| 1 | package console |
| 2 | |
| 3 | import ( |
| 4 | "bytes" |
| 5 | "context" |
| 6 | "crypto/ed25519" |
| 7 | "encoding/json" |
| 8 | "errors" |
| 9 | "strings" |
| 10 | "testing" |
| 11 | "time" |
| 12 | |
| 13 | "github.com/carlosframework/platform/internal/hostactors" |
| 14 | "github.com/carlosframework/platform/internal/hostfloor" |
| 15 | "github.com/carlosframework/platform/internal/release" |
| 16 | "github.com/carlosframework/platform/internal/status" |
| 17 | ) |
| 18 | |
| 19 | type activationFaultStore struct { |
| 20 | release.Store |
| 21 | fail bool |
| 22 | commit bool |
| 23 | hide bool |
| 24 | } |
| 25 | |
| 26 | func (s *activationFaultStore) Put(ctx context.Context, key string, raw []byte, etag string) (string, error) { |
| 27 | if key != release.SystemPointerKey("router", "stable") || !s.fail { |
| 28 | return s.Store.Put(ctx, key, raw, etag) |
| 29 | } |
| 30 | s.fail = false |
| 31 | if s.commit { |
| 32 | if _, err := s.Store.Put(ctx, key, raw, etag); err != nil { |
| 33 | return "", err |
| 34 | } |
| 35 | } |
| 36 | return "", errors.New("publication response lost") |
| 37 | } |
| 38 | func (s *activationFaultStore) Get(ctx context.Context, key string) ([]byte, string, error) { |
| 39 | if s.hide && key == release.SystemPointerKey("router", "stable") && !s.fail { |
| 40 | s.hide = false |
| 41 | return nil, "", errors.New("publication read unavailable") |
| 42 | } |
| 43 | return s.Store.Get(ctx, key) |
| 44 | } |
| 45 | |
| 46 | func activationFixture(t *testing.T) (*Server, hostactors.Directory, release.SystemRelease) { |
| 47 | t.Helper() |
| 48 | ctx := t.Context() |
| 49 | pub, key, err := ed25519.GenerateKey(nil) |
| 50 | if err != nil { |
| 51 | t.Fatal(err) |
| 52 | } |
| 53 | store := release.NewDirStore(t.TempDir()) |
| 54 | s := &Server{cfg: Config{Releases: store, Key: key, Bucket: "bucket"}} |
| 55 | sg := release.RootSigner(key) |
| 56 | old := release.SystemRelease{Version: "v1.0.0", Source: "release:v1.0.0", ReleaseTag: "v1.0.0", MovesProtocol: 1, Artifacts: map[string]release.SystemArtifact{"linux-amd64": {Key: release.SystemArtifactKey("v1.0.0", "linux-amd64"), SHA256: strings.Repeat("a", 64)}}} |
| 57 | next := release.SystemRelease{Version: "v1.1.0", Source: "release:v1.1.0", ReleaseTag: "v1.1.0", MovesProtocol: 1, Artifacts: map[string]release.SystemArtifact{"linux-amd64": {Key: release.SystemArtifactKey("v1.1.0", "linux-amd64"), SHA256: strings.Repeat("b", 64)}}} |
| 58 | for _, rel := range []release.SystemRelease{next, old} { |
| 59 | if _, err := release.PromoteSystem(ctx, store, sg, "router", "stable", rel, release.SystemPromoteOptions{}); err != nil { |
| 60 | t.Fatal(err) |
| 61 | } |
| 62 | raw, _, err := store.Get(ctx, release.SystemPointerKey("router", "stable")) |
| 63 | if err != nil { |
| 64 | t.Fatal(err) |
| 65 | } |
| 66 | if err := release.PublishSystemIdentity(ctx, store, pub, raw); err != nil { |
| 67 | t.Fatal(err) |
| 68 | } |
| 69 | } |
| 70 | if err := hostfloor.Publish(ctx, store, sg, "bucket", hostfloor.Floor{Min: 1, MinRelease: "v1.0.0"}); err != nil { |
| 71 | t.Fatal(err) |
| 72 | } |
| 73 | floor, err := hostfloor.Read(ctx, store, pub, "bucket") |
| 74 | if err != nil { |
| 75 | t.Fatal(err) |
| 76 | } |
| 77 | enrollment := strings.Repeat("e", 32) |
