mesh-controller/internal/link/stale_report_test.go

package link

import (
	"context"
	"errors"
	"testing"
)

// Supersession as a check rather than a memory (design 25 §3).
//
// Holding a message in memory let the controller drop an older report when a newer one for the
// same node arrived. On the bus being built the message belongs to the server and comes back
// whatever happened meanwhile — so the older report is redelivered *after* the newer was applied,
// and acting on it would undo the newer.
//
// The answer was already in the message: a report carries the digest of the declaration it is
// about, so "is this the past?" is a question the message answers.

// sentAndHeard records reports and knows what was last sent, which is the pair the check needs.
type sentAndHeard struct {
	sent  string
	heard []Report
	err   error
}

func (s *sentAndHeard) Heard(_ context.Context, r Report) (bool, error) {
	if s.err != nil {
		return false, s.err
	}
	s.heard = append(s.heard, r)
	return true, nil
}

func (s *sentAndHeard) Outstanding(context.Context, string) (string, error) { return s.sent, nil }

// A report about a declaration the mesh has moved past is settled and not acted on. Settled rather
// than dropped, because there is nothing wrong with the message — it is simply the past, and
// redelivering it for ever is worse than letting it go.
func TestAReportAboutASupersededDeclarationIsNotActedOn(t *testing.T) {
	store := &sentAndHeard{sent: "d2"}
	s, in := serving()
	s.listener = store
	to := &settled{}
	s.act(context.Background(), in.sends(t, to, KindReport, aReport("anchor", "d1")))

	if len(store.heard) != 0 {
		t.Fatalf("a report about a superseded declaration was acted on: %+v", store.heard)
	}
	if !to.acked {
		t.Fatalf("a superseded report was not settled, so it comes back for ever: %+v", *to)
	}
}

// The report about the declaration that *is* outstanding is acted on, and so is one from a node
// the mesh has no digest for — an older host that says nothing about which declaration it applied
// has nothing to be judged against, and refusing it would silence every node built before reports
// carried the digest.
func TestAReportAboutTheOutstandingDeclarationIsActedOn(t *testing.T) {
	for _, c := range []struct{ what, sent, declared string }{
		{"the one outstanding", "d2", "d2"},
		{"a report that says nothing about which", "d2", ""},
		{"a node nothing was ever sent", "", "d1"},
	} {
		store := &sentAndHeard{sent: c.sent}
		s, in := serving()
		s.listener = store
		to := &settled{}
		s.act(context.Background(), in.sends(t, to, KindReport, aReport("anchor", c.declared)))
		if len(store.heard) != 1 || !to.acked {
			t.Errorf("%s: was not acted on and acknowledged: heard %+v, settled %+v",
				c.what, store.heard, *to)
		}
	}
}

// **Staleness is decided before the store is waited on**, not after: a redelivery that lost its
// race is not worth holding a slot in the window that a current message needs.
func TestASupersededReportIsNotHeldForTheStore(t *testing.T) {
	store := &sentAndHeard{sent: "d2", err: errors.Join(ErrTryAgain, errors.New("starting up"))}
	s, in := serving()
	s.listener = store
	to := &settled{}
	s.act(context.Background(), in.sends(t, to, KindReport, aReport("anchor", "d1")))
	if !to.acked || len(in.held) != 0 {
		t.Fatalf("a superseded report waited for the store: %+v, %d held", *to, len(in.held))
	}
}

// **Half of a report is not about a declaration, and that half is never stale.**
//
// What the machine *is* — the tunnel it took over, the ports its own bundle holds, what an adopted
// node found and is keeping, a node moving its overlay key — reaches the mesh on a report and
// nowhere else. A rekey set aside as stale is a node whose overlay key never moves, and no retry is
// coming, because the node said it once. So a report carrying any of these is acted on whenever it
// arrives, however far the mesh has moved on.
func TestAReportCarryingWhatOnlyTheNodeKnowsIsActedOnHoweverOldItIs(t *testing.T) {
	for _, c := range []struct {
		what   string
		report Report
	}{
		{"a rekey", Report{Node: "anchor", Declared: "d1",
			Rekey: &Rekey{Previous: "k1", OverlayKey: "k2"}}},
		{"the tunnel it carried", Report{Node: "anchor", Declared: "d1",
			Tunnel: &CarriedTunnel{Interface: "wg0", State: "taken"}}},
		{"what an adopted node holds", Report{Node: "anchor", Declared: "d1",
			Held: []Held{{ID: "conf", Module: "web", Kind: "file"}}}},
		{"the firewall it found", Report{Node: "anchor", Declared: "d1", Firewall: "ufw"}},
		{"what is reachable on it", Report{Node: "anchor", Declared: "d1",
			Reachable: []Reach{{Protocol: "tcp", Port: 443}}}},
		{"the ports its own bundle holds", Report{Node: "anchor", Declared: "d1",
			Carried: []int{5432}}},
	} {
		store := &sentAndHeard{sent: "d9"}
		s, in := serving()
		s.listener = store
		to := &settled{}
		s.act(context.Background(), in.sends(t, to, KindReport, c.report))
		if len(store.heard) != 1 {
			t.Errorf("%s was set aside as stale, and the mesh will never hear it again: %+v",
				c.what, *to)
		}
	}
}

// A heartbeat is not held for the store: the next one is a minute away, and one kept for two
// minutes to be written late says nothing the one after it will not say better.
func TestAHeartbeatIsNotHeldForTheStore(t *testing.T) {
	s, in := serving()
	s.listener = heardWith{err: errors.Join(ErrTryAgain, errors.New("starting up"))}
	to := &settled{}
	s.act(context.Background(), in.sends(t, to, KindHeartbeat, Alive{Node: "anchor"}))
	if !to.acked || len(in.held) != 0 {
		t.Fatalf("a heartbeat was held for the store: %+v, %d held", *to, len(in.held))
	}
}