docs(release): add the v1.5.4 changelog, and fix v1.6.0's fail-open audit note - #3910
Merged
Conversation
Signed-off-by: cozystack-ci[bot] <274107086+cozystack-ci[bot]@users.noreply.github.com>
cozystack-ci
Bot
requested review from
Aleksei Sviridkin (lexfrei) and
myasnikovdaniil
as code owners
August 19, 2026 09:38
cozystack-ci
Bot
requested a review
from Andrei Kvapil (kvaps)
as a code owner
August 19, 2026 09:38
cozystack-ci
Bot
requested a review
from Timofei Larkin (lllamnyp)
as a code owner
August 19, 2026 09:38
The generated changelog is a good inventory but carries no operator-facing layer: no required-actions section, no runnable pre-upgrade checks, and the SeaweedFS 4.31 rename left as three isolated fix bullets that never say what to do about it. Every generated bullet is preserved; this adds to them. Widen the base to v1.5.2. v1.5.3 was tagged but its release was never published and its changelog PR was closed unmerged, so docs/changelogs/ v1.5.3.md exists on no ref and a v1.5.3 base leaves #3212 (persistent EFI/TPM state) and #3194 (filer postgres2 connection pool) documented nowhere at all. Coverage is now 82/82 commits in range. Correct three claims. The slot 45 divergence is a skip rather than an ordering problem: v1.5.4 is the first 1.5.x stamped targetVersion 46, and run-migrations.sh loops seq CURRENT (TARGET - 1), so a cluster at 46 runs 46..53 on the way to 1.6 and never executes 1.6's own slot 45. Its chart-side half was missing too, and it is not redundant -- a fresh v1.5.4 install is stamped 46 having never run any slot, so only the chart can reach it. The S3 checksum bullet named a barman-cloud plugin path that does not exist on this branch. The audit re-run warning names v1.6.0, not earlier releases. The script does not exist at v1.5.2; v1.5.4 is the first 1.5.x to carry it and carries the fail-closed version. v1.6.0 shipped the copy that silenced every kubectl failure, so a clean table from it is not evidence. Signed-off-by: Myasnikov Daniil <myasnikovdaniil2001@gmail.com>
v1.6.0's own upgrade notes tell operators to run hack/seaweedfs-naming-audit.sh, and the copy that shipped in v1.6.0 silences every kubectl call with 2>/dev/null. A timeout or an RBAC denial therefore prints an empty "all clean" table indistinguishable from a genuinely clean fleet, and that output gates a runbook step that deletes PVCs. Anyone following those notes today can still get a false clean. Verified per tag: v1.6.0 has 11 blanket redirections and no run_kubectl wrapper; v1.6.1, v1.6.2 and v1.5.4 have 4 and 18 respectively, the fail-closed shape from #3436 and its backport #3474. Published changelogs are historical records and are not normally edited. This one is an exception on purpose, because the instruction it carries is still live and still wrong. Signed-off-by: Myasnikov Daniil <myasnikovdaniil2001@gmail.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.
Adds the changelog for
v1.5.4, and corrects one still-live instruction in v1.6.0's.The generated changelog is a good inventory and every one of its bullets is kept. What it had no layer for is the operator: no required-actions section, no runnable pre-upgrade checks, and the SeaweedFS 4.31 rename left as three isolated fix bullets that never say what to do about it. That is what is added on top, in the shape v1.6.0's changelog uses.
Base widened to
v1.5.2. v1.5.3 was tagged but its release was never published and its changelog PR was closed unmerged, sodocs/changelogs/v1.5.3.mdexists on no ref and a v1.5.3 base leaves #3212 (persistent EFI/TPM state) and #3194 (filer postgres2 connection pool) documented nowhere at all. Coverage is 82/82 commits in range.Three claims corrected. The slot 45 divergence is a skip rather than an ordering problem: v1.5.4 is the first 1.5.x release stamped
targetVersion: 46, andrun-migrations.shloopsseq CURRENT (TARGET - 1), so a cluster that reaches 46 runs slots 46 through 53 on the way to 1.6 and never executes 1.6's own slot 45. Its chart-side half was missing too and is not redundant, because a fresh v1.5.4 install is stamped 46 having never run any slot, so only the chart can reach that population. And the S3 checksum bullet named a barman-cloud plugin path that does not exist onrelease-1.5.The audit re-run warning names v1.6.0 rather than an earlier release, because the script does not exist at v1.5.2 at all. v1.5.4 is the first 1.5.x release to carry it, and it carries the fail-closed version.
The second commit edits
docs/changelogs/v1.6.0.md, which is deliberate rather than an accident of scope. Published changelogs are historical records and are normally left alone, but v1.6.0's upgrade notes still tell operators to runhack/seaweedfs-naming-audit.sh, and the copy that shipped in v1.6.0 silences everykubectlcall with2>/dev/null. A timeout or an RBAC denial therefore prints an empty all-clean table indistinguishable from a genuinely clean fleet, and that output gates a runbook step that deletes PVCs. Verified per tag: v1.6.0 carries 11 blanket redirections and norun_kubectlwrapper, while v1.6.1, v1.6.2 and v1.5.4 carry 4 and 18, the fail-closed shape from #3436 and its backport #3474.