Hi everyone, The automatic termination of idle Kubernetes Computing Units (CUs) feature has been implemented and merged into Texera's main branch, as shown in PR #6046 <https://github.com/apache/texera/pull/6046>, following the design discussion in #6264 <https://github.com/apache/texera/discussions/6264>.
This feature introduces a periodic cleanup mechanism to identify and terminate Kubernetes CUs that have remained idle beyond a configurable timeout, helping reduce unnecessary resource consumption. The implementation includes: - A scheduled background task that periodically checks for idle Kubernetes CUs. - Idle detection based on workflow execution activity, ensuring that CUs with active executions are not terminated. - Configurable idle timeout and cleanup intervals through environment variables. - Automatic Kubernetes pod termination and corresponding database state updates. One remaining limitation has also been identified in the following issue #8618 <https://github.com/apache/texera/issues/8618> that I opened: a CU may not be reclaimed if an associated workflow execution remains stuck in a non-terminal state. I've also discussed possible improvements there. Thanks, Yichen Ren
