| 1 | package activator |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "errors" |
| 6 | "os" |
| 7 | "path/filepath" |
| 8 | "strings" |
| 9 | "testing" |
| 10 | "time" |
| 11 | |
| 12 | "github.com/carlosframework/platform/internal/instancekey" |
| 13 | "github.com/carlosframework/platform/internal/registry" |
| 14 | ) |
| 15 | |
| 16 | // bucketKey is ExecBackend.ConfiguredKey for a bucket whose configuration |
| 17 | // renders key ("" for none), counting its reads; down makes every read |
| 18 | // fail. |
| 19 | type bucketKey struct { |
| 20 | key string |
| 21 | down bool |
| 22 | reads int |
| 23 | } |
| 24 | |
| 25 | func (b *bucketKey) read(context.Context, registry.Route) (string, error) { |
| 26 | b.reads++ |
| 27 | if b.down { |
| 28 | return "", errors.New("bucket unreachable") |
| 29 | } |
| 30 | if b.key == "" { |
| 31 | return "", nil |
| 32 | } |
| 33 | return instancekey.Digest(b.key), nil |
| 34 | } |
| 35 | |
| 36 | // Final review I2 (Astra's P2), ruling 5: nothing is minted until the |
| 37 | // row's configuration is delivered. The files say what this box has |
| 38 | // delivered; the bucket says what the configuration sets. A key the bucket |
| 39 | // sets and the box has not delivered defers the mint; a bucket that cannot |
| 40 | // be read defers it the same way; a configuration that sets no key lets |
| 41 | // the mint proceed. The bucket is asked only when a mint is next: a key |
| 42 | // already delivered, in either file, costs no read. |
| 43 | func TestEnsureInstanceKeyAsksTheBucketBeforeItMints(t *testing.T) { |
| 44 | tok := "CARLOS_ADMIN_TOKEN=" + strings.Repeat("c", 64) + "\n" |
| 45 | for _, tc := range []struct { |
| 46 | name string |
| 47 | own, config string |
| 48 | bucket *bucketKey |
| 49 | want InstanceKeyOutcome |
| 50 | deferred bool |
| 51 | reads int |
| 52 | }{ |
| 53 | {"the configuration sets a key the box has not delivered: deferred", tok, "", &bucketKey{key: "key-A"}, InstanceKeyFailed, true, 1}, |
| 54 | {"the bucket cannot be read: deferred", tok, "", &bucketKey{down: true}, InstanceKeyFailed, true, 1}, |
| 55 | {"the configuration sets no key: minted", tok, "", &bucketKey{}, InstanceKeyMinted, false, 1}, |
| 56 | {"no configuration source: minted, nothing asked", tok, "", nil, InstanceKeyMinted, false, 0}, |
| 57 | {"the configured key is delivered: from config, nothing asked", tok, InstanceKeyVar + "=key-A\n", &bucketKey{key: "key-A"}, InstanceKeyFromConfig, false, 0}, |
| 58 | {"an own key: present, nothing asked", tok + InstanceKeyVar + "=" + strings.Repeat("K", 43) + "=\n", "", &bucketKey{key: "key-A"}, InstanceKeyPresent, false, 0}, |
| 59 | } { |
| 60 | t.Run(tc.name, func(t *testing.T) { |
| 61 | b := &ExecBackend{DataDir: t.TempDir()} |
| 62 | if tc.bucket != nil { |
| 63 | b.ConfiguredKey = tc.bucket.read |
| 64 | } |
| 65 | rt := registry.Route{Host: "shop.example.com", Kind: "instance"} |
| 66 | own := filepath.Join(b.DataDir, rt.Host+".env") |
| 67 | if err := os.WriteFile(own, []byte(tc.own), 0o600); err != nil { |
| 68 | t.Fatal(err) |
| 69 | } |
| 70 | if tc.config != "" { |
| 71 | if err := os.WriteFile(filepath.Join(b.DataDir, rt.Host+".config.env"), []byte(tc.config), 0o600); err != nil { |
| 72 | t.Fatal(err) |
| 73 | } |
| 74 | } |
| 75 | for _, call := range []string{"launch", "backfill"} { |
