K
Q

Why the multiple containers pod status is Completed even when one of the containers exit with error?

10/1/2019

I have a pod with multiple containers and one of them (containerA) exits with error:

Containers:
  containerA:
    ......
    State:          Terminated
      Reason:       Error
      Exit Code:    1
      Started:      Fri, 27 Sep 2019 16:21:53 -0700
      Finished:     Fri, 27 Sep 2019 16:21:53 -0700
    Ready:          False
    Restart Count:  0

  containerB:
    ......
    State:          Terminated
      Reason:       Completed
      Exit Code:    0
      Started:      Fri, 27 Sep 2019 16:21:54 -0700
      Finished:     Fri, 27 Sep 2019 16:21:59 -0700
    Ready:          False
    Restart Count:  0

  containerC:
    ......
    State:          Terminated
      Reason:       Completed
      Exit Code:    0
      Started:      Fri, 27 Sep 2019 16:21:54 -0700
      Finished:     Fri, 27 Sep 2019 16:21:58 -0700
    Ready:          False
    Restart Count:  0

This pod has restartPolicy: Never. And it's been controlled by a Job with backoffLimit: 9. During all the attempts, the pod status is shown as:

NAME                   READY   STATUS       RESTARTS   AGE
my-pod-2scsn           0/4     Completed    0          3d18h
my-pod-8z7qq           0/4     Completed    0          3d18h
my-pod-9cjnc           0/4     Completed    0          3d18h
my-pod-f6hxr           0/4     Completed    0          3d18h
my-pod-fz7hk           0/4     Completed    0          3d18h
.....

This Completed is confusing here (one of the containers exits with Error). Why the pod status is Completed here?

-- injoy
containers
kubernetes
kubernetes-pod

Similar Questions

Kubernetes: How to link to internal services with different namespaces?
Kubernetes kind of objects and Prometheus operator
Wait for kubernetes jobs to finish
Gitlab Runner in Kubernetes Could not resolve host
How to write a json patch so that it creates a file in /spec/kubeadmConfigSpec/files/-"?
ASP.Net Core sidecar logging in Kubernetes with Fluentd problem
Istio1.7 sidecar-injection failure , Timeout exceeded while awaiting headers

1 Answer

10/2/2019

According to the official Kubernetes documentation, Job treats a Pod failed once any of entire containers quit with a non-zero exit code or for some resources overlimits detected. Due to this fact, Pod phase actually is the main indicator in terms of generic Pod lifecycle, telling initial Job about most recent Pod status.

However, what I have observed, that kubectl get pod output STATUS column doesn't display Pod phase status, instead it retrieves a value for a particular container inside a Pod and uses .status.containerStatuses.state.terminated.reason field to display STATUS column data.

In fact, you can even get more informative output, supplying custom columns to the standard kubectl command-line tool:

$ kubectl get po -o=custom-columns=NAME:.metadata.name,PHASE:.status.phase,CONTAINERS:.spec.containers[*].name,STATUS:.status.containerStatuses[*].state.terminated.reason

Since Job reaches backoffLimit: 9 count, it will terminate with BackoffLimitExceeded warning message:

Warning BackoffLimitExceeded 54m job-controller Job has reached the specified backoff limit

-- mk_sta
Source: StackOverflow