lablup / lablup/backend.ai

Build Agent task management structure

Open
#11,650 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
670
Forks
183
Avg merge
17h 7m
Merged PRs (30d)
358

Description

Set up the Agent lifecycle handler framework in agent/stage/ that sequences the existing (currently unwired) Provisioner/Stage drafts into the live create_kernel path. The Handler base exposes execute() -> response (NO event emission). CreateTaskInfo tracks per-stage status (current stage, start time). The local phase is progress-tracking only, not authoritative state.

Retry/timeout/reschedule are NOT owned by the Agent - they belong to Manager sokovan. This framework does not classify retry/give-up.

Event emission is handled by the separate transition-event layer (BA-6851), not by this framework or the handlers.

Success Criteria:
- [ ] Handler base defined with execute() -> response interface (no event emission)
- [ ] CreateTaskInfo tracks current stage and per-stage timestamps
- [ ] agent/stage runner sequences the stage drafts; no agent-side retry/give-up classification
- [ ] No regression in existing create_kernel() behavior
- [ ] pants test passes for affected packages

JIRA Issue: BA-6075

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.