generated from amazon-archives/__template_Apache-2.0
-
Notifications
You must be signed in to change notification settings - Fork 70
Open
Description
The EFA plugin has the values.yaml overwritten in the Hyperpod Helm charts and is misconfigured as the values (helm_chart/HyperPodHelmChart/values.yaml) for the supportedInstanceLabels has many instances that actually don't have support for EFA. This causes the plugin to create pods on instances that don't have EFA and those pods go into a continunous crash loop with a message like the following:
│ 2026/02/03 15:23:32 Fetching EFA devices. │
│ 2026/02/03 15:23:32 Error to get IB device: %v. No valid EFA devices found │
│ stream closed: EOF for kube-system/hyperpod-requirements-aws-efa-k8s-device-plugin-9zlf7 (aws-efa-k8s-device-plugin)
Ask: can you correct the list of support EFA instances so the plugin doesn't create pods in nodes which don't support EFA?
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
No labels