cilium-cli/connectivity: only check container restarts on v1.17+

f900d93 ("cilium-cli/connectivity: additionally check for container
restarts") introduced a new check to the connectivity suite to validate
that no Cilium-related pods have a container restart count greater than
zero. However, this check is getting triggered way more than expected
on older stable branches. Let's temporarily enable it only for Cilium
v1.17 and later, while the discovered failures get investigated.

Signed-off-by: Marco Iorio <marco.iorio@isovalent.com>

cilium-cli/connectivity/tests/errors.go

@@ -93,13 +93,15 @@ func NoErrorsInLogs(ciliumVersion semver.Version, checkLevels []string, external
 	return &noErrorsInLogs{
 		errorMsgsWithExceptions: errorMsgsWithExceptions,
 		ScenarioBase:            check.NewScenarioBase(),
+		ciliumVersion:           ciliumVersion,
 	}
 }
 
 type noErrorsInLogs struct {
 	check.ScenarioBase
 
 	errorMsgsWithExceptions map[string][]logMatcher
+	ciliumVersion           semver.Version
 }
 
 func (n *noErrorsInLogs) Name() string {
@@ -125,10 +127,13 @@ func (n *noErrorsInLogs) Run(ctx context.Context, t *check.Test) {
 		for container, restarts := range info.containers {
 			id := fmt.Sprintf("%s/%s/%s (%s)", pod.Cluster, pod.Namespace, pod.Name, container)
 			t.NewGenericAction(n, id).Run(func(a *check.Action) {
+				// Do not check for container restarts for Cilium v1.16 and earlier.
+				ignore := n.ciliumVersion.LT(semver.MustParse("1.17.0"))
+
 				// Ignore Cilium operator restarts for the moment, as they can
 				// legitimately happen in case it loses the leader election due
 				// to temporary control plane blips.
-				ignore := container == "cilium-operator"
+				ignore = ignore || container == "cilium-operator"
 
 				// The hubble relay container can currently be restarted by the
 				// startup probe if it fails to connect to Cilium. However, this