangel-fong Hosted on GitHub Pages — Theme by mattgraham

Kubernetes: Pod - Lifecycle

Kubernetes: Pod - Lifecycle

Pod Phase

Pod Phase	Description
`Pending`	Since `object` is accepted, until one of the `container` is running
`Running`	Since At least one of containers is running
`Unknown`	The state of the Pod cannot be determined. `Kubelet` has stopped reporting.

Finite Pod: pod that is not intended to run indefinitely

Pod Phase	Description
`Succeeded`	All `containers` complete successfully
`Failed`	At least one of the `container` terminates unsuccessfully

Pod Conditions

Pod Condition
- a detailed report on whether a Pod has passed a specific milestone or “health check” in its lifecycle.
PodScheduled: Whether the pod has been scheduled to a node.
Initialized: Whether the init containers have all completed.
ContainersReady: Whether all containers in the pod are ready.
Ready: Whether the readiness gates are all reporting that all containers are ready.

Condition Type	Description
`PodScheduledHas`	Whether the Pod been assigned to a node by the `Scheduler`
`Initialized`	Whether all `Init Containers` have finished successfully
`ContainersReady`	Whether all `containers` are in the Pod currently ready
`Ready`	Whether the Pod’s `Readiness Probes` are passed
`PodReadyToStartContainers`	Whether the sandbox and network are set up so the Kubelet can start pulling images

Lab: Pod Contidtions

kubectl run web --image=nginx
# pod/web created

kubectl describe pod web
# Conditions:
#   Type                        Status
#   PodScheduled                True
#   Initialized                 True
#   ContainersReady             True
#   Ready                       True
#   PodReadyToStartContainers   True

kubectl get pod web -o yaml
# status:
#   conditions:
#     - type: PodScheduled
#       status: "True"
#       observedGeneration: 1
#       lastTransitionTime: "2026-01-02T19:26:06Z"
#       lastProbeTime: null
#     - type: Initialized
#       status: "True"
#       observedGeneration: 1
#       lastTransitionTime: "2026-01-02T19:26:06Z"
#       lastProbeTime: null
#     - type: PodReadyToStartContainers
#       status: "True"
#       observedGeneration: 1
#       lastTransitionTime: "2026-01-03T17:28:37Z"
#       lastProbeTime: null
#     - type: ContainersReady
#       status: "True"
#       observedGeneration: 1
#       lastTransitionTime: "2026-01-03T17:28:37Z"
#       lastProbeTime: null
#     - type: Ready
#       status: "True"
#       observedGeneration: 1
#       lastTransitionTime: "2026-01-03T17:28:37Z"
#       lastProbeTime: null

Pod Lifecycle

initialization stage
- when the pod’s init containers run
run stage
- when the regular containers of the pod run
termination stage
- when the pod’s containers are terminated

Initialization stage

the stage when init containers run

Pulling the container image

Before each init container is started, its container image is pulled to the worker node
containers.imagePullPolicy field
- determines whether the image is pulled every time, only the first time, or never.
- value:
  - Always:
    - pulled image every time the container is (re)started.
    - the registry still needs to be contacted
      - container will not run if registry unavailable, even the image stored locally.
  - Never:
    - never pulled from the registry
    - image must exist on the worker node beforehand, can be
      - stored locally
      - built on the node
    - if not available locally, raise ErrImageNeverPull event
  - IfNotPresent
    - pulled if it is not already present on the node
    - image is only pulled the first time it’s required.
  - if not specified
    - image tag == :latest: defaults to Always
    - Otherwise, default to IfNotPresent

Running the containers

repeated process of init containers
- image pull
- container start
- next image pull
- next container start
- …
- untill all init containers completed.
All init containers must run to completion before the regular containers can start

Restarting failed init containers

When an init container terminates with an error
- if pod’s restart policy == Always/OnFailure
  - failed init container is restarted
- if pod’s restart policy == Never
  - subsequent init containers and the pod’s regular containers are never started
  - status: Init:Error
  - fix:
    - delete and recreate the pod object

Run stage

the stage when init containers are completed and regular containers create.

