| 1 | package host |
| 2 | |
| 3 | import ( |
| 4 | "bytes" |
| 5 | "context" |
| 6 | "crypto/ecdh" |
| 7 | "encoding/base64" |
| 8 | "encoding/json" |
| 9 | "log" |
| 10 | "os" |
| 11 | "path/filepath" |
| 12 | "strings" |
| 13 | "sync" |
| 14 | "testing" |
| 15 | "time" |
| 16 | |
| 17 | "github.com/carlosframework/platform/internal/appconfig" |
| 18 | "github.com/carlosframework/platform/internal/edgesfile" |
| 19 | "github.com/carlosframework/platform/internal/moves" |
| 20 | "github.com/carlosframework/platform/internal/registry" |
| 21 | "github.com/carlosframework/platform/internal/release" |
| 22 | "github.com/carlosframework/platform/internal/status" |
| 23 | ) |
| 24 | |
| 25 | func edgesTestRegistry(t *testing.T) *registry.Registry { |
| 26 | t.Helper() |
| 27 | reg, err := registry.Open(filepath.Join(t.TempDir(), "registry.db")) |
| 28 | if err != nil { |
| 29 | t.Fatal(err) |
| 30 | } |
| 31 | t.Cleanup(func() { reg.Close() }) |
| 32 | return reg |
| 33 | } |
| 34 | |
| 35 | // seedRoute upserts a minimal ready route under the given host — any served |
| 36 | // kind does for these tests, since edgesFilePass folds every kind in |
| 37 | // (the serving-class split is the reader's business, Task 3). |
| 38 | func seedRoute(t *testing.T, reg *registry.Registry, host string) { |
| 39 | t.Helper() |
| 40 | if err := reg.Upsert(registry.Route{Host: host, Kind: "service", Addr: "10.0.0.1:3000", Status: "ready"}); err != nil { |
| 41 | t.Fatal(err) |
| 42 | } |
| 43 | } |
| 44 | |
| 45 | // putCountingStore wraps a release.Store and counts Put calls — the same |
| 46 | // wrap-and-count idiom agent_test.go's countingStore uses for Get, applied |
| 47 | // to Put so the change-gate test can pin "second identical pass writes |
| 48 | // nothing" without inspecting store internals. |
| 49 | type putCountingStore struct { |
| 50 | release.Store |
| 51 | puts int |
| 52 | } |
| 53 | |
| 54 | func (p *putCountingStore) Put(ctx context.Context, key string, body []byte, ifMatch string) (string, error) { |
| 55 | p.puts++ |
| 56 | return p.Store.Put(ctx, key, body, ifMatch) |
| 57 | } |
| 58 | |
| 59 | func TestEdgesFilePublishesHomedHosts(t *testing.T) { |
| 60 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 61 | seedRoute(t, reg, "app.example.com") |
| 62 | var lastBody []byte |
| 63 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "1.2.3.4", &lastBody); err != nil { |
| 64 | t.Fatal(err) |
| 65 | } |
| 66 | got, _, err := s.Get(context.Background(), "control/edges/annie.json") |
| 67 | if err != nil { |
| 68 | t.Fatal(err) |
| 69 | } |
| 70 | var f edgesFile |
| 71 | if err := json.Unmarshal(got, &f); err != nil { |
| 72 | t.Fatal(err) |
| 73 | } |
| 74 | if f.V != 1 || f.Box != "annie" || f.EIP != "1.2.3.4" { |
| 75 | t.Fatalf("file = %+v", f) |
| 76 | } |
| 77 | if r, ok := f.Hosts["app.example.com"]; !ok || r.Role != "home" { |
| 78 | t.Fatalf("hosts = %+v", f.Hosts) |
| 79 | } |
| 80 | if f.Generation == "" { |
| 81 | t.Fatal("generation must be stamped") |
| 82 | } |
| 83 | if got, want := edgesFileKey("annie"), "control/edges/annie.json"; got != want { |
| 84 | t.Fatalf("edgesFileKey = %q, want %q", got, want) |
| 85 | } |
| 86 | } |
| 87 | |
| 88 | // A box with no discoverable public IP still publishes the file, with |
| 89 | // "eip":"" — the DR runbook's hosts list must not depend on IMDS working. |
| 90 | func TestEdgesFilePublishesWithEmptyEIP(t *testing.T) { |
| 91 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 92 | seedRoute(t, reg, "app.example.com") |
| 93 | var lastBody []byte |
| 94 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "", &lastBody); err != nil { |
| 95 | t.Fatal(err) |
| 96 | } |
| 97 | got, _, err := s.Get(context.Background(), "control/edges/annie.json") |
| 98 | if err != nil { |
| 99 | t.Fatal(err) |
| 100 | } |
| 101 | var f edgesFile |
| 102 | if err := json.Unmarshal(got, &f); err != nil { |
| 103 | t.Fatal(err) |
| 104 | } |
| 105 | if f.EIP != "" { |
| 106 | t.Fatalf("eip = %q, want empty", f.EIP) |
| 107 | } |
| 108 | if _, ok := f.Hosts["app.example.com"]; !ok { |
| 109 | t.Fatalf("hosts = %+v, want app.example.com present despite empty eip", f.Hosts) |
| 110 | } |
| 111 | } |
| 112 | |
| 113 | func TestEdgesFileSkipsWriteWhenUnchanged(t *testing.T) { |
| 114 | real := release.NewDirStore(t.TempDir()) |
| 115 | s := &putCountingStore{Store: real} |
| 116 | reg := edgesTestRegistry(t) |
| 117 | seedRoute(t, reg, "app.example.com") |
| 118 | |
| 119 | var lastBody []byte |
| 120 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "1.2.3.4", &lastBody); err != nil { |
| 121 | t.Fatal(err) |
| 122 | } |
| 123 | if s.puts != 1 { |
| 124 | t.Fatalf("first pass puts = %d, want 1", s.puts) |
| 125 | } |
| 126 | |
| 127 | // Second pass, identical registry state: a fresh generation timestamp |
| 128 | // would differ from the first pass's, but the change-gate excludes |
| 129 | // Generation, so this must not write. |
| 130 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "1.2.3.4", &lastBody); err != nil { |
| 131 | t.Fatal(err) |
| 132 | } |
| 133 | if s.puts != 1 { |
| 134 | t.Fatalf("unchanged second pass puts = %d, want still 1 (no write)", s.puts) |
| 135 | } |
| 136 | |
| 137 | // A registry change makes the next pass write again. |
| 138 | seedRoute(t, reg, "second.example.com") |
| 139 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "1.2.3.4", &lastBody); err != nil { |
| 140 | t.Fatal(err) |
| 141 | } |
| 142 | if s.puts != 2 { |
| 143 | t.Fatalf("changed third pass puts = %d, want 2", s.puts) |
| 144 | } |
| 145 | } |
| 146 | |
| 147 | // FINDING 1 (review, 2026-08-22): readPublicIP is a single 2s best-effort |
| 148 | // IMDS attempt, re-run every tick — a transient throttle or blip returns "" |
| 149 | // exactly like a box that never had a public IP at all. Without a hold, a |
| 150 | // single hiccuped tick would blank a known-good EIP in the published file, |
| 151 | // costing every remote edge its dial address for this box's home-boxed |
| 152 | // hosts until the next successful tick (up to 5min + jitter). A pass |
| 153 | // reporting "" after a pass that reported "1.2.3.4" must keep "1.2.3.4". |
| 154 | func TestEdgesFileHoldsLastKnownGoodEIPOnTransientBlank(t *testing.T) { |
| 155 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 156 | seedRoute(t, reg, "app.example.com") |
| 157 | |
| 158 | var last []byte |
| 159 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "1.2.3.4", &last); err != nil { |
| 160 | t.Fatal(err) |
| 161 | } |
| 162 | |
| 163 | // Simulate a transient IMDS blip: this tick's probe returns "". |
| 164 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "", &last); err != nil { |
| 165 | t.Fatal(err) |
| 166 | } |
| 167 | |
| 168 | got, _, err := s.Get(context.Background(), "control/edges/annie.json") |
| 169 | if err != nil { |
| 170 | t.Fatal(err) |
| 171 | } |
| 172 | var f edgesFile |
| 173 | if err := json.Unmarshal(got, &f); err != nil { |
| 174 | t.Fatal(err) |
| 175 | } |
| 176 | if f.EIP != "1.2.3.4" { |
| 177 | t.Fatalf("eip after a blank probe = %q, want held value %q", f.EIP, "1.2.3.4") |
| 178 | } |
| 179 | } |
| 180 | |
| 181 | // FINDING 2 (review, 2026-08-22, ruling): edge.Reload (internal/edge/edge.go) |
| 182 | // skips any route whose Status != "ready" — that gate is the deployment's |
| 183 | // own definition of "serves", so a pending/creating/failed row must not |
| 184 | // claim role:"home" here either. Hibernating routes stay Status "ready", |
| 185 | // so this filter excludes nothing a remote backhaul could actually reach. |
| 186 | func TestEdgesFileExcludesNonReadyRows(t *testing.T) { |
| 187 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 188 | seedRoute(t, reg, "ready.example.com") |
| 189 | if err := reg.Upsert(registry.Route{Host: "creating.example.com", Kind: "service", Addr: "10.0.0.2:3000", Status: "creating"}); err != nil { |
| 190 | t.Fatal(err) |
| 191 | } |
| 192 | |
| 193 | var last []byte |
| 194 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "1.2.3.4", &last); err != nil { |
| 195 | t.Fatal(err) |
| 196 | } |
| 197 | got, _, err := s.Get(context.Background(), "control/edges/annie.json") |
| 198 | if err != nil { |
| 199 | t.Fatal(err) |
| 200 | } |
| 201 | var f edgesFile |
| 202 | if err := json.Unmarshal(got, &f); err != nil { |
| 203 | t.Fatal(err) |
| 204 | } |
| 205 | if _, ok := f.Hosts["ready.example.com"]; !ok { |
| 206 | t.Fatalf("hosts = %+v, want ready.example.com present", f.Hosts) |
| 207 | } |
| 208 | if _, ok := f.Hosts["creating.example.com"]; ok { |
| 209 | t.Fatalf("hosts = %+v, want creating.example.com ABSENT (not ready, nothing serves it yet)", f.Hosts) |
| 210 | } |
| 211 | } |
| 212 | |
| 213 | // A sidecar row is READY — that is what makes it runnable — so the readiness |
| 214 | // filter above lets it straight through, and this file is published to every |
| 215 | // box in the deployment. That matters more than it looks: a REMOTE edge has |
| 216 | // no sidecar row and never sees a Kind, so edge.Reload's refusal cannot |
| 217 | // protect it. It misses on its own upstream lookup, reads this file through |
| 218 | // the home map, and backhauls — dialing the home box with the sidecar's own |
| 219 | // name as SNI. The same entry also satisfies deploymentHost, which is what |
| 220 | // admits a name to the cert machinery. |
| 221 | // |
| 222 | // So the guarantee has to be made by the WRITER, exactly as |
| 223 | // publishResidentFleetBindings makes it: a sidecar is never advertised as |
