Using ICAgent to Collect Container Logs

CCE works with AOM to collect workload logs. When creating a node, CCE installs the ICAgent for you (the DaemonSet named icagent in the kube-system namespace of the cluster). After the ICAgent collects workload logs and reports them to AOM, you can view workload logs on the CCE or AOM console.

Notes and Constraints

The ICAgent only collects *.log, *.trace, and *.out text log files.

Using ICAgent to Collect Logs

  1. When creating a workload, set logging for the container.

  2. Click image1 to add a log policy.

    The following uses Nginx as an example. Log policies vary depending on workloads.

    **Figure 1** Adding a log policy

    Figure 1 Adding a log policy

  3. Set Storage Type to Host Path or Container Path.

    Table 1 Configuring log policies

    Parameter

    Description

    Storage Type

    • Host Path (hostPath): A host path is mounted to the specified container path (mount path). In the node host path, you can view the container logs output into the mount path.

    • Container Path (emptyDir): A temporary path of the node is mounted to the specified path (mount path). Log data that exists in the temporary path but is not reported by the collector to AOM will disappear after the pod is deleted.

    Host Path

    Enter a host path, for example, /var/paas/sys/log/nginx.

    Container Path

    Container path (for example, /tmp) to which the storage resources will be mounted.

    Important

    NOTICE:

    • Do not mount storage to a system directory such as / or /var/run; this action may cause a container error to occur. You are advised to mount the container to an empty directory. If the directory is not empty, ensure that there are no files affecting container startup in the directory. Otherwise, such files will be replaced, resulting in failures to start the container and create the workload.

    • When the container is mounted to a high-risk directory, you are advised to use an account with minimum permissions to start the container; otherwise, high-risk files on the host machine may be damaged.

    • AOM collects only the first 20 log files that have been modified recently. It collects files from 2 levels of subdirectories by default.

    • AOM only collects .log, .trace, and .out text log files in the mount paths.

    • For details about how to set permissions for mount points in a container, see Configure a Security Context for a Pod or Container.

    Extended Host Path

    This parameter is mandatory only if Storage Type is set to Host Path.

    Extended host paths contain pod IDs or container names to distinguish different containers into which the host path is mounted.

    A level-3 directory is added to the original volume directory/subdirectory. You can easily obtain the files output by a single Pod.

    • None: No extended path is configured.

    • PodUID: ID of a pod.

    • PodName: name of a pod.

    • PodUID/ContainerName: ID of a pod or name of a container.

    • PodName/ContainerName: name of a pod or container.

    Log Dump

    Log dump refers to rotating log files on a local host.

    • Enabled: AOM scans log files every minute. When a log file exceeds 50 MB, it is dumped immediately. A new .zip file is generated in the directory where the log file locates. For a log file, AOM stores only the latest 20 .zip files. When the number of .zip files exceeds 20, earlier .zip files will be deleted. After the dump is complete, the log file in AOM will be cleared.

    • Disabled: AOM does not dump log files.

    Note

    • AOM rotates log files using copytruncate. Before enabling log dumping, ensure that log files are written in the append mode. Otherwise, file holes may occur.

    • Currently, mainstream log components such as Log4j and Logback support log file rotation. If you have already set rotation for log files, skip the configuration. Otherwise, conflicts may occur.

    • You are advised to configure log file rotation for your own services to flexibly control the size and number of rolled files.

  4. Click OK.

YAML Example

You can set the container log storage path by defining a YAML file.

As shown in the following figure, an emptyDir volume is mounted a temporary path to /var/log/nginx. In this way, the ICAgent collects logs in /var/log/nginx. The policy field is customized by CCE and allows the ICAgent to identify and collect logs.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: testlog
  namespace: default
spec:
  selector:
    matchLabels:
      app: testlog
  template:
    replicas: 1
    metadata:
      labels:
        app: testlog
    spec:
      containers:
        - image: 'nginx:alpine'
          name: container-0
          resources:
            requests:
              cpu: 250m
              memory: 512Mi
            limits:
              cpu: 250m
              memory: 512Mi
          volumeMounts:
            - name: vol-log
              mountPath: /var/log/nginx
              policy:
                logs:
                  rotate: ''
      volumes:
        - emptyDir: {}
          name: vol-log
      imagePullSecrets:
        - name: default-secret

