When using Elastigroup with Kubernetes, pods may remain in a pending state if their node selectors do not align with Elastigroup configuration. This article explains the meaning of the message Pod is pending but its constraints differ from the group labels and how to resolve it.

Understanding the log message

When the Elastigroup autoscaler logs the warning:

WARN, AutoScaler - Attempt Scale Up, Pod <pod-name> is pending but its constraints differ from the group labels. NodeSelector: '{"kubernetes.io/os":"linux"}'

it indicates that the autoscaler cannot scale up because the pod's node selector does not match the custom labels defined at the Elastigroup level.

Why does this happen?

• In Elastigroup, the autoscaler requires the group's configured labels and the pod's nodeSelector to match exactly (excluding built-in labels).

• If the launch specifications have taints, or if the specified labels differ, the pod cannot be scheduled.

• For example:

  • Pod NodeSelector:

    {"kubernetes.io/os": "linux"}

  • Elastigroup labels:

    { "key": "yourdomain.com/application_label", "value": "utility" }

  In this case, the autoscaler needs the pod's nodeSelector to include the group's custom label in addition to any defaults.

How to fix it

1. Identify the custom labels configured in the Elastigroup third-party integration section.

2. Add those labels to the pod's nodeSelector in addition to the existing constraints.

   • Updated NodeSelector:

     {"kubernetes.io/os": "linux", "weedmaps.com/application_label": "utility"}

3. Apply the updated configuration so the pod can be scheduled and the autoscaler can scale up.

Additional information

The autoscaler supports standard Kubernetes built-in labels listed in the Kubernetes labels and annotations reference.