| 224 | // served, by this box or any other. |
| 225 | func TestEdgesFileNeverPublishesASidecar(t *testing.T) { |
| 226 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 227 | seedRoute(t, reg, "ready.example.com") |
| 228 | // Minted the way the sidecar reconcile pass mints one: ready, and |
| 229 | // carrying neither socket nor addr, because nothing dials it. |
| 230 | if err := reg.Upsert(registry.Route{ |
| 231 | Host: "bot.eleven.bdm.sidecar.internal", Kind: "sidecar", |
| 232 | App: "eleven", Account: "bdm", Status: "ready", Provisioned: true, |
| 233 | }); err != nil { |
| 234 | t.Fatal(err) |
| 235 | } |
| 236 | |
| 237 | var last []byte |
| 238 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie"}, "1.2.3.4", &last); err != nil { |
| 239 | t.Fatal(err) |
| 240 | } |
| 241 | got, _, err := s.Get(context.Background(), "control/edges/annie.json") |
| 242 | if err != nil { |
| 243 | t.Fatal(err) |
| 244 | } |
| 245 | var f edgesFile |
| 246 | if err := json.Unmarshal(got, &f); err != nil { |
| 247 | t.Fatal(err) |
| 248 | } |
| 249 | if _, ok := f.Hosts["ready.example.com"]; !ok { |
| 250 | t.Fatalf("hosts = %+v, want the ordinary route still published", f.Hosts) |
| 251 | } |
| 252 | if _, ok := f.Hosts["bot.eleven.bdm.sidecar.internal"]; ok { |
| 253 | t.Fatalf("hosts = %+v, want the sidecar ABSENT — publishing it makes every remote edge a backhaul door onto a member nothing may dial", f.Hosts) |
| 254 | } |
| 255 | } |
| 256 | |
| 257 | // FINDING 3 (review, 2026-08-22): a catch-all claim's registry row is a real |
| 258 | // "*.parent" host (Task 3's reconcile pass derives it the same way |
| 259 | // edge.go's upstream map does). This writer performs no wildcard expansion |
| 260 | // — the row's Host lands in Hosts verbatim, exactly like every other row — |
| 261 | // so an exact-match reader must apply edge.go's lookupUpstream one-label |
| 262 | // rule itself; that contract is documented at the call site and in the |
| 263 | // spec, not enforced here. |
| 264 | func TestEdgesFileWildcardRowPublishesVerbatim(t *testing.T) { |
| 265 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 266 | if err := reg.Upsert(registry.Route{Host: "*.example.com", Kind: "service", Addr: "10.0.0.3:3000", Status: "ready"}); err != nil { |
| 267 | t.Fatal(err) |
| 268 | } |
| 269 | |
| 270 | var last []byte |
| 271 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "1.2.3.4", &last); err != nil { |
| 272 | t.Fatal(err) |
| 273 | } |
| 274 | got, _, err := s.Get(context.Background(), "control/edges/annie.json") |
| 275 | if err != nil { |
| 276 | t.Fatal(err) |
| 277 | } |
| 278 | var f edgesFile |
| 279 | if err := json.Unmarshal(got, &f); err != nil { |
| 280 | t.Fatal(err) |
| 281 | } |
| 282 | if r, ok := f.Hosts["*.example.com"]; !ok || r.Role != "home" { |
| 283 | t.Fatalf("hosts = %+v, want \"*.example.com\" present verbatim with role home", f.Hosts) |
| 284 | } |
| 285 | } |
| 286 | |
| 287 | // FINDING 4 (review, 2026-08-22): every other test in this file round-trips |
| 288 | // through the edgesFile struct on both the write and the assert side, so a |
| 289 | // json-tag rename (e.g. "eip" -> "publicIP") would stay green everywhere |
| 290 | // while breaking Task 3's parser, which reads the wire bytes independently |
| 291 | // of this package's Go types. This test pins the RAW body against a |
| 292 | // hand-written literal matching the brief's exact schema string — every |
| 293 | // byte is fixed except the generation timestamp, which this fixed |
| 294 | // signature has no clock seam to inject (edgesFilePass takes no `now` |
| 295 | // parameter, per the brief's pinned interface); that one field is read |
| 296 | // back and re-inserted into the literal rather than relaxed by a regex. |
| 297 | func TestEdgesFileRawWireBodyMatchesSchema(t *testing.T) { |
| 298 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 299 | seedRoute(t, reg, "app.example.com") |
| 300 | |
| 301 | var last []byte |
| 302 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "quiet-harbour", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "203.0.113.10", &last); err != nil { |
| 303 | t.Fatal(err) |
| 304 | } |
| 305 | got, _, err := s.Get(context.Background(), "control/edges/quiet-harbour.json") |
| 306 | if err != nil { |
| 307 | t.Fatal(err) |
| 308 | } |
| 309 | |
| 310 | var probe struct { |
| 311 | Generation string `json:"generation"` |
| 312 | } |
| 313 | if err := json.Unmarshal(got, &probe); err != nil { |
| 314 | t.Fatal(err) |
| 315 | } |
| 316 | if _, err := time.Parse(time.RFC3339, probe.Generation); err != nil { |
| 317 | t.Fatalf("generation %q is not RFC3339: %v", probe.Generation, err) |
| 318 | } |
| 319 | |
| 320 | want := `{"v":1,"box":"quiet-harbour","eip":"203.0.113.10","hibernation_configured":false,"serves_shared_pool":false,"accepts_inbound":false,"moves_protocol":0,"seeds_protocol":0,"generation":"` + |
| 321 | probe.Generation + `","hosts":{"app.example.com":{"role":"home","platform":false}}}` |
| 322 | if string(got) != want { |
| 323 | t.Fatalf("raw wire body =\n%s\nwant\n%s", got, want) |
| 324 | } |
| 325 | } |
| 326 | |
| 327 | // FINDING 5 (task 1, 2026-08-22): region is additive to the pinned schema — |
| 328 | // this extends FINDING 4's raw-body pin rather than replacing it. Two |
| 329 | // bodies, byte for byte: an unresolved region (edgesFileRegion's "" case — |
| 330 | // off EC2, no CARLOS_AWS_REGION override) must produce the IDENTICAL bytes |
| 331 | // FINDING 4 already pins (no "region" key at all, omitempty), and a |
| 332 | // resolved region must insert exactly `"region":"eu-west-1"` between "eip" |
| 333 | // and "generation" — the struct field order — with nothing else disturbed. |
| 334 | // This test drives edgesFilePass directly with the resolved value; the |
| 335 | // resolution contract itself (override wins, IMDS otherwise, trimmed) is |
| 336 | // pinned separately in TestEdgesFileRegionResolution. |
| 337 | func TestEdgesFileRawWireBodyRegionField(t *testing.T) { |
| 338 | t.Parallel() |
| 339 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 340 | seedRoute(t, reg, "app.example.com") |
| 341 | |
| 342 | genOf := func(t *testing.T, body []byte) string { |
| 343 | t.Helper() |
| 344 | var probe struct { |
| 345 | Generation string `json:"generation"` |
| 346 | } |
| 347 | if err := json.Unmarshal(body, &probe); err != nil { |
| 348 | t.Fatal(err) |
| 349 | } |
| 350 | return probe.Generation |
| 351 | } |
| 352 | |
| 353 | t.Run("unset region is omitted entirely", func(t *testing.T) { |
| 354 | var last []byte |
| 355 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "quiet-harbour", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "203.0.113.10", &last); err != nil { |
| 356 | t.Fatal(err) |
| 357 | } |
| 358 | got, _, err := s.Get(context.Background(), "control/edges/quiet-harbour.json") |
| 359 | if err != nil { |
| 360 | t.Fatal(err) |
| 361 | } |
| 362 | gen := genOf(t, got) |
| 363 | want := `{"v":1,"box":"quiet-harbour","eip":"203.0.113.10","hibernation_configured":false,"serves_shared_pool":false,"accepts_inbound":false,"moves_protocol":0,"seeds_protocol":0,"generation":"` + |
| 364 | gen + `","hosts":{"app.example.com":{"role":"home","platform":false}}}` |
| 365 | if string(got) != want { |
| 366 | t.Fatalf("raw wire body (region unset) =\n%s\nwant\n%s", got, want) |
| 367 | } |
| 368 | }) |
| 369 | |
| 370 | t.Run("set region is inserted between eip and generation", func(t *testing.T) { |
| 371 | s2, reg2 := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 372 | seedRoute(t, reg2, "app.example.com") |
| 373 | var last []byte |
| 374 | if err := edgesFilePass(context.Background(), s2, reg2, edgesFile{Box: "quiet-harbour", Region: "eu-west-1", Placement: "", DataBucket: "", DataBucketRegion: ""}, "203.0.113.10", &last); err != nil { |
| 375 | t.Fatal(err) |
| 376 | } |
| 377 | got, _, err := s2.Get(context.Background(), "control/edges/quiet-harbour.json") |
| 378 | if err != nil { |
| 379 | t.Fatal(err) |
| 380 | } |
| 381 | gen := genOf(t, got) |
| 382 | want := `{"v":1,"box":"quiet-harbour","eip":"203.0.113.10","region":"eu-west-1","hibernation_configured":false,"serves_shared_pool":false,"accepts_inbound":false,"moves_protocol":0,"seeds_protocol":0,"generation":"` + |
| 383 | gen + `","hosts":{"app.example.com":{"role":"home","platform":false}}}` |
| 384 | if string(got) != want { |
| 385 | t.Fatalf("raw wire body (region set) =\n%s\nwant\n%s", got, want) |
| 386 | } |
| 387 | }) |
| 388 | } |
| 389 | |
| 390 | // Plan re-ruling (2026-08-23, review catch): CARLOS_REGION is the CARLOS |
| 391 | // placement code ("ie"), NOT an AWS region — Task 5's |
| 392 | // ChangeResourceRecordSets would reject it batch-wide, so the field's |
| 393 | // value must come from CARLOS_AWS_REGION / IMDS instead, never |
| 394 | // CARLOS_REGION. edgesFileRegion is the pure half of that contract |
| 395 | // (getenv/probe-injectable, domainsTickConfig's own shape) — this pins |
| 396 | // override-wins, unset-and-no-probe-answer → "", and trimming on both |
| 397 | // inputs, all without a live IMDS server. |
| 398 | func TestEdgesFileRegionResolution(t *testing.T) { |
| 399 | env := func(m map[string]string) func(string) string { |
| 400 | return func(k string) string { return m[k] } |
| 401 | } |
| 402 | |
| 403 | cases := []struct { |
| 404 | name string |
| 405 | env map[string]string |
| 406 | probed string |
| 407 | want string |
| 408 | }{ |
| 409 | { |
| 410 | name: "override set wins over the probed value", |
