Zhanghao Wu

AWS instances e.g. L40S GPUs can come with disks that are not mount to the instance by default. SkyPilot should mount them and when a user specify the docker image...

[Dev] Make cluster/job records a dataclass, instead of a dict

1

Dev: this urges us to make cluster/job records a dataclass, instead of a dict _Originally posted by @Michaelvll in https://github.com/skypilot-org/skypilot/issues/4840#issuecomment-2689210759_

``` ✓ Setup completed. View logs: sky api logs -l sky-2025-06-23-18-32-27-782001/setup-*.log ``` The hint showing above does not work during the jobs controller setup.

Fixes #5994 Tested (run the relevant ones): - [ ] Code formatting: install pre-commit (auto-check on commit) or `bash format.sh` - [ ] Any manual or new tests for this...

[CLI] Speed up `sky status` 40% by reducing the health check

4

This speed up CLI operation on remote API server from (reducing RTT from 14 to 9): Note, the following is with a remote API server and no managed jobs or...

Tested (run the relevant ones): - [ ] Code formatting: install pre-commit (auto-check on commit) or `bash format.sh` - [ ] Any manual or new tests for this PR (please...

A user experience an issue with autodown when launching a large amount of clusters with local API server: some of those clusters fail to have autostop column set. ``` sky...

[k8s] Unexpected error when relaunching an INIT cluster on k8s which failed due to capacity error

1

To reproduce: 1. Launch a managed job with the controller on k8s with the following `~/.sky/config.yaml` ```yaml jobs: controller: resources: cpus: 2 cloud: kubernetes ``` ```console $ sky jobs launch...