🌟 Highlights

• InstanceSet Workload: Introduced native support for InstanceSet, enabling fine-grained management for stateful workloads.
• Coordinated Rollout: Implemented synchronized updates across interdependent roles to maintain system stability during upgrades.
• In-Place Updates: Enabled efficient updates by using InstanceSet workloads without requiring pod recreation.
• Revision Management: Added full ControllerRevision support for robust version tracking and rollback capabilities.

🚀 Features

• InstanceSet Support
  • Introduced InstanceSet API and Controller.
  • Added support for using InstanceSet as a Role Workload in RoleBasedGroup (RBG).
• Controller Revision & Versioning
  • Added ControllerRevision support to RoleBasedGroup for better version control.
  • Updated rbgctl to support RBG revision operations.
  • Enabled updating roles using controllerrevision hash.
• Orchestration & Scheduling
  • Implemented rollout coordination for roles in RBG.
  • Supported parallel execution for roles sharing the same dependencies.
  • Added support for Pod Group Policy.
• Runtime & Integrations
  • Added Mooncake integration.
  • Added Engine Runtime support.
  • Supported sgl-router PD (Prefix Disaggregation) with engine runtime.
• Miscellaneous
  • Added support for role-level metadata.
  • Migrated Template, leaderWorkerSet, and RollingUpdate fields from value to pointer types.

🐛 Bug Fixes

• StatefulSet & Reconciler
  • Fixed sts_reconciler failure when retrieving historical StatefulSet revisions.
  • Updated StatefulSet service naming to meet Kubernetes requirements (compatible headless service name).
• Workload & Updates
  • Fixed an issue where updating a role did not trigger a rolling update.
  • Fixed leader pods with InstanceSet workload lacking environment variables.
  • Added max length checks for workloadName and serviceName.
• Cleanup & Resources
  • Fixed issue where corresponding PodGroups created by RBG were not deleted during gang scheduling.
  • Fixed incorrect apiVersion in auto-generated applyconfigurations.
  • Fixed InstanceSet RBAC and LWS environment build issues.

📖 Documentation

• Added missing model examples.
• Updated Gang Scheduling documentation.
• Updated sglang PD disaggregation example with sgl-router.
• Fixed link errors in README and added CI checks for markdown quality.

What's Changed

• fix build image error in Makefile by @gujingit in #41
• feat: Support parallel execution for roles with same dependencies by @tlipoca9 in #42
• Build: update rbg helm chart by @cheyang in #44
• feat: Supports using controllerrevision hash to update role by @bcfre in #34
• feat: add instanceset api by @veophi in #50
• doc: add missing model examples by @bcfre in #49
• feat: use code-generator to generate applyconfiguration code by @liubing0427 in #48
• Build docker image for supporting controller revision by @cheyang in #51
• [WIP]: Add in-place update api and core codes for InstanceSet by @veophi in #52
• fix: change stateful set service name to meet k8s requirements by @liubing0427 in #53
• feat: add engine runtime by @gujingit in #55
• bugfix: add max len check for workloadName & serviceName by @gujingit in #58
• fix: delete corresponding podgroup created by rbg when gang-schedulin… by @ShirleyDing in #57
• Update Helm chart 0.5.0-alpha.1 by @cheyang in #60
• fix: address review comments in pr-57 by @bcfre in #67
• doc: update gang scheduling by @bcfre in #62
• feat: The ControllerRevision not store the replicas for RBG Roles by @bcfre in #61
• CI:add release script by @cheyang in #71
• feat: rbgctl supports rbg revision operations by @bcfre in #54
• KEP-8: Reduce YAML Duplication via RoleTemplates by @LikiosSedo in #70
• Release 0.5.0-alpha.2 by @cheyang in #73
• feat: add instance controller by @yangsoon in #66
• fix: sts_reconciler failure to retrieve historical StatefulSet revisions by @bcfre in #77
• fix(sts-reconciler): use compatible headless service name for statefu… by @TrafalgarZZZ in #81
• [KEP-31]: Adding ControllerRevision support to the RoleBasedGroup by @bcfre in #27
• Release 0.5.0-alpha.3 by @cheyang in #78
• feat(engine-runtime): support sgl-router pd disaggregation with engine runtime by @TrafalgarZZZ in #82
• KEP-30: add role coordination kep by @gujingit in #59
• feat: add instanceset controller codes by @veophi in #83
• KEP-8: Refine API naming and preview/diff design based on community feedback by @LikiosSedo in #79
• [KEP-30]: Introduce InstanceSet Workload Support in RoleBasedGroup for Improved LLM Orchestration by @veophi in #26
• Fix(rbg): update role not trigger rolling update by @Syspretor in #89
• doc: Update sglang pd disaggregation example with sgl-router by @TrafalgarZZZ in #86
• KEP 74: Mooncake integration by @Syspretor in #75
• chore(codegen): Generate instanceSet related go-client codes by @Syspretor in #93
• chore(hack): add script to generate manifests.yaml by @Syspretor in #95
• feat(rbg): add rbac with resource instancesets/instances by @Syspretor in #97
• Build: update rbg helm chart 0.5.0-alpha.4 by @cheyang in #98
• fix(rbg): fix incorrect apiVersion in auto-generated applyconfigurations by @Syspretor in #101
• docs: fix link error in README; fix lint errors in markdown files; add a CI to automatically check markdown quality by @Phil-Fan in #100
• feat: impl rollout coordination for roles in rbg by @veophi in #91
• fix: add explicit permissions to the docs-check workflow by @Phil-Fan in #105
• rbgs support pod group policy by @nightmeng in #107
• refactor: migrate Template field from value to pointer type by @LikiosSedo in #102
• chore(rbg): migrate role.leaderWorkerSet field type from value to poi… by @Syspretor in #108
• feat(rbg): support role-level metadata by @Syspretor in #113
• feature(rbg): support to use InstanceSet as role workload by @Syspretor in #110
• chore(rbg): migrate RollingUpdate field type by @Syspretor in #114
• Update readme by @cheyang in #111
• fix(rbg): fix instanceset rbac and lws env build by @Syspretor in #116
• fix(rbg): fix leader pod with instanceset workload lack envs by @Syspretor in #122
• Build: update rbg helm chart 0.5.0 by @cheyang in #123

New Contributors

• @tlipoca9 made their first contribution in #42
• @veophi made their first contribution in #50
• @liubing0427 made their first contribution in #48
• @ShirleyDing made their first contribution in #57
• @LikiosSedo made their first contribution in #70
• @yangsoon made their first contribution in #66
• @TrafalgarZZZ made their first contribution in #81
• @Syspretor made their first contribution in #89
• @Phil-Fan made their first contribution in #100
• @nightmeng made their first contribution in #107

Full Changelog: v0.4.0...v0.5.0