| 76 | var got InstanceKeyOutcome |
| 77 | var err error |
| 78 | if call == "launch" { |
| 79 | got, err = b.ensureInstanceKeyFor(t.Context(), rt, b.pathsFor(rt)) |
| 80 | } else { |
| 81 | if err := os.WriteFile(own, []byte(tc.own), 0o600); err != nil { |
| 82 | t.Fatal(err) |
| 83 | } |
| 84 | if tc.bucket != nil { |
| 85 | tc.bucket.reads = 0 |
| 86 | } |
| 87 | got, err = b.BackfillInstanceKey(t.Context(), rt) |
| 88 | } |
| 89 | if got != tc.want || errors.Is(err, ErrInstanceKeyDeferred) != tc.deferred || (err != nil) != tc.deferred { |
| 90 | t.Fatalf("%s = %v, %v; want %v, deferred %v", call, got, err, tc.want, tc.deferred) |
| 91 | } |
| 92 | raw, _ := os.ReadFile(own) |
| 93 | if minted := len(keyLines(raw)) > len(keyLines([]byte(tc.own))); minted != (tc.want == InstanceKeyMinted) { |
| 94 | t.Fatalf("%s: a key line was appended: %v, want %v", call, minted, tc.want == InstanceKeyMinted) |
| 95 | } |
| 96 | if tc.bucket != nil && tc.bucket.reads != tc.reads { |
| 97 | t.Fatalf("%s: the bucket was read %d times, want %d", call, tc.bucket.reads, tc.reads) |
| 98 | } |
| 99 | } |
| 100 | }) |
| 101 | } |
| 102 | } |
| 103 | |
| 104 | // The mint's bucket read is bounded by configReadTimeout and by the |
| 105 | // launch's own context, whichever ends first. |
| 106 | func TestEnsureInstanceKeyBoundsTheBucketRead(t *testing.T) { |
| 107 | b := &ExecBackend{DataDir: t.TempDir()} |
| 108 | var deadline time.Time |
| 109 | b.ConfiguredKey = func(ctx context.Context, _ registry.Route) (string, error) { |
| 110 | deadline, _ = ctx.Deadline() |
| 111 | return "", ctx.Err() |
| 112 | } |
| 113 | rt := registry.Route{Host: "shop.example.com", Kind: "instance"} |
| 114 | if err := os.WriteFile(filepath.Join(b.DataDir, rt.Host+".env"), []byte("A=1\n"), 0o600); err != nil { |
| 115 | t.Fatal(err) |
| 116 | } |
| 117 | if _, err := b.ensureInstanceKeyFor(context.Background(), rt, b.pathsFor(rt)); err != nil { |
| 118 | t.Fatal(err) |
| 119 | } |
| 120 | if until := time.Until(deadline); until <= 0 || until > configReadTimeout { |
| 121 | t.Fatalf("the read's deadline is %v away, want within %v", until, configReadTimeout) |
| 122 | } |
| 123 | if err := os.WriteFile(filepath.Join(b.DataDir, rt.Host+".env"), []byte("A=1\n"), 0o600); err != nil { |
| 124 | t.Fatal(err) // keyless again, so the backfill reaches the read |
| 125 | } |
| 126 | ctx, cancel := context.WithCancel(context.Background()) |
| 127 | cancel() |
| 128 | if _, err := b.BackfillInstanceKey(ctx, rt); !errors.Is(err, ErrInstanceKeyDeferred) || !errors.Is(err, context.Canceled) { |
| 129 | t.Fatalf("a cancelled read = %v, want deferred by the cancellation", err) |
| 130 | } |
| 131 | } |
| 132 | |
| 133 | // A first start whose configuration is still on its way, end to end: the |
| 134 | // start is deferred and no app runs; nothing is minted into <host>.env; |
| 135 | // once the config sync has delivered <host>.config.env with the configured |
| 136 | // key, the start runs and the app seals with that key, the configured one, |
| 137 | // never a minted one. |
| 138 | func TestFirstStartWaitsForTheConfiguredKey(t *testing.T) { |
| 139 | b := testExecBackend(t) |
| 140 | bucket := &bucketKey{key: "key-A"} |
| 141 | b.ConfiguredKey = bucket.read |
