The issue we are seeing is our K8s AWS cluster will auto-scale to be able to handle a pipeline's workflow. The cluster then see's that it can scale down, so it performs a scale down and evicts the pod. Harness then treats this as a failure instead of restarting the step.

The ability under failure conditions for a pod eviction to tell harness to restart would also be helpful.

We would like the option to have a ci pipeline restart when a pod is evcited. A pod getting evicted by k8s is a typical scenario with node autoscaline or failover. K8s and the k8s autoscaler expects a pod to restart itself. We would like the option to have a failure strategy that allows for a ci pipeline to get restarted at the last failed step or completely restarted when a pod gets evicted.

Harness team.. As adviced. Pls also provide timeline these feature releases as well. Thanks

adding 4th item for auto retries

--------extract from Request #72479 ------------

Lijo Jacob

Further checking about this internally, we have been informed that the failure types delegate restart and delegate provisioning errors are not supported with CI stage execution. The team will work on removing these options from the failure strategy of the CI stage until this is supported.

Can we also add this request to the enhancement request created earlier to detect the other failure types?

Regarding the usage of the delegate restart failure strategy in the custom stage, we got the same result when tested on the internal account. We are currently checking with the internal team to get more details about this. Will get back to you with an update shortly

Thanks,

Lijo Jacob || Customer Success Engineer

Any updates Harness team? Could you pls revert by 14/11 COB. Thanks