deployd#3 (DD-0620): every kit for a backend registered since 2026-09-08 died at `--site`.
env-dev-08 (2026-09-11..13): two days of "identity is not readable" ticks — ziti-edge-tunnel
re-creates the file with mode 0600, the ACL mask goes to ---, group membership stops helping.
identity-acl.sh + monky-deployd-identity-acl.path re-apply the grant on every directory change.
Doc-Drift: DD-0620 fixed
Closes#3
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01ASnneBmT7rfaJLE8NGNw7S
Design merged first: monky-design-docs !225 (doc 24 §4a). Pairs with monky-tenancy!40, which
copies the estate-wide read-only Harbor robot into each environment's own prefix and marks the
manifest entry `use: registry-auth`.
- That entry is not an env var (it would otherwise land in .env and therefore in every container's
environment). The agent parses it — JSON, or `username:password` with the new `registry_host` —
and writes `<state_dir>/docker/config.json` 0600 in a directory it owns, with an explicit
DOCKER_CONFIG pointing the docker CLI at it. The unit runs as monky-deployd, so a `docker login`
by a human or by root is invisible to the agent: that is what made env-dev-01 look like it had no
credential at all after the operator had just logged in.
- `compose pull` failures now carry the registry's own message ("no basic auth credentials",
"manifest unknown", DNS) into the journal and the report instead of `rc=1`.
- Tests: both credential shapes, the refusals, 0600/0700 modes, idempotent rewrite, and that the
runner never silently falls back to a human's $HOME.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KLB7jieMNRkTsJ2epr4Ds1
DD-0523 — !7 (31586c30, 0.1.5) moved install.sh and config.example.yaml to the
443 intercept after env-qa-02 hit "service not available". The ansible role
default (monky_deployd_tenancy_port) and TenancyCfg.port still said 8081, so
an ansible-installed box or a config that omits `port` still dialled the wrong
port; both now default to 443, the proxy-mapping and config tests follow, and
PROTOCOL.md §Where and how states the intercept port separately from the
in-pod 8081 and names openziti state/overlay/configs.json as the authority.
DD-0525 — PROTOCOL.md and README said "the broker adds the attr when the
identity is created at kit reveal". monky-ziti at b44c50a4 has no such code
(app/fabric.py host_identity_attrs carries the env template only) and openziti
docs/services.md says "Nothing carries the attr yet". Both now state the
dependency: an operator adds #monky-deploy-agent/#openbao-client on the
controller until the ADR-0028 addendum lands in monky-ziti.
DD-0527 — "the old kit's grant fails at login (unknown/used jti)". The
jwt-tenancy mount keeps no replay state (openbao terraform/jwt-tenancy.tf
see_env role: signature, aud, bound_claims, exp); a superseded grant logs in
until exp and the refusal is tenancy's 401 on the first bearer call. The
second-reveal paragraph, the grant-flow diagram and README §Security model say
so; FakeBao no longer pops a grant at login (the suite's superseded-token test
already goes through FakeTenancy.superseded_jtis, which is the real model).
DD-0528 — "Both locations keep being published": release:gitea has been a
never-run manual job on every tag pipeline (6999, 7044, 7066); README and
OPERATIONS.md now say when the Gitea mirror is published and that it has not
been yet.
Gates (local, py3.12): ruff format, ruff check, pytest 50 passed,
bash -n packaging/install.sh. `git grep 8081` afterwards hits only the in-pod
listener statements.
Doc-Drift: DD-0523 fixed
Doc-Drift: DD-0525 fixed
Doc-Drift: DD-0527 fixed
Doc-Drift: DD-0528 fixed
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AW3QqEpwLV69KHn24Re45Q
mdella
Claude-Docs-Manager