Issues

• Refresh worker dependencies and roll out NICLLS across main and self-managed 1.0.x

  NVIDIA/nvcf#2424 · Filter by author sbaum1994 opened on Oct 9, 2026

• Add Spark recipes for Nemotron Nano 12B v2, Nemotron 3.5 Lightning and Qwen3.5-4B

  NVIDIA/nvcf#2411 · Filter by author jalajthanakinvidia opened on Oct 9, 2026

• nats-auth-callout: NATS connection loss is invisible to logs, metrics, and health checks

  NVIDIA/nvcf#2409 · Filter by author along-2017 opened on Oct 9, 2026

• nats-auth-callout: an nkey entry overwrites the account name for every later plugin on that account

  good-first-issueinfrastructure
  Bug

  NVIDIA/nvcf#2399 · Filter by author CloneOfAlex opened on Oct 8, 2026
  1

• function-autoscaler: DistributedLockGuard drop deletes the lock without checking ownership

  good-first-issueinvocation-plane
  Bug

  NVIDIA/nvcf#2398 · Filter by author CloneOfAlex opened on Oct 8, 2026
  1

  

• function-autoscaler: stickiness windows longer than the 10-minute lookback silently disable scaling

  good-first-issueinvocation-plane
  Bug

  NVIDIA/nvcf#2397 · Filter by author CloneOfAlex opened on Oct 8, 2026
  1

• self-managed: admin-token-issuer-proxy /v1/admin/keys is routed on the shared gateway with no caller authentication

  good-first-issueinfrastructure
  Bug

  NVIDIA/nvcf#2396 · Filter by author CloneOfAlex opened on Oct 8, 2026

• grpc-proxy: rate-limit requests omit clientAuthSubject and ownerNcaId, so per-user and owner-account limits never apply to gRPC invocations

  invocation-plane
  Bug

  NVIDIA/nvcf#2395 · Filter by author CloneOfAlex opened on Oct 8, 2026
  

• ratelimiter: Olric counter increment is a non-atomic Get+Put, so concurrent checks lose counts and over-admit

  good-first-issue
  Bug

  NVIDIA/nvcf#2394 · Filter by author CloneOfAlex opened on Oct 8, 2026
  

• test(llm-request-router): investigate failing load-balancer render check

  NVIDIA/nvcf#2369 · Filter by author sbaum1994 opened on Oct 8, 2026

• refactor(event-ledger): unify per-request identity into a single typed context value

  NVIDIA/nvcf#2367 · Filter by author shelleyshen-0 opened on Oct 8, 2026

• docs: model cache SDD and Helm model caching page are behind the implementation

  NVIDIA/nvcf#2365 · Filter by author balajinvda opened on Oct 8, 2026

• nvca: expose worker-init download concurrency and chunk size as a typed cluster setting

  NVIDIA/nvcf#2363 · Filter by author balajinvda opened on Oct 8, 2026

• NVCA operator repeatedly rolls out the agent when config keys differ only by case

  bugcompute-stackneeds-triage
  NVIDIA/nvcf#2355 · Filter by author mikeyrcamp opened on Oct 7, 2026

• llm-routing: expose a common hardware-aware recipe catalog

  NVIDIA/nvcf#2337 · Filter by author priyaselvaganesan opened on Oct 7, 2026

• llm-routing: separate shared infrastructure and independent model recipes

  NVIDIA/nvcf#2336 · Filter by author priyaselvaganesan opened on Oct 7, 2026

• llm-routing: preflight and download Jobs cannot be rerun after a failure

  bugllm-stackneeds-triage
  NVIDIA/nvcf#2335 · Filter by author kristinapathak opened on Oct 6, 2026

• llm-routing: runtime guard exits after start when the cgroup swap file is missing

  bugllm-stackneeds-triage
  NVIDIA/nvcf#2334 · Filter by author kristinapathak opened on Oct 6, 2026

• llm-routing: an interrupted model download resume can fail permanently and leave an oversized partial file

  bugllm-stackneeds-triage
  NVIDIA/nvcf#2333 · Filter by author kristinapathak opened on Oct 6, 2026

• llm-gateway-stack: demo UI key check passes when the gateway uses an external API-key Secret

  bugllm-stackneeds-triage
  NVIDIA/nvcf#2332 · Filter by author kristinapathak opened on Oct 6, 2026

• llm-routing: run model recipes on any ARM64 GPU cluster, including single-node GB300

  enhancementllm-stack
  NVIDIA/nvcf#2330 · Filter by author kristinapathak opened on Oct 6, 2026
  

• Add optional monitoring for the LLM routing demo

  NVIDIA/nvcf#2324 · Filter by author priyaselvaganesan opened on Oct 6, 2026

• nvca: shared-claim writer Job with repeated FailedCreate and no active pod is never bounded

  NVIDIA/nvcf#2323 · Filter by author balajinvda opened on Oct 6, 2026
  2

• Scope the OpenBao init-cluster Job Secret RBAC to the Secrets it manages

  NVIDIA/nvcf#2316 · Filter by author sunilthorat09 opened on Oct 6, 2026

• OpenBao init Job fails: manual raft join conflicts with retry_join and cannot recover on retry

  NVIDIA/nvcf#2313 · Filter by author sunilthorat09 opened on Oct 6, 2026