6 Pull requests merged by 5 people

• Not mutate azure dep list at runtime

  #4457 merged Dec 11, 2024

• [k8s] Fix show-gpus when running with incluster auth

  #4452 merged Dec 11, 2024

• use lazy import for runpod

  #4451 merged Dec 9, 2024

• [Feature] support spot pod on RunPod

  #4447 merged Dec 9, 2024

• smoke tests support storage mount only

  #4446 merged Dec 9, 2024

• [robustness] cover some potential resource leakage cases

  #4443 merged Dec 8, 2024

7 Pull requests opened by 5 people

• Continue storage deletion when some fail

  #4454 opened Dec 10, 2024

• [Core] Avoid high concurrency issue with control master

  #4455 opened Dec 10, 2024

• add 1, 2, 4 size H100's to GCP

  #4456 opened Dec 10, 2024

• detach the managed job controller from job submission

  #4458 opened Dec 11, 2024

• [core] skip provider.availability_zone in the cluster config hash

  #4463 opened Dec 11, 2024

• [Example] PyTorch distributed training with minGPT

  #4464 opened Dec 12, 2024

• [k8s] Add validation for pod_config #4206

  #4466 opened Dec 12, 2024

21 Issues closed by 4 people

• [Core] Provision an A10 GPU on Azure takes 20 minutes

  #3718 closed Dec 12, 2024

• [UI] Ads on the SkyPilot documentation page

  #4210 closed Dec 12, 2024

• cannot run `sky jobs logs -n <job_name>` on SUCCEEDED job

  #4235 closed Dec 12, 2024

• [UX] Unnecessary logs from ray

  #4300 closed Dec 12, 2024

• [Fluidstack] sky launch can leak instances when instance creation times out

  #4392 closed Dec 12, 2024

• Unable to use NodePort in EKS

  #3805 closed Dec 12, 2024

• [K8s] Cannot `sky show-gpus` for service account

  #4152 closed Dec 11, 2024

• [K8s] list_pod_for_all_namespaces gives ApiException: (403) if the user doesn't have necessary permissions

  #4276 closed Dec 11, 2024

• [k8s] Investigate and document `podPidsLimit` kubelet arg

  #3412 closed Dec 10, 2024

• [Spot] Auto-translated bucket leakage if the spot job is not submitted correctly

  #1225 closed Dec 10, 2024

• Race between status update and instance creation can cause resource leak

  #4431 closed Dec 9, 2024

• [Core] Ray job refused to submit jobs in PENDING status

  #4260 closed Dec 9, 2024

• [Jobs/Core] Leakage of `sky jobs cancel`

  #4410 closed Dec 9, 2024

• runpod 4090 spot not available

  #4265 closed Dec 9, 2024

• Spot instance support for runpod.

  #3927 closed Dec 9, 2024

• `sky check` from one Kubernetes cluster to another failing

  #3904 closed Dec 9, 2024

• [UX] Shortcut `k8s` for `kubernetes`

  #4089 closed Dec 9, 2024

• [k8s] Parallelize pod initialization steps

  #4229 closed Dec 9, 2024

• [k8s] Skip SSH setup for faster provisioning

  #4225 closed Dec 9, 2024

• [k8s] multinode torch distributed nccl timeout

  #3788 closed Dec 8, 2024

• RunPod H100 pricing / catalog needs refresh

  #3794 closed Dec 6, 2024

10 Issues opened by 7 people

• [docs] Make YAML keys referenceable

  #4462 opened Dec 11, 2024

• [k8s] Fail to ssh into the head node on k8s

  #4461 opened Dec 11, 2024

• [Core] Launching on a just launched existing cluster with `--fast` does not skip the provision

  #4460 opened Dec 11, 2024

• [UX] `gh repo clone` fail to work after `gh auth login` on cluster

  #4459 opened Dec 11, 2024

• [Dev] Automatically source the sky environment for dev mode

  #4453 opened Dec 10, 2024

• [UX] Additional message from OCI even though not enabled

  #4450 opened Dec 9, 2024

• [k8s] `sky check k8s` fails when current context is not available, even if that context is not in allowed_contexts

  #4449 opened Dec 9, 2024

• Latest skypilot image does not support azure accelerated networking and nccl

  #4448 opened Dec 8, 2024

• [SERVE][AUTOSCALERS] Replica scaling sampling period and stability.

