From 1737b7c2f779f9c36994e1fda844d98bf01510bb Mon Sep 17 00:00:00 2001
From: NekoPunch <engineer.jyao@gmail.com>
Date: Sun, 2 Aug 2026 03:18:23 -0700
Subject: [PATCH] fix(atenet): make xDS snapshot versions restart-unique

The in-memory version counter restarted at 1 on every router boot; if
Envoy reconnected still holding an identical version string the
snapshot cache saw a match and skipped the push, stranding Envoy on
pre-restart config. Versions now carry a per-process epoch (unix
seconds plus a random suffix) so no incarnation repeats an earlier
one's strings, even across clock jumps. Fixes #617.
---
 cmd/atenet/internal/router/xds.go      |  5 +++-
 cmd/atenet/internal/router/xds_test.go | 33 ++++++++++++++++++++++++++
 2 files changed, 37 insertions(+), 1 deletion(-)

diff --git a/cmd/atenet/internal/router/xds.go b/cmd/atenet/internal/router/xds.go
index e422376d9..6a6e0f7a8 100644
--- a/cmd/atenet/internal/router/xds.go
+++ b/cmd/atenet/internal/router/xds.go
@@ -16,6 +16,7 @@ package router
 
 import (
 	"context"
+	"crypto/rand"
 	"fmt"
 	"log/slog"
 	"net"
@@ -104,6 +105,7 @@ type XdsServer struct {
 	snapshot     cachev3.SnapshotCache
 	srv          serverv3.Server
 	versionCount int64
+	versionEpoch string
 
 	mu sync.Mutex
 
@@ -145,6 +147,7 @@ func NewXdsServer(xdsPort int) *XdsServer {
 		xdsPort:               xdsPort,
 		snapshot:              cache,
 		srv:                   srv,
+		versionEpoch:          strconv.FormatInt(time.Now().Unix(), 10) + "-" + rand.Text()[:8],
 		extprocPort:           50051, // matches default extproc port
 		extprocAddr:           "127.0.0.1",
 		ingressPort:           8080,
@@ -299,7 +302,7 @@ func (x *XdsServer) UpdateSnapshot() error {
 	defer x.mu.Unlock()
 
 	x.versionCount++
-	ver := strconv.FormatInt(x.versionCount, 10)
+	ver := x.versionEpoch + "-" + strconv.FormatInt(x.versionCount, 10)
 
 	// Clusters
 	clusters := []types.Resource{
diff --git a/cmd/atenet/internal/router/xds_test.go b/cmd/atenet/internal/router/xds_test.go
index 23f9ffdac..918b4cb08 100644
--- a/cmd/atenet/internal/router/xds_test.go
+++ b/cmd/atenet/internal/router/xds_test.go
@@ -649,3 +649,36 @@ func TestXdsServer_SetOtlpCollector_EmptyDisablesTracing(t *testing.T) {
 		t.Errorf("snapshot contains cluster %q, want it omitted when tracing is disabled", OtlpClusterName)
 	}
 }
+
+func TestSnapshotVersionsUniqueAcrossRestarts(t *testing.T) {
+	deployAndGetVersion := func(t *testing.T, x *XdsServer) string {
+		t.Helper()
+		if err := x.UpdateSnapshot(); err != nil {
+			t.Fatalf("UpdateSnapshot: %v", err)
+		}
+		snap, err := x.snapshot.GetSnapshot(NodeID)
+		if err != nil {
+			t.Fatalf("GetSnapshot: %v", err)
+		}
+		return snap.GetVersion(resourcev3.ClusterType)
+	}
+
+	seen := map[string]bool{}
+	first := NewXdsServer(0)
+	for range 3 {
+		v := deployAndGetVersion(t, first)
+		if seen[v] {
+			t.Fatalf("version %q minted twice by the same server", v)
+		}
+		seen[v] = true
+	}
+
+	restarted := NewXdsServer(0)
+	for range 3 {
+		v := deployAndGetVersion(t, restarted)
+		if seen[v] {
+			t.Fatalf("version %q reused after restart; Envoy holding that version would not receive the new config", v)
+		}
+		seen[v] = true
+	}
+}