| 411 | env: map[string]string{"CARLOS_AWS_REGION": "eu-west-1"}, |
| 412 | probed: "us-east-1", |
| 413 | want: "eu-west-1", |
| 414 | }, |
| 415 | { |
| 416 | name: "override set, untrimmed, wins and is trimmed", |
| 417 | env: map[string]string{"CARLOS_AWS_REGION": " eu-west-1 "}, |
| 418 | probed: "us-east-1", |
| 419 | want: "eu-west-1", |
| 420 | }, |
| 421 | { |
| 422 | name: "override unset falls back to the probed value", |
| 423 | env: map[string]string{}, |
| 424 | probed: "eu-west-1", |
| 425 | want: "eu-west-1", |
| 426 | }, |
| 427 | { |
| 428 | name: "override unset, probed value untrimmed, is trimmed", |
| 429 | env: map[string]string{}, |
| 430 | probed: " eu-west-1 ", |
| 431 | want: "eu-west-1", |
| 432 | }, |
| 433 | { |
| 434 | name: "CARLOS_REGION (the placement code) is never consulted", |
| 435 | env: map[string]string{"CARLOS_REGION": "ie"}, |
| 436 | probed: "", |
| 437 | want: "", |
| 438 | }, |
| 439 | { |
| 440 | name: "unset override and no probe answer (off EC2) is empty, not an error", |
| 441 | env: map[string]string{}, |
| 442 | probed: "", |
| 443 | want: "", |
| 444 | }, |
| 445 | { |
| 446 | name: "blank override (set but empty) does not shadow the probed value", |
| 447 | env: map[string]string{"CARLOS_AWS_REGION": ""}, |
| 448 | probed: "eu-west-1", |
| 449 | want: "eu-west-1", |
| 450 | }, |
| 451 | } |
| 452 | for _, c := range cases { |
| 453 | t.Run(c.name, func(t *testing.T) { |
| 454 | if got := edgesFileRegion(env(c.env), c.probed); got != c.want { |
| 455 | t.Fatalf("edgesFileRegion() = %q, want %q", got, c.want) |
| 456 | } |
| 457 | }) |
| 458 | } |
| 459 | } |
| 460 | |
| 461 | // FINDING 6 (task 1, 2026-08-22, the plan-2 final review's Important 1): a |
| 462 | // freshly ARMED loop must seed its change-gate from whatever this box last |
| 463 | // published, or two bugs follow — see seedEdgesFileGate's doc. This test |
| 464 | // pins the seed-then-pass contract directly: a process that primes its |
| 465 | // *last from seedEdgesFileGate, then runs a pass against UNCHANGED registry |
| 466 | // state, must write ZERO times. Before this task, edgesFileLoop always |
| 467 | // started with a bare `var last []byte`, so the very first tick after any |
| 468 | // restart always wrote — this is the regression that closes. |
| 469 | func TestEdgesFileSeedGateSkipsWriteOnUnchangedBucketState(t *testing.T) { |
| 470 | real := release.NewDirStore(t.TempDir()) |
| 471 | reg := edgesTestRegistry(t) |
| 472 | seedRoute(t, reg, "app.example.com") |
| 473 | |
| 474 | // A prior process (or the prior pass of this one) already published. |
| 475 | var priming []byte |
| 476 | if err := edgesFilePass(context.Background(), real, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "1.2.3.4", &priming); err != nil { |
| 477 | t.Fatal(err) |
| 478 | } |
| 479 | |
| 480 | // A fresh process arms: seed the gate from the bucket, as edgesFileLoop |
| 481 | // now does at loop arm. |
| 482 | last := seedEdgesFileGate(context.Background(), real, "annie") |
| 483 | if last == nil { |
| 484 | t.Fatal("seedEdgesFileGate returned nil, want the prior file's gate body") |
| 485 | } |
| 486 | |
| 487 | counting := &putCountingStore{Store: real} |
| 488 | if err := edgesFilePass(context.Background(), counting, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "1.2.3.4", &last); err != nil { |
| 489 | t.Fatal(err) |
| 490 | } |
| 491 | if counting.puts != 0 { |
| 492 | t.Fatalf("puts = %d, want 0 — a seeded gate against unchanged state must not write", counting.puts) |
| 493 | } |
| 494 | } |
| 495 | |
| 496 | // The restart-blip case named in the final review: a process restarts, |
| 497 | // arms, and its very FIRST tick hits both a transient IMDS blank AND a real |
| 498 | // registry change (so the pass must write). The write must still carry the |
| 499 | // bucket's held EIP, not the blank probe — seedEdgesFileGate's dual |
| 500 | // contract (gate AND EIP hold) has to survive an ACTUAL write, not just a |
| 501 | // no-op pass, or the hold half of the fix is unverified. |
| 502 | func TestEdgesFileSeedHoldsEIPThroughRestartBlip(t *testing.T) { |
| 503 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 504 | seedRoute(t, reg, "app.example.com") |
| 505 | |
| 506 | var priming []byte |
| 507 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "1.2.3.4", &priming); err != nil { |
| 508 | t.Fatal(err) |
| 509 | } |
| 510 | |
| 511 | last := seedEdgesFileGate(context.Background(), s, "annie") |
| 512 | if last == nil { |
| 513 | t.Fatal("seedEdgesFileGate returned nil, want the prior file's gate body") |
| 514 | } |
| 515 | |
| 516 | // This tick both hits a transient IMDS blip AND sees a real registry |
| 517 | // change, so the pass must write. |
| 518 | seedRoute(t, reg, "second.example.com") |
| 519 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "", &last); err != nil { |
| 520 | t.Fatal(err) |
| 521 | } |
| 522 | |
| 523 | got, _, err := s.Get(context.Background(), "control/edges/annie.json") |
| 524 | if err != nil { |
| 525 | t.Fatal(err) |
| 526 | } |
| 527 | var f edgesFile |
| 528 | if err := json.Unmarshal(got, &f); err != nil { |
| 529 | t.Fatal(err) |
| 530 | } |
| 531 | if f.EIP != "1.2.3.4" { |
| 532 | t.Fatalf("eip after restart + blank probe = %q, want held %q", f.EIP, "1.2.3.4") |
| 533 | } |
| 534 | if _, ok := f.Hosts["second.example.com"]; !ok { |
| 535 | t.Fatalf("hosts = %+v, want second.example.com present — the write must carry the real registry change too", f.Hosts) |
| 536 | } |
| 537 | } |
| 538 | |
| 539 | // Without the seed, a fresh *last is nil and edgesFilePass's own hold guard |
| 540 | // (`*last != nil`) cannot fire — this pins that this really is the bug the |
| 541 | // two tests above close, not a tautology of the hold logic that would pass |
| 542 | // regardless of seeding. |
| 543 | func TestEdgesFileUnseededLastPublishesBlankEIPOnFirstTickBlip(t *testing.T) { |
| 544 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 545 | seedRoute(t, reg, "app.example.com") |
| 546 | |
| 547 | var priming []byte |
| 548 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "1.2.3.4", &priming); err != nil { |
| 549 | t.Fatal(err) |
| 550 | } |
| 551 | |
| 552 | // The pre-fix shape: a fresh process's *last starts nil (no seed read). |
| 553 | var unseeded []byte |
| 554 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "", &unseeded); err != nil { |
| 555 | t.Fatal(err) |
| 556 | } |
| 557 | |
| 558 | got, _, err := s.Get(context.Background(), "control/edges/annie.json") |
| 559 | if err != nil { |
| 560 | t.Fatal(err) |
| 561 | } |
| 562 | var f edgesFile |
| 563 | if err := json.Unmarshal(got, &f); err != nil { |
| 564 | t.Fatal(err) |
| 565 | } |
| 566 | if f.EIP != "" { |
| 567 | t.Fatalf("eip = %q, want blank — demonstrating why the seed in the tests above is load-bearing, not incidental", f.EIP) |
| 568 | } |
| 569 | } |
| 570 | |
| 571 | // Self-review: does the seed interact correctly with the change gate when |
| 572 | // the bucket file PREDATES the region field? A box upgraded to this build |
| 573 | // finds a bucket file an older build left with no "region" key at all. |
| 574 | // Seeding *last from it must not treat this build's first pass — the one |
| 575 | // that finally sets a real region — as unchanged: Region participates in |
| 576 | // the change gate exactly like every other field (edgesFilePass's doc), so |
| 577 | // a seeded pre-region gate and a fresh body WITH region differ in bytes and |
| 578 | // must write exactly once, not zero and not on every tick after. |
| 579 | func TestEdgesFileSeedPreRegionBucketWritesOnceWhenRegionArrives(t *testing.T) { |
| 580 | real := release.NewDirStore(t.TempDir()) |
| 581 | reg := edgesTestRegistry(t) |
| 582 | seedRoute(t, reg, "app.example.com") |
| 583 | |
| 584 | // Hand-written exactly as a pre-region build would have left it in the |
| 585 | // bucket: no "region" key present. |
| 586 | preRegion := []byte(`{"v":1,"box":"annie","eip":"1.2.3.4","generation":"2026-08-01T00:00:00Z","hosts":{"app.example.com":{"role":"home"}}}`) |
| 587 | if err := putReplacing(context.Background(), real, edgesFileKey("annie"), preRegion); err != nil { |
| 588 | t.Fatal(err) |
| 589 | } |
| 590 | |
| 591 | last := seedEdgesFileGate(context.Background(), real, "annie") |
| 592 | if last == nil { |
| 593 | t.Fatal("seedEdgesFileGate returned nil") |
| 594 | } |
| 595 | |
| 596 | // This build now has CARLOS_REGION set: the first pass must write, once. |
| 597 | counting := &putCountingStore{Store: real} |
| 598 | if err := edgesFilePass(context.Background(), counting, reg, edgesFile{Box: "annie", Region: "eu-west-1", Placement: "", DataBucket: "", DataBucketRegion: ""}, "1.2.3.4", &last); err != nil { |
| 599 | t.Fatal(err) |
| 600 | } |
| 601 | if counting.puts != 1 { |
| 602 | t.Fatalf("puts = %d, want 1 — a region arriving on a seeded pre-region body is a real change", counting.puts) |
| 603 | } |
| 604 | got, _, err := real.Get(context.Background(), edgesFileKey("annie")) |
| 605 | if err != nil { |
| 606 | t.Fatal(err) |
| 607 | } |
| 608 | var f edgesFile |
| 609 | if err := json.Unmarshal(got, &f); err != nil { |
| 610 | t.Fatal(err) |
| 611 | } |
| 612 | if f.Region != "eu-west-1" { |
| 613 | t.Fatalf("region = %q, want eu-west-1", f.Region) |
| 614 | } |
| 615 | |
| 616 | // A second, identical-region pass must not write again — the gate now |
| 617 | // includes region on both sides. |
