Skip to content

move to production#1694

Open
portersupport wants to merge 40 commits into
productionfrom
main
Open

move to production#1694
portersupport wants to merge 40 commits into
productionfrom
main

Conversation

@portersupport
Copy link
Copy Markdown
Contributor

No description provided.

MauAraujo and others added 23 commits April 2, 2026 15:06
* Support cloudsql.instances field for service-connection-managed proxy config

* Add PR check tests for CloudSQL legacy and instances-only paths

* Enforce single service account for CloudSQL proxy; document legacy fields

* Consolidate cloudsql.instances into cloudsql.connections across all charts

* Remove instances[] from worker and job values.yaml; promote connections[]

* Fix mixed-scenario CI tests and stray instances[] comment references

* add multi-instance tests

---------

Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* Support cloudsql.instances field for service-connection-managed proxy config

* Add proxyVersion flag to support upgrading to Cloud SQL Proxy v2

* Flip proxyVersion logic: v2 is opt-in, v1 is default

* Simplify proxyVersion helpers and remove dead code

* Flip proxyVersion conditions: if v2 use v2, else v1

* Fix whitespace in terminationGracePeriodSeconds conditionals

* Consolidate proxyVersion CI tests: 27 steps → 12, delete 9 testdata files

* Revert unrelated terminationGracePeriodSeconds whitespace changes

* Rename cloudsql.proxyVersion to cloudsql.version


---------

Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
fix: prevent karpenter from evicting rollout-sync pods
)

Adds pod- and container-level securityContext rendering from
.Values.podSecurityContext / .Values.securityContext in the job chart's
cronjob.yaml and hook-configmap.yaml, matching the existing behavior in
the web and worker charts. This allows users to set seccompProfile (and
other security context fields) on job workloads.

Co-authored-by: David Townley <davidtownley@Davids-MacBook-Pro.local>
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
…or, and logdna

Pins SHA256 image digests to prevent silent tag mutation and supply chain
attacks. For logdna, also adds sha field support to the daemonset template
since the chart previously had no way to specify a digest.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Exposes image.repository and image.tag in values so callers can override
both. Required for pinning the k8s-dns-node-cache image by digest in the
porter monolith (RUN-2547). Defaults preserve the current image reference.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Trigger the chart-publishing workflow off pushes to main instead of
pushes to a separate production branch. The production branch was a
manual deploy gate that drifted from main over time (auto-bumped
Chart.yaml versions on production stopped flowing back to main earlier
this year), and the merge-to-production PRs were just bulk replays
rather than selective filtering.

Bump main's Chart.yaml versions to match what is currently live in
ChartMuseum (last published from production HEAD) so the next publish
increments from the right base.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Revert the version bumps on worker and the three addons. Those charts
have no functional drift from live (only whitespace), so bumping them
would trigger pointless republishes on first run. Keep the bumps on web
(ships #1690 karpenter rollout-sync fix) and job (ships #1693 job
securityContext support), which do have pending unreleased changes
sitting on main.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The staging→production promotion step described here no longer exists
after switching the publish workflow to trigger on merge to main.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
build.sh queries ChartMuseum and minor-bumps from there, overwriting
whatever's in Chart.yaml. There's no manual-override path without
modifying the script itself.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The auto-commit step pushes Chart.yaml version bumps directly to main.
The default GITHUB_TOKEN authenticates as github-actions[bot], which
cannot be added to main's "Allow specified actors to bypass required
pull requests" list (the bot identity is not selectable). Switch the
build job's checkout to use a portersupport PAT so the push happens
under that account, which can be added to the bypass list.

Requires: PORTERSUPPORT_PAT repo secret containing a fine-grained PAT
for the portersupport account, scoped to this repo with contents:write.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Add [skip ci] to the auto-commit message so the PAT-authenticated push
back to main doesn't retrigger the workflow. Without this, every chart
publish would loop: bump → push → retrigger → bump → push → ...

The GITHUB_TOKEN loop protection that prevented this on the production
branch doesn't apply to PAT pushes — PAT pushes are treated like any
human push and do trigger workflows. [skip ci] is GitHub Actions'
documented marker for opting out of that.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds helm-template assertions to pr_checks.yaml for the five addons whose
images were digest-pinned (grafana, redis, datadog, tailscale-operator,
logdna). Each step renders the chart and greps for `<repo>@sha256:` —
loose check by design, so legitimate digest bumps don't require CI updates.

Catches: digest accidentally cleared, template not wiring the digest field
through, future chart syncs wiping pinned values.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
`echo "$OUT" | grep -q` produces false positives on large helm outputs
under bash's `set -e -o pipefail` (GitHub Actions default): grep exits
early on match, echo gets SIGPIPE (141), pipefail surfaces that as the
pipeline's exit code, and the `||` error path fires even though grep
matched. Datadog's render is large enough to hit this reliably.

Switching to `grep -qE "pattern" <<< "$OUT"` eliminates the pipe and the
race.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
feat: pin image digests for Tier 1 addon charts
georgez98 and others added 2 commits May 12, 2026 13:10
PR #1695 added pinned image digests to the values.yaml of datadog,
grafana, logdna, redis, and tailscale-operator, but landed on main
five minutes before the workflow trigger was switched from production
to main (#1697). As a result those changes never made it into
ChartMuseum.

Bump each affected chart so the next publish run detects a diff in
the chart directory and republishes. The version numbers here are
just to force the diff — build.sh queries ChartMuseum for the live
version and minor-bumps from there, then overwrites Chart.yaml.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…-digests

chore: republish tier 1 addons to ship pinned image digests
rudimk and others added 3 commits May 15, 2026 14:26
Signed-off-by: Edwin Hermans <edwin@madtech.cx>
…y_worker

Allow the worker deploymentstrategy to be defined like the web service
madeddie and others added 3 commits May 22, 2026 21:10
Sync vendored karpenter chart to upstream v1.12.1 to pull in the latest
version.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
chore: upgrade karpenter 1.8.2 -> 1.12.1
madeddie added 2 commits June 3, 2026 15:19
Like command, entrypoint and healthcheck customer sometimes want to use
the shell form of lifecycle commands
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.

9 participants