prole/mock_val/onboard_engineer.sh
chrisfu 3943d1b298 feat(mock_val): add diagnostics, utilities, and misc operational scripts
Diagnostics:
  diag_gitlab_boot.sh, diag_gitlab_webservice_oom.sh, diag_gke_storage.sh

Utilities:
  ensure_default_storage_class.sh — set/verify default StorageClass
  preflight_kubecontext.sh        — validate kubecontext before ops
  onboard_engineer.sh             — new engineer onboarding script
  gen_oidc_signing_key.sh         — generate OIDC signing key
  fetch_prole_secrets.sh          — pull secrets from vault
  set-k3s-token-1password.sh      — store k3s token in 1Password
  sync_cnpg_grafana_dashboard.py  — sync CNPG dashboard to Grafana

Config/certs:
  krb5.local.conf, knoe-db-ca.crt

Updated: build-a-bao.sh, hostprobe-*.yaml, hosts.txt, knoe-db-passwwd.sh,
         repair_pipeline.sh, status.sh, status_common_services.sh
Co-authored-by: Junie <junie@jetbrains.com>
2026-05-23 21:31:40 -07:00

238 lines
7.6 KiB
Bash
Executable File

#!/usr/bin/env bash
# onboard_engineer.sh — provision a per-engineer postgres role + 24h temp
# password, then emit a one-time onboarding URL (and QR-code rendering of it)
# that the new engineer redeems at https://db.0.knoe.dev/onboard.html.
#
# This is the Phase 1 bridge while Junie's Phase 2 (libpq OAUTHBEARER) is in
# flight. When OAUTHBEARER lands, the temp-password mechanism goes away and
# this script's role-creation step is replaced by an INSERT into
# `knoe.oauth_role_map`. The engineer-facing URL stays the same (the page just
# stops showing a password and instead displays the OAUTH connection string).
#
# Usage:
# ./etc/onboard_engineer.sh <username> <email>
# ./etc/onboard_engineer.sh --revoke <username>
#
# Env (resolved from knoe_cfg.sh / KUBECONTEXT etc):
# DB_CLUSTER_KUBECONTEXT — the CNPG cluster context (knoe-dev-cnpg-0)
# DB_NAMESPACE — defaults to knoe-db-0
# DB_PRIMARY_POD — auto-discovered if unset
# ONBOARD_HOST — defaults to db.0.knoe.dev
# ONBOARD_TTL_HOURS — defaults to 24
set -euo pipefail
SCRIPT_DIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
REPO_ROOT=$(cd "$SCRIPT_DIR/.." && pwd)
DB_CTX="${DB_CLUSTER_KUBECONTEXT:-${KUBECONTEXT:-gke_plenary-truck-485623-p7_us-west3_knoe-dev-cnpg-0}}"
DB_NS="${DB_NAMESPACE:-knoe-db-0}"
ONBOARD_HOST="${ONBOARD_HOST:-db.0.knoe.dev}"
ONBOARD_TTL_HOURS="${ONBOARD_TTL_HOURS:-24}"
log() { printf '[%s] %s\n' "$(date +%H:%M:%S)" "$*"; }
warn() { printf '\033[33m[!]\033[0m %s\n' "$*" >&2; }
err() { printf '\033[31m[ERROR]\033[0m %s\n' "$*" >&2; exit 1; }
usage() {
cat <<EOF
Usage:
$(basename "$0") <username> <email> # provision new engineer
$(basename "$0") --revoke <username> # drop role + cancel access
$(basename "$0") --rotate <username> # generate a fresh 24h temp password
Examples:
$(basename "$0") ron ron@knoey.com
$(basename "$0") --revoke test-user
EOF
exit 1
}
primary_pod() {
if [[ -n "${DB_PRIMARY_POD:-}" ]]; then printf '%s' "$DB_PRIMARY_POD"; return; fi
kubectl --context="$DB_CTX" -n "$DB_NS" get pod \
-l cnpg.io/cluster=knoe-db,cnpg.io/instanceRole=primary \
-o jsonpath='{.items[0].metadata.name}'
}
run_psql() {
local sql="$1"
local pod
pod=$(primary_pod)
printf '%s' "$sql" | kubectl --context="$DB_CTX" -n "$DB_NS" exec -i "$pod" -c postgres -- \
psql -U postgres -d postgres -v ON_ERROR_STOP=1 -X --quiet 2>&1
}
ensure_dependencies() {
command -v kubectl >/dev/null 2>&1 || err "kubectl not found"
command -v openssl >/dev/null 2>&1 || err "openssl not found"
command -v base64 >/dev/null 2>&1 || err "base64 not found"
if ! command -v qrencode >/dev/null 2>&1; then
warn "qrencode not installed — output will be URL-only (brew install qrencode for QR rendering)"
fi
}
current_iso8601() {
date -u "+%Y-%m-%dT%H:%M:%SZ"
}
ttl_iso8601() {
# macOS date and GNU date have different flag syntax; both ISO 8601 a
# given offset-from-now expressed in hours. Use python as a portable fallback.
python3 - <<EOF
from datetime import datetime, timedelta, timezone
print((datetime.now(timezone.utc) + timedelta(hours=$ONBOARD_TTL_HOURS)).strftime("%Y-%m-%dT%H:%M:%SZ"))
EOF
}
validate_username() {
local u="$1"