| 618 | if err := edgesFilePass(context.Background(), counting, reg, edgesFile{Box: "annie", Region: "eu-west-1", Placement: "", DataBucket: "", DataBucketRegion: ""}, "1.2.3.4", &last); err != nil { |
| 619 | t.Fatal(err) |
| 620 | } |
| 621 | if counting.puts != 1 { |
| 622 | t.Fatalf("puts after unchanged second pass = %d, want still 1", counting.puts) |
| 623 | } |
| 624 | } |
| 625 | |
| 626 | // The class mark (#300): platform-minted rows carry platform:true; a stray |
| 627 | // operator row does not, so no remote edge derives a wildcard parent from it. |
| 628 | func TestEdgesFileMarksPlatformRows(t *testing.T) { |
| 629 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 630 | for _, r := range []registry.Route{ |
| 631 | {Host: "app.acct.apps.test", Kind: "service", Addr: "10.0.0.1:3000", Status: "ready", Platform: true}, |
| 632 | {Host: "prov.acct.apps.test", Kind: "service", Addr: "10.0.0.1:3001", Status: "ready", Provisioned: true}, |
| 633 | {Host: "stray.acct.apps.test", Kind: "service", Addr: "10.0.0.1:3002", Status: "ready"}, |
| 634 | {Host: "pending.acct.apps.test", Kind: "service", Addr: "10.0.0.1:3003", Status: "pending", Platform: true}, |
| 635 | } { |
| 636 | if err := reg.Upsert(r); err != nil { |
| 637 | t.Fatal(err) |
| 638 | } |
| 639 | } |
| 640 | var lastBody []byte |
| 641 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "1.2.3.4", &lastBody); err != nil { |
| 642 | t.Fatal(err) |
| 643 | } |
| 644 | body, _, err := s.Get(context.Background(), "control/edges/annie.json") |
| 645 | if err != nil { |
| 646 | t.Fatal(err) |
| 647 | } |
| 648 | var f struct { |
| 649 | Hosts map[string]struct { |
| 650 | Platform bool `json:"platform"` |
| 651 | } `json:"hosts"` |
| 652 | } |
| 653 | if err := json.Unmarshal(body, &f); err != nil { |
| 654 | t.Fatal(err) |
| 655 | } |
| 656 | for host, want := range map[string]bool{"app.acct.apps.test": true, "prov.acct.apps.test": true, "stray.acct.apps.test": false} { |
| 657 | if got := f.Hosts[host].Platform; got != want { |
| 658 | t.Errorf("%s platform = %v; want %v", host, got, want) |
| 659 | } |
| 660 | } |
| 661 | if _, ok := f.Hosts["pending.acct.apps.test"]; ok { |
| 662 | t.Error("a non-ready row is published regardless of class") |
| 663 | } |
| 664 | // The mark is always written, so a keyless host object can only mean a |
| 665 | // pre-mark publisher. |
| 666 | if !strings.Contains(string(body), `"stray.acct.apps.test":{"role":"home","platform":false}`) { |
| 667 | t.Errorf("mark not emitted for an unclassed row: %s", body) |
| 668 | } |
| 669 | } |
| 670 | |
| 671 | // #287: attached custom domains are published under their own names — |
| 672 | // only when their primary is served AND their DNS points here, marked as |
| 673 | // aliases, never platform-class, catch-alls never, and never over a real |
| 674 | // route. Detaching one republishes without it. |
| 675 | func TestEdgesFilePublishesPointingAliasesOfServedPrimaries(t *testing.T) { |
| 676 | withheldAliasSaid = sync.Map{} // once per PROCESS: isolate from -count>1 |
| 677 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 678 | if err := reg.Upsert(registry.Route{Host: "app.acct.apps.test", Kind: "service", Addr: "10.0.0.1:3000", Status: "ready", Platform: true}); err != nil { |
| 679 | t.Fatal(err) |
| 680 | } |
| 681 | if err := reg.Upsert(registry.Route{Host: "pending.acct.apps.test", Kind: "service", Addr: "10.0.0.1:3001", Status: "pending"}); err != nil { |
| 682 | t.Fatal(err) |
| 683 | } |
| 684 | if err := reg.Upsert(registry.Route{Host: "real.example.net", Kind: "service", Addr: "10.0.0.1:3002", Status: "ready", Platform: true}); err != nil { |
| 685 | t.Fatal(err) |
| 686 | } |
| 687 | for _, a := range []registry.RouteAlias{ |
| 688 | {Alias: "www.customer.com", Host: "app.acct.apps.test", Account: "acct", App: "app", DNSStatus: "pointing"}, |
| 689 | {Alias: "unpointed.customer.com", Host: "app.acct.apps.test", Account: "acct", App: "app", DNSStatus: "elsewhere"}, |
| 690 | {Alias: "fresh.customer.com", Host: "app.acct.apps.test", Account: "acct", App: "app"}, |
| 691 | {Alias: "*.customer.com", Host: "app.acct.apps.test", Account: "acct", App: "app", DNSStatus: "pointing"}, |
| 692 | {Alias: "orphan.example.org", Host: "pending.acct.apps.test", Account: "acct", App: "pending", DNSStatus: "pointing"}, |
| 693 | {Alias: "real.example.net", Host: "app.acct.apps.test", Account: "acct", App: "app", DNSStatus: "pointing"}, // shadowed |
| 694 | } { |
| 695 | if err := reg.UpsertAlias(a); err != nil { |
| 696 | t.Fatal(err) |
| 697 | } |
| 698 | } |
| 699 | type hostView struct { |
| 700 | Role string `json:"role"` |
| 701 | Platform bool `json:"platform"` |
| 702 | Alias bool `json:"alias"` |
| 703 | } |
| 704 | publish := func() map[string]hostView { |
| 705 | t.Helper() |
| 706 | var lastBody []byte |
| 707 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "1.2.3.4", &lastBody); err != nil { |
| 708 | t.Fatal(err) |
| 709 | } |
| 710 | body, _, err := s.Get(context.Background(), "control/edges/annie.json") |
| 711 | if err != nil { |
| 712 | t.Fatal(err) |
| 713 | } |
| 714 | var f struct { |
| 715 | Hosts map[string]hostView `json:"hosts"` |
| 716 | } |
| 717 | if err := json.Unmarshal(body, &f); err != nil { |
| 718 | t.Fatal(err) |
| 719 | } |
| 720 | return f.Hosts |
| 721 | } |
| 722 | var buf strings.Builder |
| 723 | log.SetOutput(&buf) |
| 724 | t.Cleanup(func() { log.SetOutput(os.Stderr) }) |
| 725 | hosts := publish() |
| 726 | h, ok := hosts["www.customer.com"] |
| 727 | if !ok || h.Role != "home" || !h.Alias || h.Platform { |
| 728 | t.Errorf("www.customer.com = %+v, %v; want home, alias, not platform", h, ok) |
| 729 | } |
| 730 | for _, absent := range []string{"unpointed.customer.com", "fresh.customer.com", "*.customer.com", "orphan.example.org"} { |
| 731 | if _, ok := hosts[absent]; ok { |
| 732 | t.Errorf("%s must not be published", absent) |
| 733 | } |
| 734 | } |
| 735 | if h := hosts["real.example.net"]; !h.Platform || h.Alias { |
| 736 | t.Errorf("an alias colliding with a real route displaced the route's entry: %+v", h) |
| 737 | } |
| 738 | // A withheld alias says so once per process, not once per pass. |
| 739 | publish() |
| 740 | publish() |
| 741 | if n := strings.Count(buf.String(), "unpointed.customer.com not published"); n != 1 { |
| 742 | t.Errorf("withheld line logged %d times over 3 passes; want 1:\n%s", n, buf.String()) |
| 743 | } |
| 744 | // Detach: gone on the next pass. |
| 745 | if err := reg.DeleteAlias("www.customer.com"); err != nil { |
| 746 | t.Fatal(err) |
| 747 | } |
| 748 | if _, ok := publish()["www.customer.com"]; ok { |
| 749 | t.Error("a detached alias was still published") |
| 750 | } |
| 751 | } |
| 752 | |
| 753 | // #305: delegated customer wildcard parents publish as "*.<parent>" in their |
| 754 | // own class — never platform, never over an existing key. |
| 755 | func TestEdgesFilePublishesDelegatedWildcardParents(t *testing.T) { |
| 756 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 757 | seedRoute(t, reg, "app.acct.apps.test") |
| 758 | if err := reg.Upsert(registry.Route{Host: "*.taken.example", Kind: "service", Addr: "10.0.0.1:3001", Status: "ready", Platform: true}); err != nil { |
| 759 | t.Fatal(err) |
| 760 | } |
| 761 | // customer.example has a served catch-all (Catchall claim) → dial half |
| 762 | // too; certonly.example is a Wildcard claim without Catchall → cert |
| 763 | // half only; pending.example's catch-all points at an unserved primary. |
| 764 | if err := reg.Upsert(registry.Route{Host: "pend.acct.apps.test", Kind: "service", Addr: "10.0.0.1:3002", Status: "pending"}); err != nil { |
| 765 | t.Fatal(err) |
| 766 | } |
| 767 | for _, a := range []registry.RouteAlias{ |
| 768 | {Alias: "*.customer.example", Host: "app.acct.apps.test", Account: "acct", App: "app"}, |
| 769 | {Alias: "*.pending.example", Host: "pend.acct.apps.test", Account: "acct", App: "pend"}, |
| 770 | } { |
| 771 | if err := reg.UpsertAlias(a); err != nil { |
| 772 | t.Fatal(err) |
| 773 | } |
| 774 | } |
| 775 | for _, w := range []registry.WildcardDomain{ |
| 776 | {Host: "customer.example", Account: "acct", App: "app"}, |
| 777 | {Host: "certonly.example", Account: "acct", App: "app"}, |
| 778 | {Host: "pending.example", Account: "acct", App: "pend"}, |
| 779 | {Host: "*.odd.example", Account: "acct", App: "app"}, // never written by the pass; refused here too |
| 780 | {Host: "taken.example", Account: "acct", App: "app"}, // a real route owns the key |
| 781 | } { |
| 782 | if err := reg.UpsertWildcardDomain(w); err != nil { |
| 783 | t.Fatal(err) |
| 784 | } |
| 785 | } |
| 786 | var lastBody []byte |
| 787 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "1.2.3.4", &lastBody); err != nil { |
| 788 | t.Fatal(err) |
| 789 | } |
| 790 | body, _, err := s.Get(context.Background(), "control/edges/annie.json") |
| 791 | if err != nil { |
| 792 | t.Fatal(err) |
| 793 | } |
| 794 | var f struct { |
| 795 | Hosts map[string]struct { |
| 796 | Role string `json:"role"` |
| 797 | Platform bool `json:"platform"` |
| 798 | Wildcard bool `json:"wildcard"` |
| 799 | Alias bool `json:"alias"` |
| 800 | } `json:"hosts"` |
| 801 | } |
| 802 | if err := json.Unmarshal(body, &f); err != nil { |
| 803 | t.Fatal(err) |
| 804 | } |
| 805 | // Marked alias BECAUSE the dial half is on: a catch-all we serve is a |
| 806 | // materialized claim, and DeploymentRoutes keys the whole "is this |
| 807 | // somebody's real route" question on that mark. See the end-to-end |
| 808 | // tripwire below for what an unmarked row does to the claim. |
| 809 | h, ok := f.Hosts["*.customer.example"] |
