# Panfactum Changelog — edge.24-10-21

> Improves Karpenter node size management with instance type anti-affinity and maximum size limits, fixes EBS volume mount performance, and addresses issues with Cilium memory limits and Ingress NGINX on large CPU nodes.

## Highlights

- `instance_type_spread_required` renamed to `instance_type_anti_affinity_required` in all submodules
- New `max_node_memory_mb` and `max_node_cpu` limits in `kube_karpenter_node_pools` (defaults: 64GB/32 CPUs)
- Fixed slow EBS PV mount times — fix applies to new PVs only

## Breaking Changes

- In all Panfactum submodules, `instance_type_spread_required` has been renamed to `instance_type_anti_affinity_required` as TopologySpreadConstraints were replaced with AntiAffinity rules to work around
- Two new variables for [`kube_karpenter_node_pools`](https://panfactum.com/docs/edge/reference/infrastructure-modules/direct/kubernetes/kube_karpenter_node_pools), `max_node_memory_mb` and `max_node_cpu`, limit the maximum size of provisionable nodes. Default limits are 64GB of
  - Impacts: iac-module `kube_karpenter_node_pools` — Added max_node_memory_mb and max_node_cpu limits (64GB/32 CPUs default)

## Fixes

- Prevents Karpenter from scheduling on bare metal instances, which have caused observed issues.
  - Impacts: iac-module `kube_karpenter_node_pools` — Bare metal instances excluded from provisioning
- Removes memory limits on the Cilium node agent in [`kube_cilium`](https://panfactum.com/docs/edge/reference/infrastructure-modules/direct/kubernetes/kube_cilium) as they could cause Cilium to fail on larger node
  - Impacts: iac-module `kube_cilium` — Removed memory limits on node agent to prevent failures on large nodes
  - Reference (issue-report): [Cilium memory limits cause failures on larger nodes](https://github.com/Panfactum/stack/issues/227)
- Upgrades [`kube_ingress_nginx`](https://panfactum.com/docs/edge/reference/infrastructure-modules/direct/kubernetes/kube_nginx_ingress) so it can run on nodes with a large number of CPU cores.
  - Impacts: iac-module `kube_ingress_nginx` — Fixed compatibility with high-CPU-count nodes
  - Reference (issue-report): [Ingress NGINX fails on high-CPU-count nodes](https://github.com/Panfactum/stack/issues/227)
- EBS-backed PVs with many large files now mount nearly instantly by adding the recommended remediation for a Bottlerocket OS issue.
  - Impacts: iac-module `kube_aws_ebs_csi` — Fixed slow mount times for PVs with many large files

## Related Resources

- [JSON Data](https://panfactum.com/docs/changelog/edge.24-10-21.json): Machine-readable data
- [Channel Release List](https://panfactum.com/docs/changelog/edge.json): All releases in this channel