| 142 | rt := registry.Route{Host: "shop.example.com", Kind: "instance", Socket: noControlSocket(t)} |
| 143 | own := filepath.Join(b.DataDir, rt.Host+".env") |
| 144 | tok := "CARLOS_ADMIN_TOKEN=" + strings.Repeat("c", 64) + "\n" |
| 145 | if err := os.WriteFile(own, []byte(tok), 0o600); err != nil { |
| 146 | t.Fatal(err) |
| 147 | } |
| 148 | out := stageHelperApp(t, b, rt.Host, "v1") |
| 149 | ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second) |
| 150 | defer cancel() |
| 151 | proc, _, err := b.Start(ctx, rt) |
| 152 | if proc != nil { |
| 153 | proc.Stop(context.Background()) |
| 154 | } |
| 155 | if !errors.Is(err, ErrInstanceKeyDeferred) || !strings.Contains(err.Error(), rt.Host) { |
| 156 | t.Fatalf("Start before delivery = %v, want ErrInstanceKeyDeferred naming the host", err) |
| 157 | } |
| 158 | if got := outcomeOf(out); got != "" { |
| 159 | t.Fatalf("an app ran before delivery: %q", got) |
| 160 | } |
| 161 | if raw, _ := os.ReadFile(own); string(raw) != tok { |
| 162 | t.Fatalf("the own file changed before delivery: %d bytes", len(raw)) |
| 163 | } |
| 164 | // The config sync delivers the file. |
| 165 | if err := os.WriteFile(filepath.Join(b.DataDir, rt.Host+".config.env"), []byte(InstanceKeyVar+"=key-A\n"), 0o600); err != nil { |
| 166 | t.Fatal(err) |
| 167 | } |
| 168 | // The helper exits once it has recorded its outcome, so Start's health |
| 169 | // wait never sees a live socket: the outcome is the proof, as in the |
| 170 | // replica tests, and the error only must not be the deferral. |
| 171 | ctx2, cancel2 := recordedCtx(2*time.Second, out) // ctx was spent waiting for the first outcome |
| 172 | defer cancel2() |
| 173 | proc, _, err = b.Start(ctx2, rt) |
| 174 | if proc != nil { |
| 175 | defer proc.Stop(context.Background()) |
| 176 | } |
| 177 | if errors.Is(err, ErrInstanceKeyDeferred) { |
| 178 | t.Fatalf("Start after delivery still deferred: %v", err) |
| 179 | } |
| 180 | if got := outcomeOf(out); got != "key "+instancekey.Digest("key-A") { |
| 181 | t.Fatalf("the app's outcome = %q, want the configured key", got) |
| 182 | } |
| 183 | if raw, _ := os.ReadFile(own); string(raw) != tok { |
| 184 | t.Fatal("a key was minted beside the configured one") |
| 185 | } |
| 186 | } |
| 187 | |
| 188 | // The re-review's finding: the bucket read runs outside adminTokenMu, so a |
| 189 | // bucket that is slow or blocked for one row's first wake never holds the |
| 190 | // lock every other row's mint waits on. Host A's read blocks until the |
| 191 | // test releases it; host B mints meanwhile; A then mints too. |
| 192 | func TestASlowBucketReadForOneHostDoesNotBlockAnothersMint(t *testing.T) { |
| 193 | b := &ExecBackend{DataDir: t.TempDir()} |
| 194 | release := make(chan struct{}) |
| 195 | entered := make(chan struct{}) |
| 196 | b.ConfiguredKey = func(ctx context.Context, rt registry.Route) (string, error) { |
| 197 | if rt.Host == "a.example.com" { |
| 198 | close(entered) |
| 199 | select { |
| 200 | case <-release: |
| 201 | case <-ctx.Done(): |
| 202 | return "", ctx.Err() |
| 203 | } |
| 204 | } |
| 205 | return "", nil |
| 206 | } |
| 207 | a := registry.Route{Host: "a.example.com", Kind: "instance"} |
| 208 | bb := registry.Route{Host: "b.example.com", Kind: "instance"} |
| 209 | for _, rt := range []registry.Route{a, bb} { |