| 810 | if !ok || h.Role != "home" || !h.Wildcard || h.Platform || !h.Alias { |
| 811 | t.Fatalf("*.customer.example = %+v, %v; want home, wildcard, alias, not platform", h, ok) |
| 812 | } |
| 813 | for _, name := range []string{"*.certonly.example", "*.pending.example"} { |
| 814 | h, ok := f.Hosts[name] |
| 815 | if !ok || h.Role != "wildcard" || !h.Wildcard || h.Platform || h.Alias { |
| 816 | t.Errorf("%s = %+v, %v; want the cert half only (role wildcard, unmarked)", name, h, ok) |
| 817 | } |
| 818 | } |
| 819 | if _, ok := f.Hosts["*.*.odd.example"]; ok { |
| 820 | t.Error("a malformed wildcard_domains row was published") |
| 821 | } |
| 822 | // A real route owns the key: it keeps role and class, and the parent's |
| 823 | // class is added — the reader must still load the PEM everywhere. It |
| 824 | // stays UNMARKED, which is the other half of the mark's meaning: this |
| 825 | // hostname really is served as a route, so it must go on outranking any |
| 826 | // claim another box makes on the same name. |
| 827 | if h := f.Hosts["*.taken.example"]; !h.Platform || !h.Wildcard || h.Role != "home" || h.Alias { |
| 828 | t.Errorf("route-owned key = %+v; want platform AND wildcard, role home, unmarked", h) |
| 829 | } |
| 830 | } |
| 831 | |
| 832 | // The tripwire on a loop that ran in production: a served catch-all published |
| 833 | // without the alias mark is read back by DeploymentRoutes as evidence that |
| 834 | // some live box serves that hostname as a REAL route, so the next |
| 835 | // domains.ReconcileAliases pass refuses the very claim that produced the row |
| 836 | // and deletes the alias. Deleting it turns the dial half off, which drops the |
| 837 | // row, which lets the pass after that recreate it — and the name flaps |
| 838 | // on/off forever (*.woodstar.app, every 5-6 minutes, down about half the |
| 839 | // time). Neither half is wrong alone, so only an end-to-end check catches it: |
| 840 | // publish with the real publisher, then fold with the real reader. |
| 841 | func TestEdgesFileServedCatchAllIsNotItsOwnDeploymentRoute(t *testing.T) { |
| 842 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 843 | seedRoute(t, reg, "app.acct.apps.test") |
| 844 | if err := reg.UpsertAlias(registry.RouteAlias{ |
| 845 | Alias: "*.customer.example", Host: "app.acct.apps.test", Account: "acct", App: "app", |
| 846 | }); err != nil { |
| 847 | t.Fatal(err) |
| 848 | } |
| 849 | if err := reg.UpsertWildcardDomain(registry.WildcardDomain{ |
| 850 | Host: "customer.example", Account: "acct", App: "app", |
| 851 | }); err != nil { |
| 852 | t.Fatal(err) |
| 853 | } |
| 854 | var lastBody []byte |
| 855 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie"}, "1.2.3.4", &lastBody); err != nil { |
| 856 | t.Fatal(err) |
| 857 | } |
| 858 | now := time.Now() |
| 859 | if err := status.WriteBox(context.Background(), s, status.Box{Label: "annie", GeneratedAt: now}); err != nil { |
| 860 | t.Fatal(err) |
| 861 | } |
| 862 | |
| 863 | routes := edgesfile.DeploymentRoutes(context.Background(), s, now) |
| 864 | if routes["*.customer.example"] { |
| 865 | t.Errorf("routes = %v, want the served catch-all NOT counted as a real route — counting it makes the claim refuse itself every pass", routes) |
| 866 | } |
| 867 | // The join really happened: without this, an empty set would pass the |
| 868 | // assertion above for the wrong reason (a snapshot/label mismatch reads |
| 869 | // as "nothing known", which is permissive by contract). |
| 870 | if !routes["app.acct.apps.test"] { |
| 871 | t.Fatalf("routes = %v, want the box's real route counted — the fold did not join", routes) |
| 872 | } |
| 873 | } |
| 874 | |
| 875 | // The whole EIP fold rests on ONE identity: the `box` name this loop |
| 876 | // publishes under and the `label` the status snapshot is written under must |
| 877 | // be the same string, or edgesfile.EdgeIPs' liveness join drops every edge |
| 878 | // silently and the deployment's edge set degrades to env-only forever — a |
| 879 | // failure with no error, no log line and no red test. |
| 880 | // |
| 881 | // They are the same today because both resolve through metricsBoxLabel over |
| 882 | // the same machine id: edgesFileLoop calls it directly, and statusTick's |
| 883 | // snapshot label is status.Label(Probes.MachineID), which is exactly what |
| 884 | // metricsBoxLabel falls through to off a fleet box. (A fleet box never gets |
| 885 | // here: edgesFileLoop returns early on shouldSkipPublisherElection.) This |
| 886 | // test is the tripwire on that, end to end through the real fold. |
| 887 | func TestEdgesFileBoxNameJoinsTheStatusSnapshotLabel(t *testing.T) { |
| 888 | const machineID = "ip-10-0-0-7" |
| 889 | notAFleetBox := func(string) string { return "" } |
| 890 | |
| 891 | box := metricsBoxLabel(notAFleetBox, machineID) |
| 892 | snapshot := status.CollectBox(nil, status.Probes{MachineID: machineID}, time.Now()) |
| 893 | if box != snapshot.Label { |
| 894 | t.Fatalf("edges file box %q vs status snapshot label %q — the fold's liveness join needs one name", box, snapshot.Label) |
| 895 | } |
| 896 | |
| 897 | ctx := context.Background() |
| 898 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 899 | seedRoute(t, reg, "app.example.com") |
| 900 | var last []byte |
| 901 | if err := edgesFilePass(ctx, s, reg, edgesFile{Box: box, Region: "eu-west-1", Placement: "", DataBucket: "", DataBucketRegion: ""}, "203.0.113.7", &last); err != nil { |
| 902 | t.Fatal(err) |
| 903 | } |
| 904 | if err := status.WriteBox(ctx, s, snapshot); err != nil { |
| 905 | t.Fatal(err) |
| 906 | } |
| 907 | got := edgesfile.EdgeIPs(ctx, s, nil, time.Now()) |
| 908 | if len(got) != 1 || got[0] != "203.0.113.7" { |
| 909 | t.Fatalf("fold = %v, want this box's own EIP — the two writers disagree about its name", got) |
| 910 | } |
| 911 | } |
| 912 | |
| 913 | // The placement code (issue #329): CARLOS_REGION verbatim, additive between |
| 914 | // region and generation, omitted entirely when the box has none — the |
| 915 | // console's armed-placement gate stays dormant on a fleet that never |
| 916 | // publishes it, which is what makes roll order irrelevant. |
| 917 | func TestEdgesFileRawWireBodyPlacementField(t *testing.T) { |
| 918 | t.Parallel() |
| 919 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 920 | seedRoute(t, reg, "app.example.com") |
| 921 | var last []byte |
| 922 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "quiet-harbour", Region: "eu-west-2", Placement: "uk", DataBucket: "", DataBucketRegion: ""}, "203.0.113.10", &last); err != nil { |
| 923 | t.Fatal(err) |
| 924 | } |
| 925 | got, _, err := s.Get(context.Background(), "control/edges/quiet-harbour.json") |
| 926 | if err != nil { |
| 927 | t.Fatal(err) |
| 928 | } |
| 929 | var probe struct { |
| 930 | Placement string `json:"placement"` |
| 931 | Generation string `json:"generation"` |
| 932 | } |
| 933 | if err := json.Unmarshal(got, &probe); err != nil { |
| 934 | t.Fatal(err) |
| 935 | } |
| 936 | want := `{"v":1,"box":"quiet-harbour","eip":"203.0.113.10","region":"eu-west-2","placement":"uk","hibernation_configured":false,"serves_shared_pool":false,"accepts_inbound":false,"moves_protocol":0,"seeds_protocol":0,"generation":"` + |
| 937 | probe.Generation + `","hosts":{"app.example.com":{"role":"home","platform":false}}}` |
| 938 | if string(got) != want { |
| 939 | t.Fatalf("raw wire body (placement set) =\n%s\nwant\n%s", got, want) |
| 940 | } |
| 941 | |
| 942 | // Unset placement publishes no key at all. |
| 943 | s2, reg2 := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 944 | seedRoute(t, reg2, "app.example.com") |
| 945 | var last2 []byte |
| 946 | if err := edgesFilePass(context.Background(), s2, reg2, edgesFile{Box: "quiet-harbour", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "203.0.113.10", &last2); err != nil { |
| 947 | t.Fatal(err) |
| 948 | } |
| 949 | got2, _, err := s2.Get(context.Background(), "control/edges/quiet-harbour.json") |
| 950 | if err != nil { |
| 951 | t.Fatal(err) |
| 952 | } |
| 953 | if strings.Contains(string(got2), "placement") { |
| 954 | t.Fatalf("raw wire body with no CARLOS_REGION carries a placement key: %s", got2) |
| 955 | } |
| 956 | } |
| 957 | |
| 958 | // Data-bucket coordinates (instance-move spec): published verbatim from the |
| 959 | // hibernate env pair, absent when unset — the same additive rule as |
| 960 | // placement. |
| 961 | func TestEdgesFileRawWireBodyDataBucketFields(t *testing.T) { |
| 962 | t.Parallel() |
| 963 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 964 | seedRoute(t, reg, "app.example.com") |
| 965 | var last []byte |
| 966 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "quiet-harbour", Region: "eu-west-2", Placement: "uk", DataBucket: "carlos-tito-uk-1", DataBucketRegion: "eu-west-2"}, "203.0.113.10", &last); err != nil { |
| 967 | t.Fatal(err) |
| 968 | } |
| 969 | got, _, err := s.Get(context.Background(), "control/edges/quiet-harbour.json") |
| 970 | if err != nil { |
| 971 | t.Fatal(err) |
| 972 | } |
| 973 | var probe struct { |
| 974 | DataBucket string `json:"data_bucket"` |
| 975 | DataBucketRegion string `json:"data_bucket_region"` |
| 976 | } |
| 977 | if err := json.Unmarshal(got, &probe); err != nil { |
| 978 | t.Fatal(err) |
| 979 | } |
| 980 | if probe.DataBucket != "carlos-tito-uk-1" || probe.DataBucketRegion != "eu-west-2" { |
| 981 | t.Fatalf("data bucket fields = %+v", probe) |
| 982 | } |
