Create an instance with Huge pages configuration and identify issues and causes
We started seeing OOM(Out of Memory) kill events for some of the v2gpu nodes. Huge pages is a feature of openstack that improves performance for DGX nodes. We created a flavor gpu2-cpu32-192g-hp
which schedules v2gpu nodes that use memory with large pages configuration.
We need to schedule v2gpus nodes with this flavor and identify any issues/causes that come with it. We expect to see a VM with its RAM using the Huge pages allocation when it gets scheduled.