Let an operator authenticate with a seeded key, and reset the schema
CI / chart (pull_request) Successful in 2s
CI / security (pull_request) Failing after 19s
CI / test (pull_request) Successful in 5m34s

- TERDUT_OPERATOR_KEY creates or re-keys the instance-scoped service account
  "terdut-operator" at every start, so terdut-operator needs no bootstrap
  handshake. An instance-scoped account now acts as owner of every team's
  configuration, but is not a member of any team.
- POST /api/teams takes an external_id (instance service accounts only) and
  is idempotent on it, so automation finds its own team again after a crash
  instead of adopting by display name. GET /api/teams?name= is removed.
- Integration and dead man's switch names are unique per team (409). The
  escalation PUT accepts usernames and resolves them itself.
- The 18 migrations are squashed into 001_schema.sql, with no Default team.
  TERDUT_DEADMAN_* and the env seeding of switches are removed: teams carry
  their own. Existing development databases must be recreated.

Security and robustness:
- GET /api/users no longer returns other people's email or ntfy topic to
  non-admins.
- The access log records the route pattern, so integration keys and ack
  tokens in the path are not written to the log. Server errors are logged.
- Rate limits take the client address TERDUT_TRUSTED_PROXIES hops from the
  right of X-Forwarded-For instead of trusting the first, forgeable entry.
- /api/bootstrap runs in a transaction under an advisory lock, so two
  concurrent calls cannot both create an administrator.
- API key last_used_at is written at most every five minutes.

Cleanup: remove GET /api/incidents/{id}/alerts, unused exports, SQLite
remnants in comments and config.

Claude-Session: https://claude.ai/code/session_016mBLURvJoMuUEr9cB2RpUN
This commit is contained in:
Niklas Ye
2026-10-09 14:56:13 +02:00
parent 44b2eb2cc3
commit 9029d48584
46 changed files with 1445 additions and 655 deletions
+19 -10
View File
@@ -5,6 +5,7 @@ import (
"database/sql"
"encoding/json"
"errors"
"github.com/go-chi/chi/v5"
"log"
"net/http"
"strconv"
@@ -17,8 +18,7 @@ import (
// sqlArgs accumulates query arguments and hands back the placeholder for each.
//
// Postgres numbers its placeholders, so a dynamically assembled WHERE clause has
// to keep its $1, $2, … in step with the order of the values — which SQLite's
// positional `?` did for free. Handing out the placeholder and storing the value
// to keep its $1, $2, … in step with the order of the values. Handing out the placeholder and storing the value
// in one call is what keeps them in step: a filter can be added, removed or
// reordered without renumbering anything by hand.
type sqlArgs struct{ vals []any }
@@ -31,8 +31,7 @@ func (a *sqlArgs) add(v any) string {
// addList stores every value and returns their placeholders as "$1, $2, …",
// ready to drop into an IN (…) clause. Returns an empty string for no values,
// which no caller should reach: `IN ()` is a syntax error in Postgres as it was
// in SQLite, so callers check for an empty set before building the query.
// which no caller should reach: `IN ()` is a syntax error in Postgres, so callers check for an empty set before building the query.
func (a *sqlArgs) addList(vs []any) string {
parts := make([]string, len(vs))
for i, v := range vs {
@@ -45,7 +44,7 @@ func (a *sqlArgs) addList(vs []any) string {
func (a *sqlArgs) all() []any { return a.vals }
// nowEpoch is the SQL expression for "now, as unix seconds", matching how every
// timestamp in this schema is stored. SQLite spelled it unixepoch().
// timestamp in this schema is stored.
//
// FLOOR, not a bare cast: EXTRACT returns fractional seconds and casting to
// bigint rounds half up, so a row written at .6 of a second would claim a
@@ -56,11 +55,9 @@ const nowEpoch = "FLOOR(EXTRACT(EPOCH FROM now()))::bigint"
// isUniqueViolation reports whether err is a broken unique constraint, which
// callers turn into 409 Conflict rather than 500.
//
// Postgres reports it as SQLSTATE 23505 on a typed error; the SQLite driver this
// replaced only put "UNIQUE constraint failed" in the message, which is why the
// check used to be a substring match. Matching the code means a renamed
// constraint or a translated message cannot quietly turn a conflict back into a
// 500.
// Postgres reports it as SQLSTATE 23505 on a typed error. Matching the code
// means a renamed constraint or a translated message cannot quietly turn a
// conflict back into a 500.
func isUniqueViolation(err error) bool {
var pgErr *pgconn.PgError
return errors.As(err, &pgErr) && pgErr.Code == pgerrcode.UniqueViolation
@@ -72,6 +69,18 @@ func respond(w http.ResponseWriter, status int, v any) {
json.NewEncoder(w).Encode(v)
}
// serverError answers 500 and logs why. The response stays opaque, so the log
// line is the only record of what failed.
func serverError(w http.ResponseWriter, r *http.Request, err error) {
// The route pattern, not the path: two routes carry a credential in it.
route := r.URL.Path
if rc := chi.RouteContext(r.Context()); rc != nil && rc.RoutePattern() != "" {
route = rc.RoutePattern()
}
log.Printf("%s %s: %v", r.Method, route, err)
respond(w, http.StatusInternalServerError, errResp("internal error"))
}
// maxBodyBytes caps an ordinary JSON request body. 1 MiB is far more than any
// endpoint below needs — it exists so an unauthenticated caller (signup,
// login, bootstrap) can't make the server buffer an arbitrarily large body