| 983 | s2, reg2 := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 984 | seedRoute(t, reg2, "app.example.com") |
| 985 | var last2 []byte |
| 986 | if err := edgesFilePass(context.Background(), s2, reg2, edgesFile{Box: "quiet-harbour", Region: "", Placement: "", DataBucket: "", DataBucketRegion: ""}, "203.0.113.10", &last2); err != nil { |
| 987 | t.Fatal(err) |
| 988 | } |
| 989 | got2, _, err := s2.Get(context.Background(), "control/edges/quiet-harbour.json") |
| 990 | if err != nil { |
| 991 | t.Fatal(err) |
| 992 | } |
| 993 | if strings.Contains(string(got2), "data_bucket") { |
| 994 | t.Fatalf("unset hibernate env still published a data_bucket key: %s", got2) |
| 995 | } |
| 996 | } |
| 997 | |
| 998 | // --- #162: the replica placement's edges-file entries ---------------------- |
| 999 | |
| 1000 | // TestEdgesFilePublishesReplicaEntries pins the data plane of replica |
| 1001 | // placements: a served row minted from a role=replica record — the Replica |
| 1002 | // mark set, Account/App carried so the account apex derives — publishes its |
| 1003 | // own host and the account's platform apex under role "home" with the |
| 1004 | // replica mark set. A PLAIN row never gains the mark, whatever placement the |
| 1005 | // box publishes; the mark is the row's, not the box's. (Raw map decode, not |
| 1006 | // the struct: the test is about the WIRE, and a struct field would make an |
| 1007 | // omitempty typo invisible.) |
| 1008 | func TestEdgesFilePublishesReplicaEntries(t *testing.T) { |
| 1009 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 1010 | if err := reg.Upsert(registry.Route{ |
| 1011 | Host: "acme-uk.go.tito.io", Slug: "acme-uk", Kind: "service", Addr: "10.0.0.1:3000", |
| 1012 | Status: "ready", Account: "bab", App: "tito", Region: "uk", |
| 1013 | Replica: true, |
| 1014 | }); err != nil { |
| 1015 | t.Fatal(err) |
| 1016 | } |
| 1017 | // The home row, on the same box only for the test's convenience — its |
| 1018 | // entries must NOT carry the mark even when it sits here. |
| 1019 | if err := reg.Upsert(registry.Route{ |
| 1020 | Host: "acme.go.tito.io", Slug: "acme", Kind: "service", Addr: "10.0.0.2:3000", |
| 1021 | Status: "ready", Account: "bab", App: "tito", Region: "ie", |
| 1022 | }); err != nil { |
| 1023 | t.Fatal(err) |
| 1024 | } |
| 1025 | |
| 1026 | var lastBody []byte |
| 1027 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "ukbox", Placement: "uk"}, "1.2.3.4", &lastBody); err != nil { |
| 1028 | t.Fatal(err) |
| 1029 | } |
| 1030 | got, _, err := s.Get(context.Background(), "control/edges/ukbox.json") |
| 1031 | if err != nil { |
| 1032 | t.Fatal(err) |
| 1033 | } |
| 1034 | var raw struct { |
| 1035 | Hosts map[string]struct { |
| 1036 | Role string `json:"role"` |
| 1037 | Replica bool `json:"replica"` |
| 1038 | } `json:"hosts"` |
| 1039 | } |
| 1040 | if err := json.Unmarshal(got, &raw); err != nil { |
| 1041 | t.Fatal(err) |
| 1042 | } |
| 1043 | for _, name := range []string{"acme-uk.go.tito.io"} { |
| 1044 | h, ok := raw.Hosts[name] |
| 1045 | if !ok { |
| 1046 | t.Errorf("replica row published nothing at %s (hosts: %+v)", name, raw.Hosts) |
| 1047 | continue |
| 1048 | } |
| 1049 | if h.Role != "home" || !h.Replica { |
| 1050 | t.Errorf("%s = %+v, want role home with the replica mark", name, h) |
| 1051 | } |
| 1052 | } |
| 1053 | if h := raw.Hosts["acme.go.tito.io"]; h.Replica { |
| 1054 | t.Errorf("the home row's entry carries the replica mark: %+v — the mark is the ROW's, and the home row is not a replica", h) |
| 1055 | } |
| 1056 | // The account apex is deliberately NOT published: a name the replica box |
| 1057 | // has no local row for must not be claimed (an apex key here would let |
| 1058 | // foldHomes' first-sorted-box rule hand the apex dial target to a |
| 1059 | // replica box fleet-wide). |
| 1060 | if _, ok := raw.Hosts["tito.bab"]; ok { |
| 1061 | t.Errorf("replica row published the account apex tito.bab — a name this box has no local row for must not be claimed (hosts: %+v)", raw.Hosts) |
| 1062 | } |
| 1063 | } |
| 1064 | |
| 1065 | // TestEdgesFileReplicaMarkFollowsTheRow pins the fail-closed default: a |
| 1066 | // plain served row publishes exactly as before #162 — no replica key at all |
| 1067 | // on the wire — on a box with no placement published, with no placement, or |
| 1068 | // with any placement. The mark cannot leak from the box's identity into an |
| 1069 | // unrelated row. |
| 1070 | func TestEdgesFileReplicaMarkFollowsTheRow(t *testing.T) { |
| 1071 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 1072 | if err := reg.Upsert(registry.Route{ |
| 1073 | Host: "acme.go.tito.io", Kind: "service", Addr: "10.0.0.1:3000", |
| 1074 | Status: "ready", |
| 1075 | }); err != nil { |
| 1076 | t.Fatal(err) |
| 1077 | } |
| 1078 | for _, ident := range []edgesFile{{Box: "annie"}, {Box: "annie", Placement: "uk"}} { |
| 1079 | var lastBody []byte |
| 1080 | if err := edgesFilePass(context.Background(), s, reg, ident, "1.2.3.4", &lastBody); err != nil { |
| 1081 | t.Fatal(err) |
| 1082 | } |
| 1083 | got, _, err := s.Get(context.Background(), "control/edges/annie.json") |
| 1084 | if err != nil { |
| 1085 | t.Fatal(err) |
| 1086 | } |
| 1087 | var raw struct { |
| 1088 | Hosts map[string]struct { |
| 1089 | Replica bool `json:"replica"` |
| 1090 | } `json:"hosts"` |
| 1091 | } |
| 1092 | if err := json.Unmarshal(got, &raw); err != nil { |
| 1093 | t.Fatal(err) |
| 1094 | } |
| 1095 | if h, ok := raw.Hosts["acme.go.tito.io"]; ok && h.Replica { |
| 1096 | t.Errorf("hosts = %+v, want no replica mark on a plain row (box placement %q)", raw.Hosts, ident.Placement) |
| 1097 | } |
| 1098 | } |
| 1099 | } |
| 1100 | |
| 1101 | // ---- owners (#175 P7, the edge-local door) --------------------------------- |
| 1102 | |
| 1103 | // publishHosts runs one pass on box "annie" and reads its hosts back. |
| 1104 | func publishHosts(t *testing.T, reg *registry.Registry) map[string]edgeHost { |
| 1105 | t.Helper() |
| 1106 | s := release.NewDirStore(t.TempDir()) |
| 1107 | var last []byte |
| 1108 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie"}, "1.2.3.4", &last); err != nil { |
| 1109 | t.Fatal(err) |
| 1110 | } |
| 1111 | body, _, err := s.Get(context.Background(), "control/edges/annie.json") |
| 1112 | if err != nil { |
| 1113 | t.Fatal(err) |
| 1114 | } |
| 1115 | var f edgesFile |
| 1116 | if err := json.Unmarshal(body, &f); err != nil { |
| 1117 | t.Fatal(err) |
| 1118 | } |
| 1119 | return f.Hosts |
| 1120 | } |
| 1121 | |
| 1122 | func ownedRoute(t *testing.T, reg *registry.Registry, host, account, app string) { |
| 1123 | t.Helper() |
| 1124 | if err := reg.Upsert(registry.Route{Host: host, Kind: "service", Addr: "10.0.0.1:3000", Status: "ready", Account: account, App: app}); err != nil { |
| 1125 | t.Fatal(err) |
| 1126 | } |
| 1127 | } |
| 1128 | |
| 1129 | func wantOwner(t *testing.T, hosts map[string]edgeHost, host, account, app string) { |
| 1130 | t.Helper() |
| 1131 | h, ok := hosts[host] |
| 1132 | if !ok { |
| 1133 | t.Fatalf("%s not published: %+v", host, hosts) |
| 1134 | } |
| 1135 | if h.Account != account || h.App != app { |
| 1136 | t.Errorf("%s owner = %q/%q, want %q/%q", host, h.Account, h.App, account, app) |
| 1137 | } |
| 1138 | } |
| 1139 | |
| 1140 | // A home row publishes its own account and app, which the door checks |
| 1141 | // remotely. A row with no app has no owner a door key could name. |
| 1142 | func TestEdgesFilePublishesTheOwnerOfAHomeRow(t *testing.T) { |
| 1143 | reg := edgesTestRegistry(t) |
| 1144 | ownedRoute(t, reg, "acme.go.tito.io", "bae", "titogo") |
| 1145 | ownedRoute(t, reg, "*.go.tito.io", "bae", "titogo") |
| 1146 | seedRoute(t, reg, "appless.example.com") |
| 1147 | hosts := publishHosts(t, reg) |
| 1148 | wantOwner(t, hosts, "acme.go.tito.io", "bae", "titogo") |
| 1149 | wantOwner(t, hosts, "*.go.tito.io", "bae", "titogo") |
| 1150 | wantOwner(t, hosts, "appless.example.com", "", "") |
| 1151 | } |
| 1152 | |
| 1153 | // An alias is served by its primary's route, so it carries the primary's |
| 1154 | // owner, not the alias row's own columns. |
| 1155 | func TestEdgesFilePublishesAnAliasWithItsPrimarysOwner(t *testing.T) { |
| 1156 | withheldAliasSaid = sync.Map{} |
| 1157 | reg := edgesTestRegistry(t) |
| 1158 | ownedRoute(t, reg, "acme.go.tito.io", "bae", "titogo") |
| 1159 | if err := reg.UpsertAlias(registry.RouteAlias{Alias: "tickets.acme.com", Host: "acme.go.tito.io", |
| 1160 | Account: "bae", App: "somethingelse", DNSStatus: "pointing"}); err != nil { |
| 1161 | t.Fatal(err) |
| 1162 | } |
| 1163 | wantOwner(t, publishHosts(t, reg), "tickets.acme.com", "bae", "titogo") |
| 1164 | } |
| 1165 | |
| 1166 | // The served catch-all carries the wildcard claim's owner, when the primary |
| 1167 | // it serves agrees. |
| 1168 | func TestEdgesFilePublishesTheServedCatchAllsOwner(t *testing.T) { |
| 1169 | reg := edgesTestRegistry(t) |
| 1170 | ownedRoute(t, reg, "acme.go.tito.io", "bae", "titogo") |
| 1171 | if err := reg.UpsertAlias(registry.RouteAlias{Alias: "*.acme.com", Host: "acme.go.tito.io", Account: "bae", App: "titogo"}); err != nil { |
| 1172 | t.Fatal(err) |
| 1173 | } |
| 1174 | for _, w := range []registry.WildcardDomain{ |
| 1175 | {Host: "acme.com", Account: "bae", App: "titogo"}, |
| 1176 | {Host: "certonly.com", Account: "bae", App: "titogo"}, |
| 1177 | } { |
| 1178 | if err := reg.UpsertWildcardDomain(w); err != nil { |
| 1179 | t.Fatal(err) |
| 1180 | } |
| 1181 | } |
| 1182 | hosts := publishHosts(t, reg) |
| 1183 | wantOwner(t, hosts, "*.acme.com", "bae", "titogo") |
