SUNK v6.8.0 has been released. This release adds a cleanup script for stuck jobs and nodes, introduces aDocumentation Index
Fetch the complete documentation index at: https://docs.coreweave.com/llms.txt
Use this file to discover all available pages before exploring further.
cache-dropper sidecar for compute pods, updates SCIM parameters to filter inactive users, refreshes NVIDIA device plugin options, includes build-essential in the extras image, and enhances the syncer to reconfigure when nodes are added.
Changes
Cleanup script
- Cleanup script: Script cleans up jobs and nodes stuck in a completing state.
New chart features
cache-droppersidecar is now added to compute pods.- Ability to filter out inactive users by default with SCIM parameters.
- Updated
nvidia-device-pluginsoptions to pass through the list of devices as volume mounts.
syncer updates
The syncer now automatically requests reconfigures when nodes are added to the cluster.
Bug fixes
Charts
- Added support for dependencies in s6 scripts.
- Fixed GPU counting logic to correctly handle cases with 0 GPUs in
run-partsexclusive checks. - Ensured deterministic ordering by sorting keys in s6 helpers.
Image updates
- Patched the Slurm cgroup/v2 plugin to handle a race condition on startup