View core dump files

Background information

In Linux, if a program unexpectedly terminates or exits, the operating system records the state of the working memory of the program at that time and stores the state to a file. This process is called a core dump. You can view and analyze the core dump file to identify the cause of the issue.

The following figure shows some signals that cause a core dump in Linux. These signals have a value of Core in the Action column.

For more information, see Core dump file.

Overview

For Elastic Container Instance, the names and directories of core dump files cannot be specified because the elastic container instances are managed. By default, the name of a core dump file is core.pid, and the file is stored in the current directory.

/ # cd /pod/
/pod # ls
data
/pod # sleep 10
^\Quit (Core dumped) (Press Ctrl+\ to trigger a core dump)
/pod # ls
core.45  data

Even if a core dump file is generated when a container process unexpectedly exits, the file is lost during the container exit and cannot be viewed offline for subsequent analysis.

Elastic Container Instance allows you to specify directories in which to store core dump files to solve the preceding problems. You can store core dump files in external volumes. This ensures that you can view and analyze core dump files offline even if a container exits.

You can use one of the following methods to configure the directory where to store core dump files:

• Kubernetes

  annotations:
      k8s.aliyun.com/eci-core-pattern: "/xx/xx/core"

• OpenAPI

  CorePattern = "/xx/xx/core"

Configuration examples

Typically, core dump files are analyzed offline. Therefore, it is better to store core dump files to external volumes instead of local directories. You can use disks or Network File System (NFS) file systems as external volumes. An NFS file system is used in the following example.

1. Mount an NFS file system and configure the directory where to store core dump files.

   In the following example, the NFS file system is mounted to the /pod/data/dump/ directory of the container and the /pod/data/dump/core directory is used to store core dump files. In this case, all core dump files of the container are automatically stored to the directory and synchronized to the NFS file system. You can obtain the core dump files from the NFS file system for analysis even if the container is released.

   apiVersion: v1
   kind: Pod
   metadata:
     name: test
     annotations:
       k8s.aliyun.com/eci-core-pattern: "pod/data/dump/core" 
   spec:
     containers:
     - image: nginx:latest
       name: test-container
       volumeMounts:
       - mountPath: /pod/data/dump/
         name: default-volume
     volumes:
     - name: default-volume
       nfs:
         server: 143b24****-gfn3.cn-beijing.nas.aliyuncs.com
         path: /dump/
         readOnly: false

2. Trigger a core dump in a directory of the container.

   The following example shows that the name and directory of the core dump file have taken effect.

   / # ls
   bin   dev   etc   home  pod   proc  root  sys   tmp   usr   var
   / # sleep 10
   ^\Quit (core dumped) (Press Ctrl+\ to trigger a core dump)
   / # ls
   bin   dev   etc   home  pod   proc  root  sys   tmp   usr   var
   / # cd /pod/data/dump/
   /pod/data/dump # ls
   core.12

3. Release the elastic container instance.

4. Mount the NFS file system to another elastic container instance and log on to the elastic container instance to view a core dump file.

   The following example shows that the core dump file is not lost, and you can view and analyze the file.

   / # cd /pod/data/dump/
   /pod/data/dump # ls
   core.12