| 1184 | // The cert half dials nothing, so it names no owner. |
| 1185 | wantOwner(t, hosts, "*.certonly.com", "", "") |
| 1186 | } |
| 1187 | |
| 1188 | // Review focus 2: a hostname whose path prefixes reach two apps publishes no |
| 1189 | // owner, nor does its alias, so the door refuses it remotely. A hostname whose |
| 1190 | // prefixes all reach one app still publishes that owner. |
| 1191 | func TestEdgesFilePublishesNoOwnerForAMixedPathPrefixHostname(t *testing.T) { |
| 1192 | withheldAliasSaid = sync.Map{} |
| 1193 | reg := edgesTestRegistry(t) |
| 1194 | ownedRoute(t, reg, "ti.to", "bae", "titogo") |
| 1195 | ownedRoute(t, reg, "a.go.tito.io", "bae", "titogo") |
| 1196 | ownedRoute(t, reg, "b.classic.tito.io", "bae", "classic") |
| 1197 | ownedRoute(t, reg, "one.example", "bae", "titogo") |
| 1198 | for _, p := range []registry.RoutePath{ |
| 1199 | {Host: "ti.to", Prefix: "a", Target: "a.go.tito.io", Account: "bae", App: "titogo"}, |
| 1200 | {Host: "ti.to", Prefix: "b", Target: "b.classic.tito.io", Account: "bae", App: "titogo"}, |
| 1201 | {Host: "one.example", Prefix: "a", Target: "a.go.tito.io", Account: "bae", App: "titogo"}, |
| 1202 | } { |
| 1203 | if err := reg.UpsertPath(p); err != nil { |
| 1204 | t.Fatal(err) |
| 1205 | } |
| 1206 | } |
| 1207 | if err := reg.UpsertAlias(registry.RouteAlias{Alias: "tickets.example", Host: "ti.to", Account: "bae", App: "titogo", DNSStatus: "pointing"}); err != nil { |
| 1208 | t.Fatal(err) |
| 1209 | } |
| 1210 | hosts := publishHosts(t, reg) |
| 1211 | wantOwner(t, hosts, "ti.to", "", "") |
| 1212 | wantOwner(t, hosts, "tickets.example", "", "") |
| 1213 | wantOwner(t, hosts, "one.example", "bae", "titogo") |
| 1214 | wantOwner(t, hosts, "b.classic.tito.io", "bae", "classic") |
| 1215 | } |
| 1216 | |
| 1217 | // The two eligibility fields are NOT omitempty: absent means "a writer |
| 1218 | // that predates the field" (unknown), false means "no" (spec §6.1). |
| 1219 | func TestEdgesFileRawWireBodyEligibilityFields(t *testing.T) { |
| 1220 | t.Parallel() |
| 1221 | for _, want := range []bool{true, false} { |
| 1222 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 1223 | seedRoute(t, reg, "app.example.com") |
| 1224 | var last []byte |
| 1225 | ident := edgesFile{Box: "quiet-harbour", Region: "eu-west-2", Placement: "uk", |
| 1226 | ServesSharedPool: want, AcceptsInbound: want} |
| 1227 | if err := edgesFilePass(context.Background(), s, reg, ident, "203.0.113.10", &last); err != nil { |
| 1228 | t.Fatal(err) |
| 1229 | } |
| 1230 | got, _, err := s.Get(context.Background(), "control/edges/quiet-harbour.json") |
| 1231 | if err != nil { |
| 1232 | t.Fatal(err) |
| 1233 | } |
| 1234 | var raw map[string]any |
| 1235 | if err := json.Unmarshal(got, &raw); err != nil { |
| 1236 | t.Fatal(err) |
| 1237 | } |
| 1238 | for _, k := range []string{"serves_shared_pool", "accepts_inbound"} { |
| 1239 | v, present := raw[k] |
| 1240 | if !present || v != want { |
| 1241 | t.Fatalf("%s = %v (present %v), want %v always published", k, v, present, want) |
| 1242 | } |
| 1243 | } |
| 1244 | } |
| 1245 | } |
| 1246 | |
| 1247 | // What the loop publishes is the reconcile config's own answer, for every |
| 1248 | // spelling an operator might write. CARLOS_POOLS=",tito" parses to |
| 1249 | // ["tito"] (agent.go's parser drops empty elements), so it does NOT serve |
| 1250 | // the shared pool and must publish false (codex R3-3). |
| 1251 | func TestEdgesFileEligibilityFollowsTheReconcileConfig(t *testing.T) { |
| 1252 | cases := []struct { |
| 1253 | pools, inbound string |
| 1254 | shared, accept bool |
| 1255 | }{ |
| 1256 | {"", "", true, true}, |
| 1257 | {"tito", "", false, true}, |
| 1258 | {",tito", "", false, true}, |
| 1259 | {"", "none", true, false}, |
| 1260 | {"tito", "NONE", false, false}, |
| 1261 | } |
| 1262 | for _, c := range cases { |
| 1263 | getenv := func(k string) string { |
| 1264 | switch k { |
| 1265 | case "CARLOS_POOLS": |
| 1266 | return c.pools |
| 1267 | case "CARLOS_INBOUND": |
| 1268 | return c.inbound |
| 1269 | } |
| 1270 | return "" |
| 1271 | } |
| 1272 | shared, inbound := edgesFileEligibility(getenv) |
| 1273 | cfg := provisionTickConfig(getenv) |
| 1274 | wantShared, wantInbound := cfg.SharedInstanceEligibility() |
| 1275 | if shared != wantShared || inbound != wantInbound { |
| 1276 | t.Fatalf("%+v: published (%v, %v) but the reconciler says (%v, %v)", c, shared, inbound, wantShared, wantInbound) |
| 1277 | } |
| 1278 | if shared != c.shared || inbound != c.accept { |
| 1279 | t.Fatalf("%+v: got (%v, %v), want (%v, %v)", c, shared, inbound, c.shared, c.accept) |
| 1280 | } |
| 1281 | } |
| 1282 | } |
| 1283 | |
| 1284 | // R9: the box publishes the fingerprint of its secrets key's public half, |
| 1285 | // never the key; no key, or a key that does not parse, publishes nothing. |
| 1286 | func TestEdgesFileSecretsKeyFingerprint(t *testing.T) { |
| 1287 | t.Parallel() |
| 1288 | priv, err := ecdh.P256().NewPrivateKey(bytes.Repeat([]byte{1}, 32)) |
| 1289 | if err != nil { |
| 1290 | t.Fatal(err) |
| 1291 | } |
| 1292 | privB64 := base64.StdEncoding.EncodeToString(priv.Bytes()) |
| 1293 | pubB64 := base64.StdEncoding.EncodeToString(priv.PublicKey().Bytes()) |
| 1294 | env := func(v string) func(string) string { |
| 1295 | return func(k string) string { |
| 1296 | if k == "CARLOS_SECRETS_KEY" { |
| 1297 | return v |
| 1298 | } |
| 1299 | return "" |
| 1300 | } |
| 1301 | } |
| 1302 | want := appconfig.KeyFingerprint(priv.PublicKey()) |
| 1303 | if got := edgesFileSecretsFP(env(privB64)); got != want { |
| 1304 | t.Fatalf("with a key = %q, want %q", got, want) |
| 1305 | } |
| 1306 | for _, v := range []string{"", "not base64!", base64.StdEncoding.EncodeToString([]byte("short"))} { |
| 1307 | if got := edgesFileSecretsFP(env(v)); got != "" { |
| 1308 | t.Errorf("CARLOS_SECRETS_KEY=%q published %q, want nothing", v, got) |
| 1309 | } |
| 1310 | } |
| 1311 | |
| 1312 | for _, fp := range []string{want, ""} { |
| 1313 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 1314 | seedRoute(t, reg, "app.example.com") |
| 1315 | var last []byte |
| 1316 | ident := edgesFile{Box: "quiet-harbour", Region: "eu-west-2", Placement: "uk", SecretsKeyFP: fp} |
| 1317 | if err := edgesFilePass(context.Background(), s, reg, ident, "203.0.113.10", &last); err != nil { |
| 1318 | t.Fatal(err) |
| 1319 | } |
| 1320 | body, _, err := s.Get(context.Background(), "control/edges/quiet-harbour.json") |
| 1321 | if err != nil { |
| 1322 | t.Fatal(err) |
| 1323 | } |
| 1324 | has := strings.Contains(string(body), `"secrets_key_fp"`) |
| 1325 | if has != (fp != "") { |
| 1326 | t.Fatalf("fp %q: key present = %v in %s", fp, has, body) |
| 1327 | } |
| 1328 | for _, secret := range []string{privB64, pubB64} { |
| 1329 | if strings.Contains(string(body), secret) { |
| 1330 | t.Fatal("the edges file carries key material") |
| 1331 | } |
| 1332 | } |
| 1333 | } |
| 1334 | } |
| 1335 | |
| 1336 | func TestEdgesFileCapabilityFollowsTheFlag(t *testing.T) { |
| 1337 | env := func(v string) func(string) string { |
| 1338 | return func(k string) string { |
| 1339 | if k == "CARLOS_REGION_HA" { |
| 1340 | return v |
| 1341 | } |
| 1342 | return "" |
| 1343 | } |
| 1344 | } |
| 1345 | if got := edgesFileCapabilities(env("")); len(got) != 0 { |
| 1346 | t.Fatalf("unset: %v", got) |
| 1347 | } |
| 1348 | if got := edgesFileCapabilities(env("1")); len(got) != 1 || got[0] != "region-ha-1" { |
| 1349 | t.Fatalf("set: %v", got) |
| 1350 | } |
| 1351 | if got := edgesFileCapabilities(env("0")); len(got) != 0 { |
| 1352 | t.Fatalf("\"0\" is off: %v", got) |
| 1353 | } |
| 1354 | } |
| 1355 | |
| 1356 | func TestEdgesFileRawWireBodyCarriesBoxEvidence(t *testing.T) { |
| 1357 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 1358 | seedRoute(t, reg, "app.example.com") |
| 1359 | var last []byte |
| 1360 | ident := edgesFile{Box: "annie", BoxID: "i-0abc", Capabilities: []string{"region-ha-1"}, |
| 1361 | Headroom: &edgeHeadroom{MemFreeMiB: 2210, DiskFreeMiB: 31200, Awake: 14}} |
| 1362 | if err := edgesFilePass(context.Background(), s, reg, ident, "1.2.3.4", &last); err != nil { |
| 1363 | t.Fatal(err) |
| 1364 | } |
| 1365 | body, _, err := s.Get(context.Background(), "control/edges/annie.json") |
| 1366 | if err != nil { |
| 1367 | t.Fatal(err) |
| 1368 | } |
| 1369 | var raw map[string]json.RawMessage |
| 1370 | if err := json.Unmarshal(body, &raw); err != nil { |
| 1371 | t.Fatal(err) |
| 1372 | } |
| 1373 | for key, want := range map[string]string{ |
| 1374 | "box_id": `"i-0abc"`, |
| 1375 | "capabilities": `["region-ha-1"]`, |
| 1376 | "headroom": `{"mem_free_mib":2210,"disk_free_mib":31200,"awake":14}`, |
| 1377 | } { |
| 1378 | if string(raw[key]) != want { |
| 1379 | t.Errorf("%s = %s, want %s", key, raw[key], want) |
| 1380 | } |
| 1381 | } |
| 1382 | if _, ok := raw["draining"]; ok { |
| 1383 | t.Error("draining:false must be omitted") |
| 1384 | } |
| 1385 | } |
| 1386 | |
| 1387 | func TestEdgesFileOmitsBoxEvidenceWhenAbsent(t *testing.T) { |
| 1388 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 1389 | var last []byte |
| 1390 | if err := edgesFilePass(context.Background(), s, reg, edgesFile{Box: "annie"}, "1.2.3.4", &last); err != nil { |
| 1391 | t.Fatal(err) |
| 1392 | } |
| 1393 | body, _, _ := s.Get(context.Background(), "control/edges/annie.json") |
| 1394 | for _, key := range []string{"box_id", "capabilities", "draining", "headroom"} { |
| 1395 | if bytes.Contains(body, []byte(`"`+key+`"`)) { |