`post-start hook` error impact a container

Sequential Start, Parallel Execution
- Regular containers are created synchronously in the order they are defined in the pod’s spec.
- All regular containers execute parallelly.
  - No Blocking: it does not wait for one container to be “ready” or “successful” before starting the next.
post-start hook runs asynchronously with the Regular container process
- execution execution of the post-start hook handler blocks the creation and start of the its container.
- container might restart based on the restart policy
- No Impact on Others

termination of containers is performed in parallel.
pre-stop hooks of the containers are all invoked at the same time
- pre-stop hook blocks the shutdown of the container in which it is defined
- does not block the shutdown of other containers

Pulling the container image

image is pulled from the image registry, following the pod’s imagePullPolicy.
- Once the image is pulled, the container is created.
Even if a container image can’t be pulled, the other containers in the pod are started nevertheless.
- risk:
  - it could raise dependency issue across containers if the dependent container take time pulling image before start.

Running the container

The container starts when the main container process starts.
post-start hook runs asynchronously
startup probe started
- followed by liveness probe is started

Terminating and restarting the container on failures

container is terminated if:
- startup probe fails
- liveness probe fails
- post-start hook fails
Container restarts based on the restartPolicy
- if restartPolicy == Never, container remain in the Terminated state
- note:
  - startup hook fails + restartPolicy == Never: pod status == Completed

Termination Grace Period

termination grace period:
- the amount of time to terminate a container
pod.spec.terminationGracePeriodSeconds field
- default: 30s
- timer starts when
  - pre-stop hook is called
  - the TERM signal is sent if no hook is defined
- If the process is still running after the termination grace period has expired, it’s terminated by force via the KILL signal.
  - This terminates the container.

Termination stage

the stage when the pod object is deleted.
- pod status is changed to Terminating.

Containers Termination

the pod’s containers are terminated in parallel
- if pre-stop hook is not called
  - TERM signals are sent to container
- if pre-stop hook is called,
  - executes pre-stop hook
  - then TERM signals are sent to container
  - terminate process by KILL signal if the deletion grace period expires
After all the containers in the pod have stopped running, the pod object is deleted.

Deletion Grace Period

the amount of time for the containers to shut down on their own.
field metadata.deletionGracePeriodSeconds
- By default, it gets its value from the spec.terminationGracePeriodSeconds field
  - 30s
it is advisable to extend it if the application usually needs more time to shut down gracefully.
Example: manifest

apiVersion: v1
kind: Pod
metadata:
  name: kiada-ssl-shortgraceperiod
spec:
  terminationGracePeriodSeconds: 50
  containers:

Imperative Command

# give the pod 10s to shut down
kubectl delete po kiada-ssl --grace-period 10

# check the log for TERM signal
kubectl logs POD_NAME -c CONTAINER_NAME -f

Lab: Pod phase

Running a pod

# demo_pod_state.yaml
apiVersion: v1
kind: Pod
metadata:
  labels:
    run: web
  name: web
spec:
  containers:
    - image: nginx
      name: web

Apply

# terminal A:
kubectl apply -f demo_pod_state.yaml
# pod/web created

# terminal B:
kubectl get pod --watch
# NAME   READY   STATUS    RESTARTS   AGE
# web    0/1     Pending   0          0s
# web    0/1     Pending   0          0s
# web    0/1     ContainerCreating   0          0s
# web    1/1     Running             0          8s

Update a pod manifest file

# demo_pod_state.yaml
apiVersion: v1
kind: Pod
metadata:
  labels:
    run: web-app # updated
  name: web
spec:
  containers:
    - image: nginx
      name: web

Apply

# terminal A:
kubectl apply -f demo_pod_state.yaml
# pod/web configured

# terminal B:
kubectl get pod --watch
# NAME   READY   STATUS    RESTARTS   AGE
# web    1/1     Running   0          14s
# web    1/1     Running   0          64s
# web    1/1     Running   0          64s
# web    1/1     Running   1 (5s ago)   69s

edit

# edit: update image: nginx:1.29.4-alpine
kubectl edit pod web
# pod/web edited

# terminal B:
kubectl get pod --watch
# NAME   READY   STATUS    RESTARTS      AGE
# web    1/1     Running   1 (42s ago)   106s

Deleting a pod

# terminal A:
kubectl delete -f demo_pod_state.yaml
# pod "web" deleted from default namespace

# terminal B:
kubectl get pod --watch
# NAME   READY   STATUS    RESTARTS        AGE
# web    1/1     Running   1 (4m50s ago)   5m54s
# web    1/1     Terminating   1 (7m12s ago)   8m16s
# web    1/1     Terminating   1 (7m12s ago)   8m16s
# web    0/1     Completed     1 (7m13s ago)   8m17s
# web    0/1     Completed     1               8m18s
# web    0/1     Completed     1               8m18s
# web    0/1     Completed     1               8m18s

