Deploy and Troubleshoot Workloads

CKS clusters support running different types of workloads using Pods, Deployments, StatefulSets, DaemonSets, Jobs, and CronJobs. Before deployment, select the workload type based on whether the task needs to run continuously, requires stable identities, node-level replicas, or needs to be scheduled periodically.

Select Workload Type

Workload	Applicable Scenario
Pod	Validating images, debugging commands, or running temporary tasks. For long-running services, prioritize workloads managed by controllers.
Deployment	Runs stateless services and maintains the desired number of Pods through replicas.
StatefulSet	Runs stateful services that require stable network identities or persistent storage.
DaemonSet	Runs one Pod on each matching node, for example, for log collection, node monitoring, or node-level agents.
Job	Runs one-time tasks that exit upon completion.
CronJob	Creates and runs Jobs on a scheduled basis.

Prepare Workload Runtime Conditions

Based on the workload configuration, first prepare the resources required immediately after submission:

Ensure that the target namespace has been created.
If runtime configuration or sensitive information needs to be read, first create a ConfigMap or Secret in the same namespace.
If data needs to be retained, first create a PVC in the same namespace and ensure that the PVC status is Bound.
If a private image needs to be pulled, prepare the image access credentials required by the workload.
If GPU, RDMA, Ingress, or other extended capabilities are required, ensure that the corresponding cluster components are installed and operating normally.

Prepare the Workload YAML

Prepare the YAML based on the selected workload type, and adjust the configuration according to the business image, namespace, resource specifications, environment variables, and storage mount requirements.

Pods are suitable for validating images, debugging commands, or running temporary tasks. For long-running services, use a Deployment, StatefulSet, or DaemonSet for management.

apiVersion: v1
kind: Pod
metadata:
  name: nginx-pod
  namespace: default
  labels:
    app: nginx
spec:
  containers:
    - name: nginx
      image: nginx:1.14.2
      ports:
        - containerPort: 80

A Deployment is suitable for stateless services and can control the number of Pods through the number of replicas.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx-deployment
  namespace: default
spec:
  replicas: 3
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
        - name: nginx
          image: nginx:1.14.2
          ports:
            - containerPort: 80

StatefulSets are suitable for stateful services that require stable network identities or persistent storage. The Headless Service in this example provides stable network identities for the StatefulSet.

apiVersion: v1
kind: Service
metadata:
  name: nginx
  namespace: default
  labels:
    app: nginx
spec:
  ports:
    - port: 80
      name: web
  clusterIP: None
  selector:
    app: nginx
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: web
  namespace: default
spec:
  serviceName: nginx
  replicas: 3
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
        - name: nginx
          image: nginx:1.14.2
          ports:
            - containerPort: 80
              name: web

A DaemonSet is suitable for running one Pod on each matching node, for example, for log collection, node monitoring, or node-level agents.

apiVersion: apps/v1
kind: DaemonSet
metadata:
  name: fluentd
  namespace: default
spec:
  selector:
    matchLabels:
      name: fluentd
  template:
    metadata:
      labels:
        name: fluentd
    spec:
      containers:
        - name: fluentd
          image: fluent/fluentd:v1.16

Jobs are suitable for one-time tasks. After a task is completed, the Pod is retained for a period of time to facilitate viewing its status and Logs.

apiVersion: batch/v1
kind: Job
metadata:
  name: pi
  namespace: default
spec:
  template:
    spec:
      restartPolicy: Never
      containers:
        - name: pi
          image: perl:5.34.0
          command: ["perl", "-Mbignum=bpi", "-wle", "print bpi(2000)"]
  backoffLimit: 4

CronJob is suitable for tasks that run on a scheduled basis. Adjust schedule as needed.

apiVersion: batch/v1
kind: CronJob
metadata:
  name: hello
  namespace: default
spec:
  schedule: "*/5 * * * *"
  jobTemplate:
    spec:
      template:
        spec:
          restartPolicy: OnFailure
          containers:
            - name: hello
              image: busybox:1.28
              command:
                - /bin/sh
                - -c
                - date; echo Hello from the Kubernetes cluster

Submit Workload

Log in to the GPU Cloud console and go to the Clusters page.
Click the name of the target cluster to access the cluster workspace.
In the left navigation pane, go to the Workloads page.
Switch to the tab corresponding to the selected workload type.
Use the creation entry provided on the page, select a namespace, enter the prepared YAML, and then submit to create.
Return to the relevant tab and verify that the newly created resource is displayed.

Confirm Workload Status

On the Workloads page, switch to the tab for the corresponding resource.
Search for the target resource as needed.
Confirm the workload status. When checking Pods, focus on the node, namespace, container readiness, and restart count.

If the workload is not running as expected, check the Pod logs, resource details, and events in the following order.

Troubleshoot Workloads That Are Not Running Properly

View Pod Logs

On the Pods tab, locate the target pod.
In the Actions column, click Log.
View the Pod logs.

To filter logs by time range, keyword, or field, go to the Logs page. For more information, see Querying and Analyzing Logs.

View Resource Details and Events

On the corresponding resource tab, locate the target workload.
In the Actions column, click View.
View resource configuration, status, events, and other information.

If a Pod remains in Pending status for an extended period or restarts frequently, check the node status, PVC binding status, Logs, and Monitor metrics to identify the cause.

Get Started

Compute

Storage

Cloud networks

Observability

Deploy and Troubleshoot Workloads

Select Workload Type

Prepare Workload Runtime Conditions

Prepare the Workload YAML

Submit Workload

Confirm Workload Status

Troubleshoot Workloads That Are Not Running Properly

View Pod Logs

View Resource Details and Events

​Select Workload Type

​Prepare Workload Runtime Conditions

​Prepare the Workload YAML

​Submit Workload

​Confirm Workload Status

​Troubleshoot Workloads That Are Not Running Properly

​View Pod Logs

​View Resource Details and Events

Select Workload Type

Prepare Workload Runtime Conditions

Prepare the Workload YAML

Submit Workload

Confirm Workload Status

Troubleshoot Workloads That Are Not Running Properly

View Pod Logs

View Resource Details and Events