e07b484a48
Setzt die Foundation aus 1.0.70 fort — bisher war ha_nodes.config_hash
noch NULL und das UI konnte keinen Drift erkennen.
internal/cluster/confighash.go:
- ComputeConfigHash() berechnet SHA-256 (truncated auf 16 hex chars)
über alle replizierbaren Tabellen. Pattern 1:1 aus mail-gateway/
internal/handlers/cluster_status.go (driftHashSpec).
- Pro Tabelle: md5((to_jsonb(t) - id - updated_at - created_at -
excludes)::text) per row, dann string_agg ORDER BY rh.
- Singleton-Tabellen (dns_settings, ntp_settings, mail_config-Stil)
hashen direkt ohne agg.
- 23 Tabellen: domains, backends, backend_servers, routing_rules,
network_interfaces, ip_addresses, tls_certs (mit ExtraExclude
last_renewed_at + last_error damit cert-renewal keinen drift
erzeugt), firewall_zones+address_objects+address_groups+services+
service_groups+rules+nat_rules, wireguard_interfaces+peers,
forward_proxy_acls, dns_zones+records+settings, ntp_pools+settings,
static_routes.
- RefreshLocalHash() schreibt den Hash in die eigene ha_nodes-Row.
Scheduler:
- 5-min-Tick ruft RefreshLocalHash. Pro-Mutation-Refresh wäre zu
teuer (jede UI-Action triggert sonst 23 jsonb-Queries).
- Initial-Refresh beim Scheduler-Boot damit /cluster/status nicht
5 min auf den ersten Wert wartet.
handlers/cluster.go:
- Status() ruft RefreshLocalHash mit 2s-Timeout on-demand. Damit
sieht das UI auch zwischen den Scheduler-Ticks immer frische
Werte; bei Timeout fallback auf den DB-Wert (eventuell stale).
Verifiziert auf 1.0.71: ha_nodes-Row hat config_hash=728834dce5ca4e48,
scheduler-log "config-hash refresh enabled tick=5m0s".
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
60 lines
2.0 KiB
Go
60 lines
2.0 KiB
Go
// Command edgeguard-ctl is the admin CLI for setup, migrations and
|
|
// (later) cluster ops. v1 wires migrate + initdb so postinst can
|
|
// initialise a fresh node; cluster-* and promote remain stubs until
|
|
// Phase 3.
|
|
package main
|
|
|
|
import (
|
|
"fmt"
|
|
"os"
|
|
)
|
|
|
|
var version = "1.0.71"
|
|
|
|
const usage = `edgeguard-ctl — EdgeGuard CLI
|
|
|
|
Usage:
|
|
edgeguard-ctl <command> [args]
|
|
|
|
Commands:
|
|
version Print version and exit
|
|
migrate up Apply pending migrations
|
|
migrate down Roll back the most recent migration (dev only)
|
|
migrate check Validate embedded migrations (no DB connect)
|
|
migrate dump [dir] Write embedded SQL files to dir (default: ./migrations)
|
|
initdb Create PostgreSQL role + database (idempotent)
|
|
render-config Regenerate haproxy / nftables configs from PG (--no-reload, --only=)
|
|
wg-import [--path <dir>] Import existing /etc/wireguard/*.conf files into the DB
|
|
cluster-join Join an existing cluster (Phase 3, not yet implemented)
|
|
promote Promote this node's PG to primary (Phase 3, not yet implemented)
|
|
dump-config Print effective config (Phase 3, not yet implemented)
|
|
`
|
|
|
|
func main() {
|
|
if len(os.Args) < 2 {
|
|
fmt.Fprint(os.Stderr, usage)
|
|
os.Exit(2)
|
|
}
|
|
switch os.Args[1] {
|
|
case "-h", "--help", "help":
|
|
fmt.Print(usage)
|
|
case "version", "--version":
|
|
fmt.Println(version)
|
|
case "migrate":
|
|
os.Exit(cmdMigrate(os.Args[2:]))
|
|
case "initdb":
|
|
os.Exit(cmdInitDB(os.Args[2:]))
|
|
case "render-config":
|
|
os.Exit(cmdRenderConfig(os.Args[2:]))
|
|
case "wg-import":
|
|
os.Exit(cmdWGImport(os.Args[2:]))
|
|
case "cluster-join", "cluster-leave", "promote", "dump-config":
|
|
fmt.Fprintf(os.Stderr, "edgeguard-ctl: %q is a Phase-3 stub — not yet implemented\n", os.Args[1])
|
|
os.Exit(1)
|
|
default:
|
|
fmt.Fprintf(os.Stderr, "edgeguard-ctl: unknown command %q\n", os.Args[1])
|
|
fmt.Fprint(os.Stderr, usage)
|
|
os.Exit(2)
|
|
}
|
|
}
|