release-train: staging -> main - #1044
Merged
Merged
Conversation
…plied marker (#1033) On recorded==latest with the workload on :tag, the image-refresh loop could not tell a fresh install from a helm-revert of an applied pin, so every fresh install paid the shared #563 flap path (resource-monitor DaemonSet rollout under tolerations: Exists) plus a Recreate downtime on jobs-manager for byte-identical content. Gate the roll on two durable per-image markers: first-observed-<image> (stamped by first-observation) and digest-applied-<image> (stamped only after a successful re-image). Skip the roll only on positive fresh-install evidence (first_observed present AND applied absent); an established edge (applied present) rolls the client-runtime#199 repair, and a pre-marker/legacy edge (neither marker) also rolls so the existing fleet is not stranded on :tag on the upgrade hop. Both marker reads and the digest-record read are fail-closed. Split from #1008 (item 2). Bumps chart to 1.9.112.
… same 1.9.112 that #1032 already shipped (#1040) The image-refresh CronJob fix (#1033) bumped Chart.yaml 1.9.111 -> 1.9.112 on its own branch, and so did the control-plane trim (#1032). Both merged; the second bump was a no-op, so develop's chart version now equals staging's while the templates differ, and the chart-version guard rightly refuses the staging promotion. This bump carries #1033 to 1.9.113. No template change.
release-train: develop -> staging
…rol-plane images (#1038) The four tracebloc-published control-plane images (jobs-manager init+api, pods-monitor, requests-proxy, resource-monitor) now pull from ghcr.io by default. The control-plane packages are public on GHCR and published there by digest alongside the Docker Hub copies (the GHCR migration), so this changes where the bytes come from, not which bytes run. images.traceblocRegistry stays the per-edge knob: "docker.io" remains a valid value and is the one-flag rollback (--set images.traceblocRegistry=docker.io; user-supplied, so it survives the fleet auto-upgrade until cleared). global.imageRegistry (a mirror) still wins over everything. Unchanged on purpose: tracebloc/mysql-client (frozen, digest-pinned, Docker Hub only), busybox and the other third-party images, and the training-image host JOB_IMAGE_HOST. No new egress: ghcr.io (+ pkg-containers.githubusercontent.com for layer redirects) is already required for the ingestor image and probed by the installer preflight. The upgrade rolls the control-plane pods once (the image reference changes); IfNotPresent behaves as before from then on. Digest pins are registry-agnostic. The tracebloc.tbRegistry helper's empty/absent-key fallback moves with the default so "empty renders the chart default" (values.schema.json) stays true. That literal restates values.yaml's default unavoidably (a template cannot read chart defaults apart from merged values), so the suite pins both to one value: the chart-default tests read values.yaml, the EMPTY-knob and images-absent tests read the literal (mutation M3 below is what makes that a check). Tests: every default-values assertion on a control-plane image flips to ghcr.io (client_env_alias, global_image_registry, image_refresh, jobs_manager, requests_proxy, resource_monitor, tracebloc_registry). Because ghcr.io is now the default, every override test in tracebloc_registry_test.yaml that set the knob to ghcr.io would have exercised nothing; they now set docker.io (the rollback) or quay.example, and the CronJob's IMAGE_REGISTRY=ghcr.io plus IMAGE_REGISTRY_RESOLVABLE=true are asserted at defaults. mysql-client, busybox, JOB_IMAGE_HOST and every mirror/explicit-knob assertion are untouched. Mutation table (helm unittest client; restored byte-for-byte, re-run 729/729): M1 values default back to "docker.io" -> 21 failed / 729 (every default-value assertion across 7 suites) M2 helper ignores images.traceblocRegistry -> 10 failed / 729 (every real override: 3 re-homes, digest pin, IMAGE_REGISTRY follows, NOTES follows, quay.example inert + 3x Always; the default-pinning tests cannot redden here by construction, the literal IS the default) M3 helper literal drifts to "docker.io" -> 4 failed / 729 (2 EMPTY-knob tests + 2 images-absent tests: the agreement check) Suites: helm unittest 41 suites / 729 tests green; helm lint --strict green on aks/bm/eks/oc values + ingestor; helm template on all four platforms renders ghcr.io/tracebloc/ for every control-plane image and docker.io/tracebloc/ for mysql-client only; make drift 48/48 guards green (incl. check-drift, mirror-enumeration-complete, customer-copy-no-ticket-refs, gate-default-prose); chart-env-vocabulary 177/177; bats scripts/tests/*.bats 1973/1973 (incl. the three image-refresh script harnesses); chart-version-guard bumped 1.9.112 -> 1.9.113 (version + appVersion). MIGRATION.md gains an "Upgrading to 1.9.113" entry; INSTALL.md's egress table and registry paragraph name ghcr.io as the default and docker.io as the rollback.
release-train: develop -> staging
Contributor
Author
|
bugbot run |
There was a problem hiding this comment.
✅ Bugbot reviewed your changes and found no new issues!
Comment @cursor review or bugbot run to trigger another review on this PR
Reviewed by Cursor Bugbot for commit 6acc2f9. Configure here.
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
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.
Automated promotion by the release train (RFC-0008 D14). Head is the train-managed
release-train/to-mainbranch (a mirror ofstaging), so it never collides with a human PR. Merged only when the fr-gate is green.Note
Medium Risk
Changes default control-plane image references (one rollout per edge) and image-refresh rollout logic for fresh vs legacy fleets; mitigated by dual-published digests, rollback knob, and extensive tests.
Overview
Chart 1.9.113 switches the default registry for tracebloc control-plane images (jobs-manager, pods-monitor, resource-monitor, requests-proxy) from Docker Hub to
ghcr.ioviaimages.traceblocRegistryandtracebloc.tbRegistry. Image-refresh’sIMAGE_REGISTRYand install docs/egress tables follow that default;docker.ioremains the documented per-edge rollback, and digest pins /global.imageRegistrybehavior are unchanged.The image-refresh CronJob (#1008) changes when it re-pins workloads that still run a floating
:tagwhile the recorded digest matches latest: it addsfirst-observedanddigest-applieddeployment annotations so genuine fresh installs skip the expensive re-pin rollout and stay on:taguntil an upstream digest change, while established edges with a reverted pin still repair, and pre-marker legacy edges get one repair roll on upgrade so the fleet is not left on a stale tag. Digest annotation reads are fail-closed on kubectl/jq errors instead of being treated as “first observation.”Helm unit tests and
image-refresh-repin-on-revert.batsare updated for the new defaults and marker gating.Reviewed by Cursor Bugbot for commit 6acc2f9. Bugbot is set up for automated code reviews on this repo. Configure here.