[[ "$u" =~ ^[a-z][a-z0-9_-]{1,30}$ ]] || err "username must match [a-z][a-z0-9_-]{1,30}: $u"
}
validate_email() {
local e="$1"
[[ "$e" =~ ^[^@[:space:]]+@knoey\.com$ ]] || err "email must be *@knoey.com (got: $e). To override, edit the script."
}
action_provision() {
local user="$1" email="$2"
validate_username "$user"
validate_email "$email"
local pw exp ttl_hint
pw=$(openssl rand -base64 24)
exp=$(ttl_iso8601)
ttl_hint="${ONBOARD_TTL_HOURS}h"
log "==> provisioning postgres role: $user (email: $email, valid for $ttl_hint)"
# SQL is idempotent: works for fresh CREATE and re-onboarding.
local sql
sql=$(cat <<SQL
DO \$do\$ BEGIN
IF NOT EXISTS (SELECT FROM pg_roles WHERE rolname='$user') THEN
CREATE ROLE "$user" LOGIN INHERIT VALID UNTIL '$exp';
RAISE NOTICE 'created role %', '$user';
ELSE
RAISE NOTICE 'role % exists; updating password and VALID UNTIL', '$user';
END IF;
END \$do\$;
ALTER ROLE "$user" PASSWORD '$pw' VALID UNTIL '$exp';
GRANT knoe_developer TO "$user";
\\echo '=== verification ==='
SELECT rolname, rolcanlogin, rolvaliduntil
FROM pg_roles WHERE rolname='$user';
SELECT 'is_developer' AS check, pg_has_role('$user','knoe_developer','MEMBER') AS ok;
SQL
)
run_psql "$sql"
# URL-safe base64: replace + with -, / with _, drop = padding.
local pw_url
pw_url=$(printf '%s' "$pw" | base64 | tr -d '\n' | tr '+/' '-_' | tr -d '=')
# Wait — pw is already base64 from openssl rand. Encode the *literal text* of
# the password in url-safe base64 so the page can decode and display it
# exactly as typed back into psql. atob() in the browser handles both
# standard and url-safe base64 (after replacement).
local url
url="https://${ONBOARD_HOST}/onboard.html#user=${user}&pw=${pw_url}&exp=${exp}"
printf '\n'
log "==> onboarding link"
printf '\n %s\n\n' "$url"
if command -v qrencode >/dev/null 2>&1; then
log "==> QR code (scan with engineer's phone camera)"
printf '\n'
qrencode -t ANSI256 -m 1 "$url"
printf '\n'
fi
log "==> engineer's plaintext password (fallback if URL is unusable)"
printf '\n %s\n\n' "$pw"
cat <<EOF
==================================================================
DELIVERY:
• Best: Show the QR above on screenshare while $user scans with phone.
Phone opens URL → page shows password + connection string.
• Async: Compose Gmail to $email, paste the URL above, send.
Recipient clicks → page shows password + connection string.
• Last: Hand the plaintext password over a secure channel (Signal).
Less ideal — the page bundles the connection-string UX.
THE TEMP PASSWORD EXPIRES IN ${ttl_hint}.
Engineer should rotate immediately on first connect via: postgres=> \\password
==================================================================
EOF
}
action_revoke() {
local user="$1"
validate_username "$user"
log "==> revoking role: $user"
local sql
sql=$(cat <<SQL
DO \$do\$ BEGIN
IF EXISTS (SELECT FROM pg_roles WHERE rolname='$user') THEN
REVOKE knoe_developer FROM "$user";
DROP ROLE "$user";
RAISE NOTICE 'dropped role %', '$user';
ELSE
RAISE NOTICE 'role % does not exist; nothing to do', '$user';
END IF;
END \$do\$;
SQL
)
run_psql "$sql"
log "==> revocation complete"
}
action_rotate() {
local user="$1"
validate_username "$user"
local pw exp pw_url url
pw=$(openssl rand -base64 24)
exp=$(ttl_iso8601)
log "==> rotating temp password for $user (valid ${ONBOARD_TTL_HOURS}h)"
local sql
sql=$(cat <<SQL
DO \$do\$ BEGIN
IF NOT EXISTS (SELECT FROM pg_roles WHERE rolname='$user') THEN
RAISE EXCEPTION 'role % does not exist; use provision action instead', '$user';
END IF;
END \$do\$;
ALTER ROLE "$user" PASSWORD '$pw' VALID UNTIL '$exp';
SQL
)
run_psql "$sql"
pw_url=$(printf '%s' "$pw" | base64 | tr -d '\n' | tr '+/' '-_' | tr -d '=')
url="https://${ONBOARD_HOST}/onboard.html#user=${user}&pw=${pw_url}&exp=${exp}"
printf '\n %s\n\n' "$url"
if command -v qrencode >/dev/null 2>&1; then
qrencode -t ANSI256 -m 1 "$url"; printf '\n'
fi
printf '\n plaintext: %s\n\n' "$pw"
}
# --- main ---
ensure_dependencies
case "${1:-}" in
""|"-h"|"--help") usage ;;
"--revoke") [[ $# -eq 2 ]] || usage; action_revoke "$2" ;;
"--rotate") [[ $# -eq 2 ]] || usage; action_rotate "$2" ;;
*) [[ $# -eq 2 ]] || usage; action_provision "$1" "$2" ;;
esac