Skip to content

docs(lore-0088): covering-operator backfill handoff runbook - #154

Merged
karczuRF merged 6 commits into
developfrom
docs/0088_backfill-handoff-covering-operator
Jul 24, 2026
Merged

docs(lore-0088): covering-operator backfill handoff runbook#154
karczuRF merged 6 commits into
developfrom
docs/0088_backfill-handoff-covering-operator

Conversation

@karczuRF

Copy link
Copy Markdown
Collaborator

What

Adds docs/runbooks/backfill-handoff-covering-operator.md — a self-contained handoff for a stand-in operator (no prior project context) to run the SDEX historical backfill (task 0088) to completion while the primary operator is away.

Why

Pass 1 finishes ~2026-07-27/28; pass 2 + pre-roll + cleanup re-enable will likely need to happen during a coverage window. This gives a covering operator an unambiguous, checkpointed procedure.

Contents

  • Live status (frontier 38,719,999 @ 2026-07-24, ~145k ledgers/hr, ETA ~07-27/28).
  • Full sequence: monitor pass 1 → clear markers → pass 2 [1, 23,423,999] → incremental pre-roll → re-enable cleanup.
  • Every step tagged by machine ([LAPTOP] vs [FISHUSER-HERO]) with ✅ checkpoints.
  • The cleanup-stays-DISABLED-until-the-end invariant surfaced repeatedly.
  • Access prerequisites incl. the fishuser-hero private-LAN caveat (the one thing that can block pass 2).
  • Commands lifted from the verified runbooks (preroll-incremental-presoroban.md, fix-backfill-history-loss-and-rerun.md); no secrets included.

Verification status

Docs only. Kept open intentionally per the "keep docs PRs open until verified" rule — the underlying pass-1/pass-2/pre-roll work is not yet complete in prod. Merge once the recovery has actually run.

karczuRF added 6 commits July 24, 2026 14:21
Step-by-step handoff for a stand-in operator with no project context to
monitor pass 1 to completion, run pass 2 and the incremental pre-roll,
and re-enable cleanup while the primary operator is away.

Records live status (frontier 38,719,999 @ 2026-07-24, ETA ~07-27/28),
tags every step by machine (laptop vs fishuser-hero), and adds ✅
checkpoints plus the cleanup-stays-disabled invariant.
Phase 2 detection query now emits a status column that reads
'IN PROGRESS' or 'DONE AND READY TO START PHASE 2' (plus a gaps-present
guard), so the covering operator does not have to interpret raw
frontier numbers.
Split the launch step: new §6.3 prepares the binary — check the repo
exists (clone if not), prefer the existing pass-1 binary, and only if
missing checkout develop, pull, build, and verify build success — then
§6.4 launches. Renumber the monitor step and fix the §9 cross-ref.
Add a warning that Phase 4 must run only after BOTH passes finish, and
turn the pre-flight into a hard gate: a verdict query that confirms the
full [1, 50457423] span is present and contiguous before pre-rolling,
so incomplete history is never baked in before cleanup re-enables.
Run tmux creation and the backfill process as two separate steps done
one at a time: Step 1 creates the session (with why tmux is needed —
survives SSH disconnect over the ~4-5 day run), Step 2 runs the process
inside it. Spell out detach (Ctrl-b then d) and reattach.
Marker progress alone does not prove candles landed (a marker can
outlive a dropped partition). Add §4.5 with four prod-verified checks —
candle-vs-marker gap (the key one), landing trend, year coverage, and a
value eyeball — plus a healthy reference reading. Reference it from the
pass-2 monitor step (§6.5).
@karczuRF
karczuRF merged commit 43ee993 into develop Jul 24, 2026
3 checks passed
@karczuRF
karczuRF deleted the docs/0088_backfill-handoff-covering-operator branch July 24, 2026 12:54
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant