build(fleet): advance the two literal image pins to b15 - #284
Merged
Conversation
Three standard classes name the floating alias nddev-ubuntu-24.04-amd64-container-current, so the b15 promotion moved them. Two name the exact image instead -- nddev-priority-standard and nddev-linux-release -- and that is deliberate: the release lane carries OIDC authority and should not follow a promotion silently. Deliberate does not mean automatic. Their scale sets stayed on b14 while the provider policy moved to b15, and the provider then refused every create for both classes: image alias "...-b14" is not the configured alias "...-b15" for pool "nddev-linux-release" nddev-linux-release is the only class permitted tcp/22 egress, so the Almaty staging deploy queued thirty-nine minutes with no runner. Found by that repository measuring completed placements per class -- 56 for standard, 0 for release -- rather than by anything on the fleet side noticing. This is the third step of promote, repin, migrate, and it is the one that makes an exact pin honest: a class that names an image by name has to be told when the image changes. The estate's own contract says the same thing from the other end, requiring every standard pool to bind one promoted fingerprint and one exact N-1 rollback, which a split fleet cannot satisfy. The constants alone change nothing live. Moving the scale sets needs the controller rebuilt and reconcile-garm run with explicit image migration, which disables each scale set while it moves -- so it waits until the release lane is idle rather than interrupting the deploy it just unblocked. Claude-Session: https://claude.ai/code/session_017rA31sF7EzvtbRmi18Uzz9
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.
Three standard classes name the floating alias
nddev-ubuntu-24.04-amd64-container-current, so the b15 promotion moved them. Two name the exact image —nddev-priority-standardandnddev-linux-release— and that is deliberate: the release lane carries OIDC authority and should not follow a promotion silently.Deliberate does not mean automatic. Their scale sets stayed on b14 while the provider policy moved to b15, so the provider refused every create for both:
nddev-linux-releaseis the only class permitted tcp/22 egress, so the Almaty staging deploy queued 39 minutes with no runner. Found by that repository measuring completed placements per class — 56 for standard, 0 for release — not by anything fleet-side noticing.This is the third step of promote, repin, migrate, and the one that makes an exact pin honest: a class naming an image by name has to be told when the image changes. The estate contract says the same from the other end — every standard pool must bind one promoted fingerprint and one exact N-1 rollback, which a split fleet cannot satisfy.
The constants alone change nothing live. Moving the scale sets needs the controller rebuilt and
reconcile-garmrun with explicit image migration, which disables each scale set while it moves — so it waits until the release lane is idle rather than interrupting the deploy it just unblocked.go build,go test ./...clean. Not provider-domain source, so no provider release is required.https://claude.ai/code/session_017rA31sF7EzvtbRmi18Uzz9