Lab: Pod state - ErrImagePull

# terminal A:
kubectl run web --image=xnign
# pod/web created

kubectl delete pod web
# pod "web" deleted from default namespace

# terminal B:
kubectl get pod --watch
# NAME   READY   STATUS    RESTARTS   AGE
# web    0/1     Pending   0          0s
# web    0/1     Pending   0          0s
# web    0/1     ContainerCreating   0          0s
# web    0/1     ErrImagePull        0          1s
# web    0/1     ImagePullBackOff    0          16s
# web    0/1     ErrImagePull        0          31s
# web    0/1     ImagePullBackOff    0          42s
# web    0/1     ErrImagePull        0          56s
# web    0/1     ImagePullBackOff    0          71s
# web    0/1     ErrImagePull        0          101s
# web    0/1     ImagePullBackOff    0          116s
# web    0/1     Terminating         0          2m59s
# web    0/1     Terminating         0          2m59s
# web    0/1     Terminating         0          3m
# web    0/1     ContainerStatusUnknown   0          3m1s
# web    0/1     ContainerStatusUnknown   0          3m1s
# web    0/1     ContainerStatusUnknown   0          3m1s
# ...

Lab: Pod state - One-off Complete

indefinite/one-off pod:
- --restart=Never

# terminal A:
kubectl run demo --image=busybox --restart=Never -- sleep 10
# pod/demo created

# terminal B:
kubectl get pod --watch
# NAME   READY   STATUS    RESTARTS   AGE
# demo   0/1     Pending   0          0s
# demo   0/1     Pending   0          0s
# demo   0/1     ContainerCreating   0          0s
# demo   1/1     Running             0          3s
# demo   0/1     Completed           0          13s
# demo   0/1     Completed           0          14s

Delete

# terminal A:
kubectl get pod
# NAME   READY   STATUS      RESTARTS   AGE
# demo   0/1     Completed   0          78

kubectl delete pod demo
# pod "demo" deleted from default namespace

kubectl get pod
# No resources found in default namespace.

# terminal B:
kubectl get pod --watch
# NAME   READY   STATUS      RESTARTS   AGE
# demo   0/1     Completed   0          114s
# demo   0/1     Completed   0          2m
# demo   0/1     Completed   0          2m

Lab: Pod state - one-off Failed

indefinite/one-off pod:
- --restart=Never

# terminal A:
kubectl run demo --image=busybox --restart=Never -- slep 10 # incoreect command
# pod/demo created

kubectl get pod
# NAME   READY   STATUS               RESTARTS   AGE
# demo   0/1     ContainerCannotRun   0          67s

# terminal B:
kubectl get pod --watch
# NAME   READY   STATUS    RESTARTS   AGE
# demo   0/1     Pending   0          0s
# demo   0/1     Pending   0          0s
# demo   0/1     ContainerCreating   0          0s
# demo   0/1     ContainerCannotRun   0          3s
# demo   0/1     ContainerCannotRun   0          4s

Delete

# terminal A:
kubectl delete pod demo
# pod "demo" deleted from default namespace

kubectl get pod
# No resources found in default namespace.

# terminal B:
kubectl get pod --watch
# NAME   READY   STATUS               RESTARTS   AGE
# demo   0/1     ContainerCannotRun   0          2m42s
# demo   0/1     ContainerCannotRun   0          2m49s
# demo   0/1     ContainerCannotRun   0          2m49s

Note_Tech

Kubernetes: Pod - Lifecycle

Pod Phase

Pod Conditions

Lab: Pod Contidtions

Pod Lifecycle

Initialization stage

Pulling the container image

Running the containers

Restarting failed init containers

Run stage

post-start hook error impact a container

Pulling the container image

Running the container

Terminating and restarting the container on failures

Termination Grace Period

Termination stage

Containers Termination

Deletion Grace Period

Lab: Pod phase

Running a pod

Update a pod manifest file

edit

Deleting a pod

Lab: Pod state - ErrImagePull

Lab: Pod state - One-off Complete

Lab: Pod state - one-off Failed

`post-start hook` error impact a container