Baseline before the clean-up (PD-13): the PD-11 reader setup and the PD-12 app
dashboards as they ran on the local stack, uncommitted until now. Not deployed,
not independently reviewed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
SPEC.md follows the owner's decisions of 24 Oct. The "eligible" baseline state
asks for >= 30 matched days (was 60). Modelled L/h is in v1 as table B11 on
Baseline & Evidence (v_modelled_lph): titled as modelled from an assumed
runtime, not measured, and never in S1, the map or the M3 sites list; the
datasets table and acceptance check 4 include it. The migrations are
unchanged: the threshold is applied by the loader's baseline propose, not by
SQL.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
No runner was ever active, so the pipeline never ran. Removes
.gitea/workflows/deploy.yml and the CI key and runner IP from
host/bootstrap.sh (a re-run removes the key an earlier bootstrap installed).
The README makes "Manual deploy" the one documented procedure: commit on main,
shellcheck in Docker (-S warning; the pipeline's lint step), git archive over
root SSH, deploy.sh as user deploy, the outside checks, then push.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Adds stacks/fuel-monitoring/fuel-db/migrations/0001-0003 with their README, and
petrolyst-insights/dashboards/atoma/SPEC.md. Nothing applies the migrations:
deploy.sh never touches fuel-db, and a person applies them by hand in phase P3.
Verified in scratch Docker (timescaledb 2.31.0-pg17 with this repository's
01-roles.sh, trino 484 with its trino/etc): 150 SQL checks pass; the 18 views
have 303 columns with identical types in PostgreSQL and Trino; fm cannot be
read through Trino. Codex (gpt-6-astra) inspected the change in four rounds;
the round-4 fixes are committed without a further inspection, on the owner's
instruction.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
forwardAuth answers unauthenticated requests itself, so a middleware listed
after it never sees that redirect: the dashboard's 302 to Keycloak went out
without HSTS, nosniff and frame options.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
`sudo -u deploy deploy.sh` (manual deploy, re-deploy, rollback) keeps
root's working directory /root, which deploy cannot read; GNU find then
fails in preflight and the run stops. Every path is absolute, so lib.sh
now changes to / first.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
`pbpaste | ssh petrolyst /root/secrets-init.sh keycloak-secret` sends the
secret without a trailing newline. read then sets the value but returns
non-zero, and set -e ended the script silently before storing it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Traefik as the server-wide edge (TLS via Let's Encrypt, dashboard on
traefik.petrolyst.com behind oauth2-proxy and Keycloak). The
fuel-monitoring stack: Superset on atoma.petrolyst.com with Keycloak login
and Petrolyst branding, Trino reading the TimescaleDB fuel-db as catalog
"fuel". Customization lives in petrolyst-insights/, deployment in the other
folders.
Deploys run server-side (deploy/deploy.sh) from a Gitea workflow or
manually; releases are immutable, Superset metadata is backed up before
migrations, and deploy/rollback.sh restores it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Nasim SadatandClaude Opus 5.5