Zhanghao Wu comments

Results 315 comments of


                                            Zhanghao Wu

Revamp installation page to be more concise

This is awesome @SeungjinYang! I think for the first section, we can add a one line cloud installation after the `pip install skypilot`, so that people who are just trying...

[UX] Allow k8s CPU/memory info in sky show-gpus and show CPU/memory on infra dashboard

@codex review

FUSE performance is worse on Kubernetes compared to VM deployment

To be clear, it seems the performance is already better than GKE with `smarter-device-plugin`. How does it compare with GKE builtin solution?

[Core] Add User Specific Exit Codes for Retry

/quicktest-core /quicktest-core --base-branch releases/0.10.5

[API Server] Local API Server is too Resource Intensive

Thanks for raising this @kyuds! This is a known issue as we were trying to mimic the pre-API server case, i.e. there is no upper limit for the requests you...

Fix remote workdir still having deleted files

> That's a good point. There are definitely possible race conditions if two concurrent launches happen with the same workdir. I hadn't anticipated that particular workflow when putting this together....

Fix remote workdir still having deleted files

Thanks for the proposals @dan-blanchard! Keep track of a single `upload_id` for the upload and the later launch/exec request makes sense to me. I am a bit confused for why...

Fix remote workdir still having deleted files

/smoke-test --kubernetes --remote-server /smoke-test --remote-server

Fix remote workdir still having deleted files

Another quick question, since we have a new folder for launch request now, should we clean up those folders once they are used once?

Fix remote workdir still having deleted files

> My intention was for `cleanup_upload_ids` to periodically clean up the uploaded files like it was cleaning up the chunks. When I went to double-check that this was happening, I...