  #4444 opened Dec 5, 2024

• [SERVE] Allow adjustment of scaling policies without redeployment

  #4442 opened Dec 5, 2024

43 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Support buildkite CICD and restructure smoke tests

  #4396 commented on Dec 12, 2024 • 21 new comments

• Preliminary Vast AI support

  #4365 commented on Dec 12, 2024 • 7 new comments

• Add Envoy as an alternative Sky Serve load balancer implementation

  #4256 commented on Dec 10, 2024 • 4 new comments

• [Release] Release 0.7.1

  #4438 commented on Dec 11, 2024 • 2 new comments

• [DigitalOcean] droplet integration

  #3832 commented on Dec 10, 2024 • 1 new comment

• [k8s] Run GPU labeller automatically on new nodes added to cluster

  #3432 commented on Dec 6, 2024 • 0 new comments

• [k8s] Add validation for `pod_config`

  #4206 commented on Dec 11, 2024 • 0 new comments

• [K8s] Error in k8s secret fetching breaks the provision failover loop

  #4148 commented on Dec 11, 2024 • 0 new comments

• [k8s] Jobs controller on stale context needs better error messages

  #4268 commented on Dec 11, 2024 • 0 new comments

• [k8s] Leaked kubectl port-forward processes

  #4343 commented on Dec 11, 2024 • 0 new comments

• Azure image-id from marketplace with :latest fails

  #4435 commented on Dec 12, 2024 • 0 new comments

• [DeepSpeed Example] Fail on AWS T4 due to package import issue

  #4434 commented on Dec 12, 2024 • 0 new comments

• [UX] Automatically source the skypilot runtime when ssh to the cluster and SKYPILOT_DEV=1

  #4372 commented on Dec 12, 2024 • 0 new comments

• [Core/UX] Improve the display of returncode for multi-node

  #4232 commented on Dec 12, 2024 • 0 new comments

• [UI] Empty Accelerator should raise an issue

  #4153 commented on Dec 12, 2024 • 0 new comments

• [UX] An annoying message in the provision log

  #4102 commented on Dec 12, 2024 • 0 new comments

• [Docs] Add docs for installing SkyPilot with pipx

  #3490 commented on Dec 12, 2024 • 0 new comments

• `sky launch` takes ~5s to print out optimizer table, which is slow

  #3159 commented on Dec 12, 2024 • 0 new comments

• [Core] Allow more PENDING jobs to be scheduled concurrently (1.4x faster)

  #4311 commented on Dec 6, 2024 • 0 new comments

• Mount cached mode

  #4369 commented on Dec 11, 2024 • 0 new comments

• [docs] Change urls to docs.skypilot.co, add 404 page

  #4413 commented on Dec 6, 2024 • 0 new comments

• [Serve] Add and adopt least load policy as default poicy.

  #4439 commented on Dec 10, 2024 • 0 new comments

• [Serve][k8s] K8s replica ports not detected

  #3798 commented on Dec 6, 2024 • 0 new comments

• [Bug][UX] Meaning of `DEVICE_MEM` for multi-GPU instance type is not aligned in `sky show-gpus`

  #3434 commented on Dec 8, 2024 • 0 new comments

• [k8s] Ambiguity when GPU labels overlap with an existing accelerator

  #3562 commented on Dec 9, 2024 • 0 new comments

• decorated functions are not properly typechecked

  #4353 commented on Dec 9, 2024 • 0 new comments

• Run with UV package manager

  #4428 commented on Dec 9, 2024 • 0 new comments

• [AWS] Bucket on eu-south-1 fail to copy/mount

  #3405 commented on Dec 10, 2024 • 0 new comments

• [Storage] `sky storage delete -a` aborted when deletion of one storage failed

  #4050 commented on Dec 10, 2024 • 0 new comments

• [K8s] Add a dedicated doc page for multiple kubernetes

  #4000 commented on Dec 11, 2024 • 0 new comments

• [k8s] Fix /dev/fuse access on Kubernetes

  #4108 commented on Dec 11, 2024 • 0 new comments

• [k8s] Support non-debian custom images

  #4110 commented on Dec 11, 2024 • 0 new comments

• [k8s] Change GPU base image to CUDA `devel` instead of `runtime`

  #4122 commented on Dec 11, 2024 • 0 new comments

• [k8s] Remote identity support when multiple contexts are configured

  #4131 commented on Dec 11, 2024 • 0 new comments

• [k8s] Requesting `--cpus 1.5` and starting a user Ray program crashes

  #4190 commented on Dec 11, 2024 • 0 new comments

• [k8s][gke][dws] autodown not toggled if file sync fails

  #4170 commented on Dec 11, 2024 • 0 new comments

• [k8s] Prevent mounting of /dev/shm in pods

  #4233 commented on Dec 11, 2024 • 0 new comments

• [k8s] Default image cannot install conda package in base env

  #4374 commented on Dec 11, 2024 • 0 new comments

• sky jobs launch on Kubernetes seems not working now

  #4346 commented on Dec 11, 2024 • 0 new comments

• [k8s] Support exec based auth kubeconfigs on controllers

  #4379 commented on Dec 11, 2024 • 0 new comments

• [k8s] L40 GPUs get detected as L4s

  #4404 commented on Dec 11, 2024 • 0 new comments

• [Storage] Refactor S3Store/R2Store to an abstract S3CompatibleStore class

  #2687 commented on Dec 11, 2024 • 0 new comments

• [k8s] Support multiple Kubernetes clusters

  #2937 commented on Dec 11, 2024 • 0 new comments