Issues

• Nebius cloud half installs DCGM exporter, so modelplane_gpu_* and energy metrics are empty

  bug
  modelplaneai/modelplane#583 · Filter by author ldalorion 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

• 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

• 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
  1

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

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

• Gateway retries, timeouts and ejections are collected and then dropped

  modelplaneai/modelplane#536 · Filter by author dennis-upbound opened on Oct 17, 2026

• Collect the gateway's usage records

  modelplaneai/modelplane#535 · Filter by author dennis-upbound opened on Oct 17, 2026

• Gateway metrics can't be grouped by deployment or service

  modelplaneai/modelplane#534 · Filter by author dennis-upbound opened on Oct 17, 2026

• No per-deployment request outcome by HTTP status

  modelplaneai/modelplane#533 · Filter by author dennis-upbound opened on Oct 17, 2026

• Run an InferenceGateway on a cluster without GPUs

  enhancementProvisioning
  modelplaneai/modelplane#532 · Filter by author pluna opened on Oct 16, 2026

• The first model on an InferenceGateway reports RoutingReady while the gateway restarts and returns HTTP 500

  bug
  modelplaneai/modelplane#531 · Filter by author dennis-upbound opened on Oct 16, 2026
  4

• Run a model from its ModelCache without contacting Hugging Face

  Cachingenhancement
  modelplaneai/modelplane#530 · Filter by author pluna opened on Oct 16, 2026

• New GPU node on AKS offers no GPUs to engines, so the model stays Pending

  bugProvisioning
  modelplaneai/modelplane#528 · Filter by author pluna opened on Oct 16, 2026
  1

  

• No validation of the InferenceCluster name length

  bug
  modelplaneai/modelplane#527 · Filter by author haarchri opened on Oct 16, 2026