Conversation
|
/help |
GiGL Automation@ 01:17:57UTC : 🤖 Available PR CommandsYou can trigger the following workflows by commenting on this PR:
💡 Usage: Simply comment on this PR with any of the commands above (e.g., ⏱️ Note: Commands may take some time to complete. Progress updates will be posted as comments. |
|
/lint_test |
GiGL Automation@ 01:18:24UTC : 🔄 @ 01:25:59UTC : ❌ Workflow failed. |
kmontemayor2-sc
left a comment
There was a problem hiding this comment.
Is this too complicated? I don't think we should swap things out for users like this. It's not hard to retry the existing run with a new disk config.
If the issue is that we don't expose boot_disk_type to users at all, then why not do that via the VertexAIResourceConfig?
I'd rather this all be explicit and we require users to set the boot disk. I feel like going down the path of doing things for users will overcomplicate our code and have us chasing down GCP requirements
|
Synced offline - I think it's fine to auto-infer this since we never expose disk to user and they don't use disk anyways. But let's add this to |
Scope of work done
If g4 machines are specified,
pd-ssdare not supported.We need to default to
hyperdisk-balancedWhere is the documentation for this feature?: N/A
Did you add automated tests or write a test plan?
Updated Changelog.md? NO
Ready for code review?: YES