| 78 | roster := hostactors.Roster{V: 1, DeploymentBucket: "bucket", Revision: 1, Publication: "published", FloorSHA256: floor.Digest, Boxes: map[string]hostactors.Box{"box": {BoxID: "box", Namespace: "platform", Label: "box", Region: "ie", Pool: "default", Cloud: hostactors.Cloud{Provider: "aws", Account: "123", Region: "eu-west-1", InstanceID: "i-box"}, Enrollment: enrollment, EvidenceSHA256: strings.Repeat("f", 64), State: "active"}}, Consoles: map[string]hostactors.Console{}, Holds: map[string]hostactors.Reservation{}, Activations: map[string]hostactors.Reservation{}} |
| 79 | raw, err := release.SignRootEnvelope(sg, roster) |
| 80 | if err != nil { |
| 81 | t.Fatal(err) |
| 82 | } |
| 83 | if _, err := store.Put(ctx, hostactors.Key, raw, ""); err != nil { |
| 84 | t.Fatal(err) |
| 85 | } |
| 86 | box := status.Box{GeneratedAt: time.Now(), Metrics: &status.BoxMetrics{Enrollment: enrollment, InstanceID: "i-box", Release: old.Version, ReleaseTag: old.ReleaseTag, Platform: "linux-amd64", BinarySHA256: old.Artifacts["linux-amd64"].SHA256, MovesProtocol: 1, BootID: "boot", FloorSHA256: floor.Digest, FloorGuardReady: true, HostClaimEnabled: true, SystemChannel: "stable"}} |
| 87 | raw, err = json.Marshal(box) |
| 88 | if err != nil { |
| 89 | t.Fatal(err) |
| 90 | } |
| 91 | if _, err := store.Put(ctx, "status/boxes/box.json", raw, ""); err != nil { |
| 92 | t.Fatal(err) |
| 93 | } |
| 94 | return s, hostactors.Directory{Store: store, PublicKey: pub, Bucket: "bucket"}, next |
| 95 | } |
| 96 | |
| 97 | // Legacy promotion needs both floor objects absent; either half alone is |
| 98 | // an unfinished publication (review round 4). |
| 99 | func TestHostUpdateReservationRefusesIncompleteFloorPair(t *testing.T) { |
| 100 | for _, drop := range []string{hostfloor.Key, hostfloor.AuthKey} { |
| 101 | t.Run(drop, func(t *testing.T) { |
| 102 | s, _, target := activationFixture(t) |
| 103 | if err := s.cfg.Releases.Delete(t.Context(), drop); err != nil { |
| 104 | t.Fatal(err) |
| 105 | } |
| 106 | if id, err := s.reserveHostUpdate(t.Context(), &target, "stable", "promote", release.SystemPromoteOptions{}); !errors.Is(err, hostfloor.ErrFloor) || id != "" { |
| 107 | t.Fatalf("incomplete pair without %s = %q, %v; want ErrFloor", drop, id, err) |
| 108 | } |
| 109 | }) |
| 110 | } |
| 111 | } |
| 112 | |
| 113 | func TestHostUpdateReservationRetriesPublication(t *testing.T) { |
| 114 | for _, fault := range []string{"crash", "failed", "committed", "unreadable"} { |
| 115 | t.Run(fault, func(t *testing.T) { |
| 116 | s, d, target := activationFixture(t) |
| 117 | id, err := s.reserveHostUpdate(t.Context(), &target, "stable", "promote", release.SystemPromoteOptions{}) |
| 118 | if err != nil || id == "" { |
| 119 | t.Fatal(id, err) |
| 120 | } |
| 121 | first, err := d.Read(t.Context()) |
| 122 | if err != nil { |
| 123 | t.Fatal(err) |
| 124 | } |
| 125 | if fault != "crash" { |
| 126 | broken := &activationFaultStore{Store: d.Store, fail: true, commit: fault != "failed", hide: fault == "unreadable"} |
| 127 | s.cfg.Releases = broken |
| 128 | _, err = s.promoteHostUpdate(t.Context(), target, "router", "stable", release.SystemPromoteOptions{}) |
| 129 | if (err == nil) != (fault == "committed") { |
| 130 | t.Fatalf("first publication: %v", err) |
| 131 | } |
| 132 | } |
| 133 | // A new console process has no in-memory job or reservation state. |
