Kubernetes pods does not work on "Pod sandbox, changed, it will be killed and recreated"

In the Google Container Engine (GKE) cluster, I sometimes see a block (or more) that does not start and is not viewed in its events, I see the following

The Pod sandbox has changed; it will be killed and recreated.

If I wait, he just keeps trying again.
If I remove the module and allow them to be recreated using the deployment replica set, it will start correctly.

The behavior is inconsistent.

Kubernet options 1.7.6 and 1.7.8

Any ideas?

+4
source share
2 answers

I see the following message sent to Google Cloud Status Dashboard :

" , Google Container Engine (GKE), node, .

, GKE, Kubernetes v1.6.11, v1.7.8 v1.8.1.

: v1.6.11, , 1.6.10. v1.7.8, , 1.1.6. v1.8.1, , 1.7.6.

. , ".

+5

node GKE 1.8.1 ( ). :

  • , node , , node. , node 1.
  • node :

    kubectl drain <node>
    

    , .

  • node Compute Engine. GKE node, .

+2

Source: https://habr.com/ru/post/1687779/


All Articles