[Roadmap] Q3 2025

[peterschmidt85]

This issue outlines the major items planned for Q3 2025:

• major [Feature]: Scheduled tasks #2836
• major [Feature]: Allow to configure a service healthcheck #2181
• major [Feature]: NVIDIA GPU passive health checks #2930
• major For fleets, support ranges (0.., 1..10, 1.., etc) in addition to a fixed number of nodes #1448
• medium Performance-related issues
• medium [Feature]: Generate Docker registry password using AWS creds #1868
• medium [Feature]: Fleet Retry #2921
• medium Forbid fleet auto-creation
• medium Batch cluster provisioning #1489
• medium [Feature] Support repos in run configurations (and deprecate dstack init) #2851 (single repo only)
• medium K8S improvements
• GPU quotas
• Sub-projects
• AMD metrics exporter (similar to DCGM)
• Named SSH volumes, see [Feature]: Volume lifecycle management #2835
• GPU healthcheck (active)
• GCP A3 Ultra/A4 instances
• AWS P6 instances
• [Feature] Improve working_dir behaviour #2857
• [Feature]: Support IAP tunneling when connecting to GCP VMs #2554
• [Feature]: Allow to use SSH via AWS Session Manager #2562
• [Feature]: Use GCP IAM when pulling docker images from Google Artifact Registry #2556
• Hot Aisle
• DigitalOcean
• CoreWeave

Note

Not all items may be completed in Q3—any unfinished items will be moved to Q4.

Warning

If you notice a critical issue that’s not listed above and is blocking you from using dstack, please share it in the comments along with a link to the corresponding issue.