| 134 | s = &Server{cfg: Config{Releases: d.Store, Key: s.cfg.Key, Bucket: "bucket"}} |
| 135 | p, err := s.promoteHostUpdate(t.Context(), target, "router", "stable", release.SystemPromoteOptions{}) |
| 136 | if err != nil || p.ActivationID != id || p.Current.Version != target.Version { |
| 137 | t.Fatal(p, err) |
| 138 | } |
| 139 | after, err := d.Read(t.Context()) |
| 140 | if err != nil || len(after.Roster.Activations) != 1 { |
| 141 | t.Fatal(after, err) |
| 142 | } |
| 143 | if !bytes.Equal(first.Roster.Activations[id].Publication.After, after.Roster.Activations[id].Publication.After) { |
| 144 | t.Fatal("retry replaced publication identity") |
| 145 | } |
| 146 | }) |
| 147 | } |
| 148 | } |
| 149 | |
| 150 | func TestHostRollbackResumesWithoutPoppingAgain(t *testing.T) { |
| 151 | s, d, _ := activationFixture(t) |
| 152 | broken := &activationFaultStore{Store: d.Store, fail: true} |
| 153 | s.cfg.Releases = broken |
| 154 | if _, err := s.rollbackHostUpdate(t.Context(), "router", "stable", "operator"); err == nil { |
| 155 | t.Fatal("missing injected failure") |
| 156 | } |
| 157 | before, err := d.Read(t.Context()) |
| 158 | if err != nil || len(before.Roster.Activations) != 1 { |
| 159 | t.Fatal(before, err) |
| 160 | } |
| 161 | var id string |
| 162 | for id = range before.Roster.Activations { |
| 163 | } |
| 164 | s.cfg.Releases = d.Store |
| 165 | first, err := s.rollbackHostUpdate(t.Context(), "router", "stable", "operator") |
| 166 | if err != nil || first.ActivationID != id || first.Current.Version != "v1.1.0" { |
| 167 | t.Fatal(first, err) |
| 168 | } |
| 169 | second, err := s.rollbackHostUpdate(t.Context(), "router", "stable", "operator") |
| 170 | if err != nil || second.ActivationID != id || len(second.History) != len(first.History) { |
| 171 | t.Fatal(second, err) |
| 172 | } |
| 173 | } |
| 174 | |
| 175 | func TestHostUpdateCancellationBlocksDelayedPublication(t *testing.T) { |
| 176 | s, d, target := activationFixture(t) |
| 177 | id, err := s.reserveHostUpdate(t.Context(), &target, "stable", "promote", release.SystemPromoteOptions{}) |
| 178 | if err != nil { |
| 179 | t.Fatal(err) |
| 180 | } |
| 181 | before, err := d.Read(t.Context()) |
| 182 | if err != nil { |
| 183 | t.Fatal(err) |
| 184 | } |
| 185 | stale := *before.Roster.Activations[id].Publication |
| 186 | if err := d.CancelActivation(t.Context(), release.RootSigner(s.cfg.Key), id); err != nil { |
| 187 | t.Fatal(err) |
| 188 | } |
| 189 | if _, err := release.PublishSystemPublication(t.Context(), d.Store, d.PublicKey, stale); !errors.Is(err, release.ErrCASConflict) { |
| 190 | t.Fatalf("delayed publication: %v", err) |
| 191 | } |
| 192 | after, err := d.Read(t.Context()) |
| 193 | if err != nil || len(after.Roster.Activations) != 0 { |
| 194 | t.Fatal(after, err) |
| 195 | } |
| 196 | p, err := s.promoteHostUpdate(t.Context(), target, "router", "stable", release.SystemPromoteOptions{}) |
| 197 | if err != nil || p.ActivationID == id { |
| 198 | t.Fatal(p, err) |
| 199 | } |
| 200 | if err := d.CancelActivation(t.Context(), release.RootSigner(s.cfg.Key), p.ActivationID); !errors.Is(err, release.ErrCASConflict) { |
| 201 | t.Fatalf("cancelled published activation: %v", err) |
| 202 | } |
| 203 | after, err = d.Read(t.Context()) |
| 204 | if err != nil || len(after.Roster.Activations) != 1 { |
| 205 | t.Fatal(after, err) |
| 206 | } |
| 207 | } |
| 208 | |