feat(deploy): Telegram + SNS alert on 3 canary rollback paths (L221 — 2/5)#184
Merged
Conversation
Independent-channel surveillance on the 3 canary failure paths (inference / regime / regime-eval) so a silent auto-rollback is no longer the failure mode. 2-day silent rollback chain in alpha-engine-data #274 retrospective is the recurrence class. The regime + regime-eval alerts include the names + versions of the upstream Lambdas already promoted at that point — operator triage information critical when an image-wide bug surfaces only on the later canary. Best-effort; trailing || true never overrides exit 1. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Merged
3 tasks
cipher813
added a commit
that referenced
this pull request
May 22, 2026
…b pin v0.17.0→v0.24.0 (#185) Adds ``--dedup-key "canary-fail-${LAMBDA}-v${VERSION}"`` to all three canary-failure alert publish calls in ``infrastructure/deploy.sh`` (inference Lambda + regime Lambda + regime-eval Lambda). An image-wide rebuild that breaks N Lambdas' canaries within the hour now collapses to one alert per (Lambda, version) — closes the N-emails-for-one-event class on the 2026-05-21 L221 canary-alert surveillance shipped 5/21 late evening (PR #184 merged 23:43Z). Originally filed as a P1 owed before the 4 L221 canary PRs merged; those PRs merged 5/21 23:42-23:51Z so this becomes a post-merge retrofit per the ROADMAP's downgrade rule. Lib pin v0.17.0 → v0.24.0 in lockstep across requirements.txt + requirements-lambda.txt (drift between them shipped a stale lib to prod 2026-05-07 per the existing comment). v0.24.0 ships the ``--dedup-key`` CLI flag this retrofit relies on; v0.18-0.23 additive features come transitively. Suite: 1131 passed. Composes with [[reference_alpha_engine_lib_alerts_v0_24_0_dedup]] (consumer migration #4 of the v0.24.0 substrate). Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
2/5 of the L221 fleet pass. Predictor's `infrastructure/deploy.sh` has 3 canary failure-exit paths — inference (L266), regime (L375), regime-eval (L473) — each gets a `python3 -m alpha_engine_lib.alerts publish` before `exit 1`.
The regime + regime-eval alerts include the names + versions of the upstream Lambdas already promoted at that point, since an image-wide bug surfacing on the later canary requires operator triage of the already-live siblings.
Implementation
3 minimal inserts; `|| true` ensures the alert publish never overrides the deploy's `exit 1`. Lib alerts v0.21.0 already in requirements.
Fleet pass scope
Test plan
🤖 Generated with Claude Code