Issues

• InferenceCluster conditions should report progressive status until they're Ready

  enhancement
  modelplaneai/modelplane#603 · Filter by author RedbackThomson opened on Oct 30, 2026

• AMD GPU pools get whole-node scheduling because the cluster agent only tolerates the NVIDIA taint

  enhancement
  modelplaneai/modelplane#601 · Filter by author negz opened on Oct 30, 2026

• GKE GPU node pools come up untainted, so system pods schedule onto GPU nodes and scale them up

  bug
  modelplaneai/modelplane#599 · Filter by author negz opened on Oct 30, 2026

• InferenceClass doesn't include the fabric its devices need

  enhancement
  modelplaneai/modelplane#598 · Filter by author negz opened on Oct 29, 2026

• Document running an InferenceGateway as a Claude Code gateway in front of Anthropic

  enhancement
  modelplaneai/modelplane#597 · Filter by author dennis-upbound opened on Oct 29, 2026

• An InferenceGateway can't serve Claude Code on people's own Claude subscriptions

  enhancement
  modelplaneai/modelplane#596 · Filter by author dennis-upbound opened on Oct 29, 2026
  1

• Serve /anthropic from vLLM's native Messages API for ModelDeployments

  enhancement
  modelplaneai/modelplane#595 · Filter by author dennis-upbound opened on Oct 28, 2026

• Vultr: VKE rejects the hard-coded system pool plan vc2-6c-16gb with "Invalid NodePool plan"

  modelplaneai/modelplane#589 · Filter by author ldalorion opened on Oct 28, 2026
  1

• EKS's default system pool is too small for its own serving stack, so an nvsentinel pod never schedules

  bug
  modelplaneai/modelplane#586 · Filter by author dennis-upbound opened on Oct 27, 2026

• An InferenceCluster reports Ready while its serving stack's pods can't schedule

  bug
  modelplaneai/modelplane#585 · Filter by author dennis-upbound opened on Oct 27, 2026

• Foreground deletion of an InferenceCluster can finish before its cloud resources are deleted

  bug
  modelplaneai/modelplane#582 · Filter by author negz opened on Oct 26, 2026

• EKS InferenceCluster node pools can only use a region's first three Availability Zones

  bug
  modelplaneai/modelplane#581 · Filter by author negz opened on Oct 26, 2026

• One node pool that can't provision blocks the serving stack for the whole InferenceCluster

  bug
  modelplaneai/modelplane#580 · Filter by author negz opened on Oct 26, 2026

• EKS InferenceCluster reports only Creating when a node pool can't get instances

  bug
  modelplaneai/modelplane#579 · Filter by author negz opened on Oct 26, 2026

• E2E Clouds placement audits can run before the serving stack's pods exist

  bug
  modelplaneai/modelplane#569 · Filter by author negz opened on Oct 26, 2026

• Count tool calls whose arguments don't match the request's tool schema, at the gateway

  enhancement
  modelplaneai/modelplane#565 · Filter by author dennis-upbound opened on Oct 25, 2026

• Serving-path components on an InferenceCluster's system nodes run as one pod with no disruption budget, so node maintenance takes a model offline

  bug
  modelplaneai/modelplane#554 · Filter by author dennis-upbound opened on Oct 21, 2026
  1

• InferenceCluster node pools have an EKS-only capacityBlock field

  enhancement
  modelplaneai/modelplane#553 · Filter by author negz opened on Oct 21, 2026
  2

  

• Detect a PrefillDecode deployment that reports Ready but doesn't disaggregate

  modelplaneai/modelplane#552 · Filter by author dennis-upbound opened on Oct 21, 2026

• Stop installing a Prometheus on every inference cluster now that the collector exports metrics

  enhancement
  modelplaneai/modelplane#551 · Filter by author dennis-upbound opened on Oct 21, 2026
  2

• A MetricMapping can't keep one of a metric's series, so SGLang's running and waiting counts double

  enhancement
  modelplaneai/modelplane#549 · Filter by author dennis-upbound opened on Oct 21, 2026

• Map vLLM's KV offload and speculative decoding metrics, and GPU interconnect fields, by default

  modelplaneai/modelplane#547 · Filter by author dennis-upbound opened on Oct 20, 2026

• Changing an engine from Standalone to Leader/Worker leaves the old Deployment running, holding its GPUs

  bug
  modelplaneai/modelplane#545 · Filter by author sumbry opened on Oct 20, 2026

• API keys on an InferenceGateway can't be limited to specific models

  enhancementRouting
  modelplaneai/modelplane#541 · Filter by author pluna opened on Oct 19, 2026
  2

• Telemetry docs describe labels and aggregation the collector doesn't produce

  modelplaneai/modelplane#537 · Filter by author dennis-upbound opened on Oct 17, 2026