dispatcher: Windows bootstrap one-job policy, transport fallback, ready-wait knobs; volume probes documented #10

Merged

founder merged 1 commit from claude/w6-windows-bootstrap-policy into main 2026-08-19 06:50:23 +00:00

Conversation 0 Commits 1 Files changed 5 +1117 -34

founder commented 2026-08-19 06:50:19 +00:00

Owner

Copy link

Windows lane hygiene: uniform one-job "no task" handling (exit 75 → guest destroyed in seconds, not ~10 min), bounded by FORGEJO_RUNNER_ONE_JOB_WAIT_TIMEOUT (per-target ExtraEnv), WinRM→ssh-tunnel fallback, README with the measured Windows cache-volume probes (no request shape mounts anything on nsc 0.0.557 / Compute v1beta — consistent with main's "never attach cache volumes on Windows"). Tests: policy (fake transports), e2e fragments, opt-in live probe. go vet + go test ./... green (-race x3).

Companion key.store PR: common.nix sets the wait timeout on the windows targets + docs; then a pin bump + forge deploy (restarts the dispatcher — do it while no Windows job is in flight).

🤖 Generated with Claude Code

Windows lane hygiene: uniform one-job "no task" handling (exit 75 → guest destroyed in seconds, not ~10 min), bounded by `FORGEJO_RUNNER_ONE_JOB_WAIT_TIMEOUT` (per-target ExtraEnv), WinRM→ssh-tunnel fallback, README with the measured Windows cache-volume probes (no request shape mounts anything on nsc 0.0.557 / Compute v1beta — consistent with main's "never attach cache volumes on Windows"). Tests: policy (fake transports), e2e fragments, opt-in live probe. `go vet` + `go test ./...` green (-race x3). Companion key.store PR: common.nix sets the wait timeout on the windows targets + docs; then a pin bump + forge deploy (restarts the dispatcher — do it while no Windows job is in flight). 🤖 Generated with [Claude Code](https://claude.com/claude-code)

founder added 1 commit 2026-08-19 06:50:19 +00:00

dispatcher: Windows bootstrap one-job policy, transport fallback, ready-wait knobs; volume probes documented ... 85ba1bc0b2

Namespace Windows lane hygiene (KSE-0186 D5 follow-up):
- the guest bootstrap treats forgejo-runner's "no task received" (v12.10.2
  exit 2) and "could not fetch task" (v12.7.x) uniformly, bounded by
  FORGEJO_RUNNER_ONE_JOB_WAIT_TIMEOUT (ExtraEnv per target), and reports
  the outcome (exit 75 = no task) so the dispatcher destroys idle guests in
  seconds instead of ~10 min; a failing job no longer trips the "one-job
  failed" throw (forgejo-runner one-job returns nil either way).
- WinRM bootstrap failure falls back to the ssh tunnel transport briefly
  (measured: all 11 launches since 08-17 07:15 used transport=ssh-tunnel
  after 0 s; the "did not answer on winrm" warnings were the old first-try
  path).
- README documents the live volume probes with nsc 0.0.557 / Compute
  v1beta: no Windows cache-volume request shape mounts anything (CLI
  --volume can't express drive letters; API CACHE volumes accepted but the
  instance never leaves PENDING; experimental persistent_volumes RUNNING
  but no disk in the guest) — matches main's "never attach cache volumes
  on Windows" (bcaad3c3).
- tests: bootstrap policy (fake transports), e2e fragments under real
  PowerShell semantics, opt-in live volume probe.

go vet/build/test green (internal/nsc under -race x3); go.mod/go.sum
unchanged.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

founder merged commit 85ba1bc0b2 into main 2026-08-19 06:50:23 +00:00

founder deleted branch claude/w6-windows-bootstrap-policy 2026-08-19 06:50:23 +00:00

Sign in to join this conversation.

Reviewers

No reviewers

Labels

Clear labels

No items

No labels

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

1 participant

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

key-store/forgejo-nsc!10

No description provided.