| 210 | if err := os.WriteFile(filepath.Join(b.DataDir, rt.Host+".env"), []byte("X=1\n"), 0o600); err != nil { |
| 211 | t.Fatal(err) |
| 212 | } |
| 213 | } |
| 214 | aDone := make(chan error, 1) |
| 215 | go func() { |
| 216 | _, err := b.ensureInstanceKeyFor(context.Background(), a, b.pathsFor(a)) |
| 217 | aDone <- err |
| 218 | }() |
| 219 | select { |
| 220 | case <-entered: // A is inside its bucket read before B starts |
| 221 | case <-time.After(2 * time.Second): |
| 222 | t.Fatal("host A never reached its bucket read") |
| 223 | } |
| 224 | bDone := make(chan error, 1) |
| 225 | go func() { |
| 226 | _, err := b.ensureInstanceKeyFor(context.Background(), bb, b.pathsFor(bb)) |
| 227 | bDone <- err |
| 228 | }() |
| 229 | select { |
| 230 | case err := <-bDone: |
| 231 | if err != nil { |
| 232 | t.Fatalf("host B's mint: %v", err) |
| 233 | } |
| 234 | case <-time.After(2 * time.Second): |
| 235 | t.Fatal("host B's mint waited on host A's bucket read") |
| 236 | } |
| 237 | select { |
| 238 | case <-aDone: |
| 239 | t.Fatal("host A minted before its bucket answered") |
| 240 | default: |
| 241 | } |
| 242 | close(release) |
| 243 | select { |
| 244 | case err := <-aDone: |
| 245 | if err != nil { |
| 246 | t.Fatalf("host A's mint after the bucket answered: %v", err) |
| 247 | } |
| 248 | case <-time.After(2 * time.Second): |
| 249 | t.Fatal("host A never minted") |
| 250 | } |
| 251 | for _, rt := range []registry.Route{a, bb} { |
| 252 | raw, _ := os.ReadFile(filepath.Join(b.DataDir, rt.Host+".env")) |
| 253 | if len(keyLines(raw)) != 1 { |
| 254 | t.Fatalf("%s: %d key lines, want one", rt.Host, len(keyLines(raw))) |
| 255 | } |
| 256 | } |
| 257 | } |
| 258 | |
| 259 | // A key that appears between the two phases is never overwritten: the |
| 260 | // bucket read happens outside the lock, so the files it looked at may |
| 261 | // change before the decision. The locked read decides. Here the bucket |
| 262 | // reader itself, standing in for a concurrent writer, puts a key into the |
| 263 | // own file, or delivers the configuration with one, before it answers. |
| 264 | func TestAKeyThatAppearsDuringTheBucketReadIsKept(t *testing.T) { |
| 265 | key := strings.Repeat("K", 43) + "=" |
| 266 | for _, tc := range []struct { |
| 267 | name string |
| 268 | plant func(dataDir, host string) |
| 269 | want InstanceKeyOutcome |
| 270 | }{ |
| 271 | {"into the own file", func(d, h string) { |
| 272 | f, err := os.OpenFile(filepath.Join(d, h+".env"), os.O_APPEND|os.O_WRONLY, 0) |
| 273 | if err != nil { |
| 274 | t.Fatal(err) |
| 275 | } |
| 276 | f.WriteString(InstanceKeyVar + "=" + key + "\n") |
| 277 | f.Close() |
| 278 | }, InstanceKeyPresent}, |
| 279 | {"delivered in the configuration", func(d, h string) { |
| 280 | if err := os.WriteFile(filepath.Join(d, h+".config.env"), []byte(InstanceKeyVar+"="+key+"\n"), 0o600); err != nil { |
| 281 | t.Fatal(err) |
| 282 | } |
| 283 | }, InstanceKeyFromConfig}, |
| 284 | } { |
| 285 | t.Run(tc.name, func(t *testing.T) { |
| 286 | b := &ExecBackend{DataDir: t.TempDir()} |
| 287 | rt := registry.Route{Host: "shop.example.com", Kind: "instance"} |
| 288 | own := filepath.Join(b.DataDir, rt.Host+".env") |
| 289 | if err := os.WriteFile(own, []byte("X=1\n"), 0o600); err != nil { |
| 290 | t.Fatal(err) |
| 291 | } |
| 292 | reads := 0 |
