memory: snapshot — run 7 training on gx10; run 6 TRANSFERRED after CSAM adjudication; erp-tune-v6-nvfp4a16 live as trial; ESH/YTVC/webhook repairs; ana-ml2 routes persisted; tank/zroot actions deferred to next session. Index 830→271 lines: 27 decisions + 8 abandoned archived, superseded in-flight blocks archived verbatim

This commit is contained in:
vh
2026-09-09 00:26:44 -07:00
parent 3e18a044bd
commit 5ad948bf31
23 changed files with 2951 additions and 2800 deletions
@@ -1,34 +0,0 @@
# [2026-08-23] Every secret-bearing `.env` on ana-docker tightened to 0600
Found while taking uptime ownership of hrafn: its `.env` was mode 0644 with a live
bearer token. Not a hrafn lapse — **0644 was the de facto pattern on the host**.
Eight stacks carried secret-shaped vars in world-readable `.env` files on a box with
four interactive accounts, verified as real exposure by reading one as `nobody`.
Swept: **vaultwarden, traefik**, beszel, gitea-runner, miniflux, news-digest,
searxng, vor. (hrafn and nevermore were fixed separately the same day.) Six other
stacks already used 0600, so this converged on the existing house pattern rather
than inventing one. Post-sweep the host has **zero** secret-bearing `.env` readable
by `nobody`.
Playbook: `playbooks/tighten-env-perms.yaml`, one run per stack, re-runnable.
## The check that matters
Every run asserts `docker compose config` still renders **as the deploy user**
(`lkraven`), not as root. Checking the mode proves the bits changed; only rendering
as the deploy user proves the next deploy can still resolve its variables.
## Two gotchas recorded in the playbook
- **vaultwarden looked like it bind-mounted its `.env`** — which would mean the
*container's* UID reads it and 0600 could break the password vault. It does not:
that `- .env` is under `env_file:`, not `volumes:`. My grep matched the YAML list
item without checking its parent key. The playbook now **refuses** any stack that
genuinely bind-mounts its `.env`, since that case is read by the container UID.
- **elway prompted for a sudo password.** The `ana-docker` ssh alias resolves to
`lkraven`, who needs one; **`infra-ops@10.250.50.70` has NOPASSWD**. `corviduo-dev`
was repointed to infra-ops at some point and `ana-docker` was not. Run elway against
the infra-ops target on this host.
Commit `a896c0a`.