Skip to content

envoy: move to distroless-v1.39.6 for the TLS and CA fixes since 1.39.1 - #42

Open
TakGN wants to merge 1 commit into
mainfrom
envoy-1-39-6
Open

TakGN wants to merge 1 commit into
mainfrom
envoy-1-39-6

Conversation

@TakGN

@TakGN TakGN commented Nov 17, 2026

Copy link
Copy Markdown
Collaborator

Comes out of a read-only security review of the boxes (2026-11-16), alongside ipni/infra#275 and ipni/infra#276.

Why

All four Envoys on sing-1, lith-1 and chic-1 run envoyproxy/envoy distroless-v1.39.1 (2026-08-27). They terminate TLS for the internet on 4443 (bootstrap WSS), 2053 (content gateway) and 9443 (pinning API), and for Cloudflare on 443 (route-origin). Fixed since, in the 1.39 line:

Release Fix Applies here?
1.39.2 BoringSSL CVE-2026-35189: excessive allocation parsing certificates with certain CRL distribution points, a remote DoS during the TLS handshake Yes, every listener here does TLS
1.39.3 GHSA-3wvj-j6qm-62mr (critical, Brotli decompressor), GHSA-8vc2-jrm4-835w (oauth2), GHSA-47vj-9r25-wv5j (api_key_auth) No: none of those filters is in our configs (we use basic_auth, local_ratelimit, router, tcp_proxy)
1.39.4 Bug fixes (RDS route-config races) Not that I can see
1.39.5, 1.39.6 Images stop trusting the compromised eMudhra intermediate CAs (1.39.6 fixes the distroless image, which 1.39.5 missed) Only for upstream TLS; ours goes to loopback

Staying on 1.39: 1.40 is out (1.40.2), but a minor version is more than a security bump should carry.

Change

  • k8s/route-origin/kustomization.yaml, k8s/bootstrap-wss/kustomization.yaml.j2, k8s/content-gateway/kustomization.yaml, k8s/content-pin/kustomization.yaml: digest sha256:eb2c01c1… to sha256:1cac8b3a2f72d5a0918f54d7b49e68aae0b0ba325c2d50c8a9a3fff7a4d1f994, the multi-arch index of envoyproxy/envoy:distroless-v1.39.6 (read from Docker Hub on 2026-11-17), and the comments.
  • README.md: the version in the route-origin section.

Checked

  • scripts/render-manifests.yml on main and on this branch, then kubectl kustomize of every rendered component: the only difference is the Envoy image digest, in route-origin, bootstrap-wss, content-gateway and content-pin.
  • ansible-playbook --syntax-check passes. yamllint and ansible-lint were not run locally; CI runs them.

Not checked: nothing applied; the new image has not run on a box.

Rolling out

None of these playbooks applies from CI, so merging changes nothing on the boxes. One box at a time, dry run first, e.g.:

ansible-playbook routing.yml --check --diff -l sing-1
ansible-playbook routing.yml -l sing-1

then bootstrap.yml and content.yml the same way. Each Envoy pod restarts; needle does not, so there is no needle warm-up. chic-1's bootstrap-b WSS goes with bootstrap.yml too.

🤖 Generated with Claude Code

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@claude

claude Bot commented Nov 17, 2026

Copy link
Copy Markdown

Claude review

This is a pure Envoy image bump from distroless-v1.39.1 to v1.39.6. It swaps one digest, sha256:eb2c01c1… to sha256:1cac8b3a…, in the four kustomizations (route-origin, bootstrap-wss, content-gateway, content-pin), plus the matching comments and the README version. I read the full diff and searched the repo for any other reference to the old digest or to 1.39.x, and found none left behind. The four digests agree with each other. The pin is still by index digest, as the repo convention requires. The PR's claim that it changes nothing on the boxes until routing.yml, bootstrap.yml and content.yml are run by hand matches AGENTS.md, which lists only site.yml, platform.yml and cve.yml as CI-applied. The rollout plan (-l <box>, dry run first, one box at a time) is consistent with the repo's guidance.

No issues found.

Not verified by me: I did not confirm that sha256:1cac8b3a… is the multi-arch index of envoyproxy/envoy:distroless-v1.39.6 on Docker Hub. The reviewer environment had no approved network path to the registry, so I relied on the PR description. Before the first box is rolled, a check like docker buildx imagetools inspect envoyproxy/envoy:distroless-v1.39.6 should show the same digest. A wrong digest would put the Envoy pods into ImagePullBackOff on that box.

claude-sonnet-5-5 at high effort · requested by TakGN · re-run with @claude review [sonnet|opus|fable] [low|medium|high|xhigh|max]

This branch has not been deployed

No deployments
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.

1 participant