Configuring Container Health Check

Scenario

Health check regularly checks the health status of containers during container running. If the health check function is not configured, a pod cannot detect application exceptions or automatically restart the application to restore it. This will result in a situation where the pod status is normal but the application in the pod is abnormal.

Kubernetes provides the following health check probes:

• Liveness probe (livenessProbe): checks whether a container is still alive. It is similar to the ps command that checks whether a process exists. If the liveness check of a container fails, the cluster restarts the container. If the liveness check is successful, no operation is executed.
• Readiness probe (readinessProbe): checks whether a container is ready to process user requests. Upon that the container is detected unready, service traffic will not be directed to the container. It may take a long time for some applications to start up before they can provide services. This is because that they need to load disk data or rely on startup of an external module. In this case, although the application process has started, the application cannot provide services. To address this issue, this health check probe is used. If the container readiness check fails, the cluster masks all requests sent to the container. If the container readiness check is successful, the container can be accessed.
• Startup probe (startupProbe): checks when a containerized application has started. If such a probe is configured, it disables liveness and readiness checks until it succeeds, ensuring that those probes do not interfere with the application startup. This can be used to adopt liveness checks on slow starting containers, avoiding them getting terminated by the kubelet before they are started.

Check Method

• HTTP request

  This health check mode applies to containers that provide HTTP/HTTPS services. The cluster periodically initiates an HTTP/HTTPS GET request to such containers. If the return code of the HTTP/HTTPS response is within 200–399, the probe is successful. Otherwise, the probe fails. In this health check mode, you must specify a container listening port and an HTTP/HTTPS request path.

  For example, for a container that provides HTTP services, the HTTP check path is /health-check, the port is 80, and the host address is optional (which defaults to the container IP address). Here, 172.16.0.186 is used as an example, and we can get such a request: GET http://172.16.0.186:80/health-check. The cluster periodically initiates this request to the container. You can also add one or more headers to an HTTP request. For example, set the request header name to Custom-Header and the corresponding value to example.

  Figure 1 HTTP request-based check
  
• TCP port

  For a container that provides TCP communication services, the cluster periodically establishes a TCP connection to the container. If the connection is successful, the probe is successful. Otherwise, the probe fails. In this health check mode, you must specify a container listening port.

  For example, if you have an Nginx container with service port 80, after you specify TCP port 80 for container listening, the cluster will periodically initiate a TCP connection to port 80 of the container. If the connection is successful, the probe is successful. Otherwise, the probe fails.

  Figure 2 TCP port-based check
  
• CLI

  CLI is an efficient tool for health check. When using the CLI, you must specify an executable command in a container. The cluster periodically runs the command in the container. If the command output is 0, the health check is successful. Otherwise, the health check fails.

  The CLI mode can be used to replace the HTTP request-based and TCP port-based health check.

  • For a TCP port, you can use a program script to connect to a container port. If the connection is successful, the script returns 0. Otherwise, the script returns –1.
  • For an HTTP request, you can use the script command to run the wget command to detect the container.

    wget http://127.0.0.1:80/health-check

    Check the return code of the response. If the return code is within 200–399, the script returns 0. Otherwise, the script returns –1.

    Figure 3 CLI-based check
    
    

    • Put the program to be executed in the container image so that the program can be executed.
    • If the command to be executed is a shell script, do not directly specify the script as the command, but add a script parser. For example, if the script is /data/scripts/health_check.sh, you must specify sh/data/scripts/health_check.sh for command execution.
• gRPC check
  gRPC checks can configure startup, liveness, and readiness probes for your gRPC application without exposing any HTTP endpoint, nor do you need an executable. Kubernetes can connect to your workload via gRPC and obtain its status.

  

  • The gRPC check is supported only in CCE clusters of v1.25 or later.
  • To use gRPC for check, your application must support the gRPC health checking protocol.
  • Similar to HTTP and TCP probes, if the port is incorrect or the application does not support the health checking protocol, the check fails.

  Figure 4 gRPC check
  

Common Parameters

YAML Example

apiVersion: v1
kind: Pod
metadata:
  labels:
    test: liveness
  name: liveness-http
spec:
  containers:
  - name: liveness
    image: <image_address>
    args:
    - /server
    livenessProbe:                 # Liveness probe
      httpGet:                     # Checking an HTTP request is used as an example.
        path: /healthz             # The HTTP check path is /healthz.
        port: 80                   # The check port number is 80.
        httpHeaders:               # (Optional) The request header name is Custom-Header and the value is Awesome.
        - name: Custom-Header
          value: Awesome
      initialDelaySeconds: 3
      periodSeconds: 3
    readinessProbe:                # Readiness probe
      exec:                        # Checking an execution command is used as an example.
        command:                   # Command to be executed
          - cat
          - /tmp/healthy
      initialDelaySeconds: 5
      periodSeconds: 5
    startupProbe:                  # Startup probe
      httpGet:                     # Checking an HTTP request is used as an example.
        path: /healthz             # The HTTP check path is /healthz.
        port: 80                   # The check port number is 80.
      failureThreshold: 30
      periodSeconds: 10