The following shows how to use a hostPath volume. Compared with emptyDir, the type of volumes is changed to hostPath, and the path on the host needs to be configured for this hostPath volume. In the following example, /tmp/log on the host is mounted to /var/log/nginx. In this way, the ICAgent can collects logs in /var/log/nginx, without deleting the logs from /tmp/log.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: testlog
  namespace: default
spec:
  replicas: 1
  selector:
    matchLabels:
      app: testlog
  template:
    metadata:
      labels:
        app: testlog
    spec:
      containers:
        - image: 'nginx:alpine'
          name: container-0
          resources:
            requests:
              cpu: 250m
              memory: 512Mi
            limits:
              cpu: 250m
              memory: 512Mi
          volumeMounts:
            - name: vol-log
              mountPath: /var/log/nginx
              readOnly: false
              extendPathMode: PodUID
              policy:
                logs:
                  rotate: Hourly
                  annotations:

                    format: ''
      volumes:
        - hostPath:
            path: /tmp/log
          name: vol-log
      imagePullSecrets:
        - name: default-secret
Table 2 Parameter description

Parameter

Description

Description

extendPathMode

Extended host path

Extended host paths contain pod IDs or container names to distinguish different containers into which the host path is mounted.

A level-3 directory is added to the original volume directory/subdirectory. You can easily obtain the files output by a single Pod.

  • None: No extended path is configured.

  • PodUID: ID of a pod.

  • PodName: name of a pod.

  • PodUID/ContainerName: ID of a pod or name of a container.

  • PodName/ContainerName: name of a pod or container.

policy.logs.rotate

Log dump

Log dump refers to rotating log files on a local host.

  • Enabled: AOM scans log files every minute. When a log file exceeds 50 MB, it is dumped immediately. A new .zip file is generated in the directory where the log file locates. For a log file, AOM stores only the latest 20 .zip files. When the number of .zip files exceeds 20, earlier .zip files will be deleted. After the dump is complete, the log file in AOM will be cleared.

  • Disabled: AOM does not dump log files.

Note

  • AOM rotates log files using copytruncate. Before enabling log dumping, ensure that log files are written in the append mode. Otherwise, file holes may occur.

  • Currently, mainstream log components such as Log4j and Logback support log file rotation. If you have set rotation for log files, skip the configuration. Otherwise, conflicts may occur.

  • You are advised to configure log file rotation for your own services to flexibly control the size and number of rolled files.

policy.logs.annotations.format

Multi-line log matching

Some program logs (for example, Java program logs) contain a log that occupies multiple lines. By default, the log collection system collects logs by line. If you want to display logs as a single log message in the log collection system, you can enable the multi-line log function and use the log time or regular pattern mode. When a line of log message matches the preset time format or regular expression, it is considered as the start of a log message and the next line starts with this line of log message is considered as the end identifier of the log message.

The format is as follows:

{
    "multi": {
        "mode": "time",
        "value": "YYYY-MM-DD hh:mm:ss"
    }
}

multi indicates the multi-line mode.

  • time: log time. Enter a time wildcard. For example, if the time in the log is 2017-01-01 23:59:59, the wildcard is YYYY-MM-DD hh:mm:ss.

  • regular: regular pattern. Enter a regular expression.

Viewing Logs

After a log collection path is configured and the workload is created, the ICAgent collects log files from the configured path. The collection takes about 1 minute.

After the log collection is complete, go to the workload details page and click Logs in the upper right corner to view logs.

You can also view logs on the AOM console.

You can also run the kubectl logs command to view the standard output of a container.

# View logs of a specified pod.
kubectl logs <pod_name>
kubectl logs -f <pod_name> # Similar to tail -f

# View logs of a specified container in a specified pod.
kubectl logs <pod_name> -c <container_name>

kubectl logs pod_name -c container_name -n namespace (one-off query)
kubectl logs -f <pod_name> -n namespace (real-time query in tail -f mode)