| 1396 | t.Errorf("%s present in %s", key, body) |
| 1397 | } |
| 1398 | } |
| 1399 | } |
| 1400 | |
| 1401 | func TestEdgesFileHeadroomAloneDoesNotRewriteTheFile(t *testing.T) { |
| 1402 | st := &putCountingStore{Store: release.NewDirStore(t.TempDir())} |
| 1403 | reg := edgesTestRegistry(t) |
| 1404 | seedRoute(t, reg, "app.example.com") |
| 1405 | var last []byte |
| 1406 | ident := edgesFile{Box: "annie", Headroom: &edgeHeadroom{MemFreeMiB: 100}} |
| 1407 | if err := edgesFilePass(context.Background(), st, reg, ident, "1.2.3.4", &last); err != nil { |
| 1408 | t.Fatal(err) |
| 1409 | } |
| 1410 | ident.Headroom = &edgeHeadroom{MemFreeMiB: 99} |
| 1411 | if err := edgesFilePass(context.Background(), st, reg, ident, "1.2.3.4", &last); err != nil { |
| 1412 | t.Fatal(err) |
| 1413 | } |
| 1414 | if st.puts != 1 { |
| 1415 | t.Fatalf("puts = %d, want 1: a headroom change alone is not a reason to write", st.puts) |
| 1416 | } |
| 1417 | } |
| 1418 | |
| 1419 | // The loop's timed pass refreshes headroom even when nothing else moved. |
| 1420 | func TestEdgesFileTimedPassRefreshesHeadroom(t *testing.T) { |
| 1421 | st := &putCountingStore{Store: release.NewDirStore(t.TempDir())} |
| 1422 | reg := edgesTestRegistry(t) |
| 1423 | seedRoute(t, reg, "app.example.com") |
| 1424 | var last []byte |
| 1425 | ident := edgesFile{Box: "annie", Headroom: &edgeHeadroom{MemFreeMiB: 100}} |
| 1426 | if err := edgesFileWrite(context.Background(), st, reg, ident, "1.2.3.4", &last, false); err != nil { |
| 1427 | t.Fatal(err) |
| 1428 | } |
| 1429 | ident.Headroom = &edgeHeadroom{MemFreeMiB: 42} |
| 1430 | if err := edgesFileWrite(context.Background(), st, reg, ident, "1.2.3.4", &last, true); err != nil { |
| 1431 | t.Fatal(err) |
| 1432 | } |
| 1433 | if st.puts != 2 { |
| 1434 | t.Fatalf("puts = %d, want 2: the timed pass must publish the new reading", st.puts) |
| 1435 | } |
| 1436 | body, _, err := st.Get(context.Background(), "control/edges/annie.json") |
| 1437 | if err != nil { |
| 1438 | t.Fatal(err) |
| 1439 | } |
| 1440 | var f edgesFile |
| 1441 | if err := json.Unmarshal(body, &f); err != nil || f.Headroom == nil || f.Headroom.MemFreeMiB != 42 { |
| 1442 | t.Fatalf("published headroom = %+v, err %v", f.Headroom, err) |
| 1443 | } |
| 1444 | // A box that publishes no headroom keeps today's behaviour exactly: |
| 1445 | // a timed pass with nothing changed writes nothing. |
| 1446 | plain := &putCountingStore{Store: release.NewDirStore(t.TempDir())} |
| 1447 | var plainLast []byte |
| 1448 | for i := 0; i < 2; i++ { |
| 1449 | if err := edgesFileWrite(context.Background(), plain, reg, edgesFile{Box: "betty"}, "1.2.3.4", &plainLast, true); err != nil { |
| 1450 | t.Fatal(err) |
| 1451 | } |
| 1452 | } |
| 1453 | if plain.puts != 1 { |
| 1454 | t.Fatalf("puts = %d, want 1 for a box with no headroom", plain.puts) |
| 1455 | } |
| 1456 | } |
| 1457 | |
| 1458 | // A restart must not rewrite an unchanged file just because the stored |
| 1459 | // copy carries a headroom reading. |
| 1460 | func TestEdgesFileSeededGateIgnoresStoredHeadroom(t *testing.T) { |
| 1461 | st := &putCountingStore{Store: release.NewDirStore(t.TempDir())} |
| 1462 | reg := edgesTestRegistry(t) |
| 1463 | seedRoute(t, reg, "app.example.com") |
| 1464 | var last []byte |
| 1465 | ident := edgesFile{Box: "annie", Headroom: &edgeHeadroom{MemFreeMiB: 100}} |
| 1466 | if err := edgesFilePass(context.Background(), st, reg, ident, "1.2.3.4", &last); err != nil { |
| 1467 | t.Fatal(err) |
| 1468 | } |
| 1469 | seeded := seedEdgesFileGate(context.Background(), st, "annie") |
| 1470 | ident.Headroom = &edgeHeadroom{MemFreeMiB: 7} |
| 1471 | if err := edgesFilePass(context.Background(), st, reg, ident, "1.2.3.4", &seeded); err != nil { |
| 1472 | t.Fatal(err) |
| 1473 | } |
| 1474 | if st.puts != 1 { |
| 1475 | t.Fatalf("puts = %d, want 1", st.puts) |
| 1476 | } |
| 1477 | } |
| 1478 | |
| 1479 | func TestEdgesFileStampIsOmittedWhenZero(t *testing.T) { |
| 1480 | b, _ := json.Marshal(edgeHost{Role: "home"}) |
| 1481 | if bytes.Contains(b, []byte("stamp")) { |
| 1482 | t.Fatalf("zero stamp on the wire: %s", b) |
| 1483 | } |
| 1484 | b, _ = json.Marshal(edgeHost{Role: "home", Stamp: 41}) |
| 1485 | if !bytes.Contains(b, []byte(`"stamp":41`)) { |
| 1486 | t.Fatalf("stamp missing: %s", b) |
| 1487 | } |
| 1488 | } |
| 1489 | |
| 1490 | // The move door (Task 4) admits a move only when every edges file of both |
| 1491 | // regions shows the box's armed floor and its verdict on carlos.prev, so |
| 1492 | // the writer publishes them from the box's own files at every write. |
| 1493 | func TestEdgesFilePublishesMovesProtocolFloorAndPrev(t *testing.T) { |
| 1494 | t.Parallel() |
| 1495 | type published struct { |
| 1496 | MovesProtocol *int `json:"moves_protocol"` |
| 1497 | Floor *string `json:"moves_floor_min_release"` |
| 1498 | Prev *string `json:"prev_release"` |
| 1499 | Allowed *string `json:"prev_allowed_for"` |
| 1500 | } |
| 1501 | publish := func(t *testing.T, dd, opt string) (published, string) { |
| 1502 | t.Helper() |
| 1503 | s, reg := release.NewDirStore(t.TempDir()), edgesTestRegistry(t) |
| 1504 | seedRoute(t, reg, "app.example.com") |
| 1505 | var last []byte |
| 1506 | ident := withMovesFields(edgesFile{Box: "quiet-harbour"}, dd, opt) |
| 1507 | if err := edgesFilePass(context.Background(), s, reg, ident, "203.0.113.10", &last); err != nil { |
| 1508 | t.Fatal(err) |
| 1509 | } |
| 1510 | body, _, err := s.Get(context.Background(), "control/edges/quiet-harbour.json") |
| 1511 | if err != nil { |
| 1512 | t.Fatal(err) |
| 1513 | } |
| 1514 | var p published |
| 1515 | if err := json.Unmarshal(body, &p); err != nil { |
| 1516 | t.Fatal(err) |
| 1517 | } |
| 1518 | return p, string(body) |
| 1519 | } |
| 1520 | |
| 1521 | // Nothing armed, no carlos.prev. |
| 1522 | dd, opt := t.TempDir(), t.TempDir() |
| 1523 | p, body := publish(t, dd, opt) |
| 1524 | if p.MovesProtocol == nil || *p.MovesProtocol != moves.ProtocolVersion || !strings.Contains(body, `"moves_protocol":1`) { |
| 1525 | t.Fatalf("moves_protocol missing: %s", body) |
| 1526 | } |
| 1527 | if p.Floor != nil || p.Allowed != nil || p.Prev == nil || *p.Prev != "none" { |
| 1528 | t.Fatalf("unarmed box with no carlos.prev published %s", body) |
| 1529 | } |
| 1530 | |
| 1531 | // A carlos.prev with no sidecar: its release is unknown, so absent. |
| 1532 | os.WriteFile(systemPrevPath(opt), []byte("#!/bin/sh\n"), 0o755) |
| 1533 | if p, body = publish(t, dd, opt); p.Prev != nil { |
| 1534 | t.Fatalf("carlos.prev without a sidecar published %s", body) |
| 1535 | } |
| 1536 | |
| 1537 | // Armed, judged: every value as the box's files hold it, one line each. |
| 1538 | if err := writeLocalFloor(dd, moves.Floor{Min: 1, MinRelease: "v0.30.0"}); err != nil { |
| 1539 | t.Fatal(err) |
| 1540 | } |
| 1541 | if err := writePrevVersion(opt, "v0.31.0"); err != nil { |
| 1542 | t.Fatal(err) |
| 1543 | } |
| 1544 | writeFileDurable(filepath.Join(opt, moves.PrevAllowedFile), []byte("v0.30.0\n"), 0o644) |
| 1545 | p, body = publish(t, dd, opt) |
| 1546 | if p.Floor == nil || *p.Floor != "v0.30.0" || p.Prev == nil || *p.Prev != "v0.31.0" || p.Allowed == nil || *p.Allowed != "v0.30.0" { |
| 1547 | t.Fatalf("armed box published %s", body) |
| 1548 | } |
| 1549 | |
| 1550 | // No verdict: absent, which the door reads as below the floor. |
| 1551 | os.Remove(filepath.Join(opt, moves.PrevAllowedFile)) |
| 1552 | if p, body = publish(t, dd, opt); p.Allowed != nil { |
| 1553 | t.Fatalf("no verdict, yet published %s", body) |
| 1554 | } |
| 1555 | } |
| 1556 | |
| 1557 | // Region HA spec 2 §7 reads a box's edges file as stale after 15 minutes |
| 1558 | // (boxchoice.MaxEdgesAge), and the gate clears Generation, so a quiet box |
| 1559 | // stays placeable only if its timed pass rewrites an UNCHANGED body. With |
| 1560 | // headroom published (region HA on) it does, every timed pass (at most six |
| 1561 | // minutes apart); an untimed pass with nothing changed does not; and with |
| 1562 | // region HA off (no headroom) nothing is ever rewritten. |
| 1563 | func TestEdgesFileQuietBoxStaysFreshOnTimedPassesOnlyWithRegionHA(t *testing.T) { |
| 1564 | ctx, reg := context.Background(), edgesTestRegistry(t) |
| 1565 | seedRoute(t, reg, "app.example.com") |
| 1566 | run := func(ident edgesFile) (untimedPuts, timedPuts int) { |
| 1567 | st := &putCountingStore{Store: release.NewDirStore(t.TempDir())} |
| 1568 | var last []byte |
| 1569 | if err := edgesFileWrite(ctx, st, reg, ident, "1.2.3.4", &last, false); err != nil { |
| 1570 | t.Fatal(err) |
| 1571 | } |
| 1572 | if err := edgesFileWrite(ctx, st, reg, ident, "1.2.3.4", &last, false); err != nil { |
| 1573 | t.Fatal(err) |
| 1574 | } |
| 1575 | untimedPuts = st.puts |
| 1576 | if err := edgesFileWrite(ctx, st, reg, ident, "1.2.3.4", &last, true); err != nil { |
| 1577 | t.Fatal(err) |
| 1578 | } |
| 1579 | return untimedPuts, st.puts |
| 1580 | } |
| 1581 | on := edgesFile{Box: "annie", Capabilities: []string{capRegionHA1}, |
| 1582 | Headroom: &edgeHeadroom{MemFreeMiB: 100, DiskFreeMiB: 100, Awake: 1, MemTotalMiB: 1000, DiskTotalMiB: 1000}} |
| 1583 | if untimed, timed := run(on); untimed != 1 || timed != 2 { |
| 1584 | t.Fatalf("region HA on: puts = %d untimed, %d after a timed pass; want 1, 2", untimed, timed) |
| 1585 | } |
| 1586 | if untimed, timed := run(edgesFile{Box: "annie"}); untimed != 1 || timed != 1 { |
| 1587 | t.Fatalf("region HA off: puts = %d untimed, %d after a timed pass; want 1, 1", untimed, timed) |
| 1588 | } |
| 1589 | } |
| 1590 | |
amadan