| 293 | b.ConfiguredKey = func(context.Context, registry.Route) (string, error) { |
| 294 | reads++ |
| 295 | tc.plant(b.DataDir, rt.Host) |
| 296 | return "", nil // the bucket says: no configured key, mint away |
| 297 | } |
| 298 | got, err := b.ensureInstanceKeyFor(context.Background(), rt, b.pathsFor(rt)) |
| 299 | if err != nil || got != tc.want || reads != 1 { |
| 300 | t.Fatalf("ensure = %v, %v after %d reads; want %v", got, err, reads, tc.want) |
| 301 | } |
| 302 | raw, _ := os.ReadFile(own) |
| 303 | ownKeys := keyLines(raw) |
| 304 | if tc.want == InstanceKeyPresent && (len(ownKeys) != 1 || ownKeys[0] != key) { |
| 305 | t.Fatalf("own file keys %d, want the one that appeared, kept", len(ownKeys)) |
| 306 | } |
| 307 | if tc.want == InstanceKeyFromConfig && len(ownKeys) != 0 { |
| 308 | t.Fatal("a key was minted beside the one the configuration delivered") |
| 309 | } |
| 310 | }) |
| 311 | } |
| 312 | } |
| 313 | |
| 314 | // The other side of the two phases: a mint that looks next only under the |
| 315 | // lock has no bucket verdict, and defers rather than mint on nothing. The |
| 316 | // first read saw a key in the own file (or, on the backfill, no file at |
| 317 | // all), so the bucket was not asked; by the locked read the key is gone |
| 318 | // (or a keyless file has appeared). Nothing is minted, the bucket is |
| 319 | // still unasked, and the next attempt asks it. |
| 320 | func TestAMintThatLooksNextOnlyUnderTheLockDefers(t *testing.T) { |
| 321 | key := strings.Repeat("K", 43) + "=" |
| 322 | for _, tc := range []struct { |
| 323 | name string |
| 324 | before string // the own file at the first read; "" is no file |
| 325 | after string // the own file at the locked read |
| 326 | create bool |
| 327 | }{ |
| 328 | {"a key seen by the first read is gone", "X=1\n" + InstanceKeyVar + "=" + key + "\n", "X=1\n", true}, |
| 329 | {"a keyless file appears after the backfill's first read", "", "X=1\n", false}, |
| 330 | } { |
| 331 | t.Run(tc.name, func(t *testing.T) { |
| 332 | b := &ExecBackend{DataDir: t.TempDir()} |
| 333 | bucket := &bucketKey{} |
| 334 | b.ConfiguredKey = bucket.read |
| 335 | rt := registry.Route{Host: "shop.example.com", Kind: "instance"} |
| 336 | own := filepath.Join(b.DataDir, rt.Host+".env") |
| 337 | if tc.before != "" { |
| 338 | if err := os.WriteFile(own, []byte(tc.before), 0o600); err != nil { |
| 339 | t.Fatal(err) |
| 340 | } |
| 341 | } |
| 342 | instanceKeyHooks.betweenPhases = func(string) { |
| 343 | if err := os.WriteFile(own, []byte(tc.after), 0o600); err != nil { |
| 344 | t.Error(err) |
| 345 | } |
| 346 | } |
| 347 | t.Cleanup(func() { instanceKeyHooks.betweenPhases = nil }) |
| 348 | got, err := b.ensureInstanceKeyAt(context.Background(), rt, b.pathsFor(rt), tc.create, "minted") |
| 349 | if got != InstanceKeyFailed || !errors.Is(err, ErrInstanceKeyDeferred) { |
| 350 | t.Fatalf("ensure = %v, %v; want deferred", got, err) |
| 351 | } |
| 352 | if bucket.reads != 0 { |
| 353 | t.Fatalf("the bucket was read %d times, want 0", bucket.reads) |
| 354 | } |
| 355 | raw, _ := os.ReadFile(own) |
| 356 | if string(raw) != tc.after { |
| 357 | t.Fatalf("own file changed by the deferred ensure (%d bytes, want %d)", len(raw), len(tc.after)) |
| 358 | } |
| 359 | }) |
| 360 | } |
| 361 | } |
| 362 | |
amadan