randomparity / randomparity/kdive

Hosted TCG gate: measure ppc64le build-fs wall-time and set a real 30-min-vs-actual timeout

Open
#1,327 1 comment 0 reactions 0 assignees View on GitHub
area:build-install effort:S provider:powervm risk:daytime-only status:ready type:test
Dominant language
Python
Stars
0
Forks
0
Avg merge
1h 26m
Merged PRs (30d)
311

Description

## Context

Follow-up to #1320 (superseded). During the #1319–#1323 campaign, #1320's chosen approach — fetch a prebuilt kdive-ready ppc64le rootfs to avoid emulating the build under TCG — was found **infeasible**: no published, cross-run-fetchable artifact exists. Every `rootfs_catalog.toml` source is a `cloud-image` (upstream base needing build-fs customization), the warm store is host-local (unreachable from the ephemeral hosted TCG runner), and the stack S3/MinIO is per-run ephemeral. Standing up a persistent artifact + publish pipeline is effort:L and a phantom feature until backed by a real artifact.

The concrete adjacent bug #1320 surfaced — the `${TCG_IMAGE_INPUT:-fedora-ppc64le}` schedule/push fallback resolving to an invalid catalog name, plus a fail-loud guard in `produce_rootfs_and_kernel` — was **already fixed and merged in #1324**.

What remains unanswered is #1320's actual headline: **does ppc64le build-fs under TCG emulation complete within the gate's timeout?**

## Scope (ADR-0389 Task 7)

1. Run the hosted `tcg` job in `.github/workflows/live.yml` once (nightly or manual `workflow_dispatch`) and **measure** the wall-time of the ppc64le customize-boot build-fs + proof-boot path under TCG.
2. Replace the placeholder `timeout-minutes: 30` (`.github/workflows/live.yml` ~line 49, "30 is the floor default until measured") with a **measured** value plus headroom.
3. If the measured time exceeds a tolerable CI budget, reconsider the fetch-prebuilt redesign — but only paired with a separate issue to stand up the artifact publish pipeline (do not build fetch infra with nothing to fetch).

## Acceptance criteria

- [ ] A recorded wall-time measurement of the ppc64le TCG build-fs+boot path from a real gate run.
- [ ] `live.yml` timeout set from that measurement (not the placeholder floor).
- [ ] ADR-0389 Task 7 marked done, or a documented decision to pursue the publish-pipeline redesign instead.

## Notes
- Requires a live hosted-TCG run; not resolvable from a CI-only code change.
- Refs #1320, #1324. ADR-0389, ADR-0391.

Contributor guide

Open the contributing guide

Research direction

Run the hosted tcg job in .github/workflows/live.yml through nightly execution or workflow_dispatch, then measure the ppc64le customize-boot build-fs and proof-boot wall-time. Replace the placeholder timeout-minutes: 30 with the measured value plus headroom, and mark ADR-0389 Task 7 done or document the separate publish-pipeline decision if the budget is exceeded.

Written by the indexing model from the issue text.

Assessment

Tech stack
github-actions
Domain
ci-cd
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.