Repository Issue Activity (beta)

#9569 Support of deploy on docker based existing machines

#9461 Better DAG/Parallel support

#8413 [Test] Add a test for logs download functionality in the dashboard

#10457 [Slurm] Transient empty status query can delete live cluster metadata

#10658 CNCF AI Technical Community Group Discussion/Presentation

#10520 [API server] Client with DB-backed config serializes config path as JSON "null" -> override_skypilot_config crashes every request with TypeError (list + NoneType)

#9455 [Storage] Support AWS S3 Files (NFS access to S3 buckets)

#9421 Allow setting registry for us-docker.pkg.dev/sky-dev-465/skypilotk8s/skypilot:latest

#8562 serve: autoscaler `latest_version` resets on controller restart, causing scale churn

#10409 Loopback auth exemption is reachable via kubectl port-forward on Kubernetes deployments

#10491 [k8s] network_tier: best sizes EFA from a warm node of a different accelerator, making the pod unschedulable

#10506 /validate fails with Workspace 'default' not found for non-admin users — RBAC-gated workspace check reached with an unresolved workspace

#10507 /validate: 'Workspace not found in SkyPilot configuration' error message is misleading when access-filtered list is empty

#10547 [API Server] API Server leaks one fd per request log — unlinked logs hold GiBs until the pod restarts, wedging the server (0.13.0)

#6820 Flaky case test_multi_echo on remote API server test

#10444 [SERVE] Load balancing failure to decrement on errors and additional concerns

#10641 [Kubernetes] Ingress port cleanup stops at the first already-deleted service and leaks the rest

#10639 [API Server] add job-level metrics with job ids and names

#10636 SSH proxy websocket handler blocks the event loop on a synchronous `import kubernetes`

#10628 Autostop: support an admin-controlled idleness provider

#5041 [k8s] Propagate pod termination signals to the job

#9499 [Cost-Report] Make cost-report export into csv format

#10588 [Core] With slurm.submit_as_user enabled, the enabled-clouds cache becomes per-user for ALL clouds — new users get "Kubernetes is not enabled" until they manually run `sky check`

#9733 GPU mismatch between sky launch and Vast

#9478 [Admin Policy] Users can bypass server-side autostop policy via post-launch `sky autostop`